POPULARITY
Categories
536 episodes with Python
461 episodes with Python
319 episodes with Python
245 episodes with Python
211 episodes with Python
138 episodes with Python
112 episodes with Python
89 episodes with Python
294 episodes with Python
86 episodes with Python
195 episodes with Python
131 episodes with Python
57 episodes with Python
357 episodes with Python
67 episodes with Python
55 episodes with Python
66 episodes with Python
88 episodes with Python
55 episodes with Python
88 episodes with Python
58 episodes with Python
108 episodes with Python
101 episodes with Python
49 episodes with Python
53 episodes with Python
63 episodes with Python
35 episodes with Python
34 episodes with Python
28 episodes with Python
35 episodes with Python
60 episodes with Python
29 episodes with Python
46 episodes with Python
25 episodes with Python
23 episodes with Python
25 episodes with Python
18 episodes with Python
35 episodes with Python
31 episodes with Python
22 episodes with Python
27 episodes with Python
49 episodes with Python
25 episodes with Python
20 episodes with Python
28 episodes with Python
16 episodes with Python
17 episodes with Python
17 episodes with Python
21 episodes with Python
32 episodes with Python
47 episodes with Python
19 episodes with Python
16 episodes with Python
25 episodes with Python
BONUS: When AI Decisions Go Wrong at Scale—And How to Prevent It We've spent years asking what AI can do. But the next frontier isn't more capability—it's something far less glamorous and far more dangerous if we get it wrong. In this episode, Ran Aroussi shares why observability, transparency, and governance may be the difference between AI that empowers humans and AI that quietly drifts out of alignment. The Gap Between Demos and Deployable Systems "I've noticed that I watched well-designed agents make perfectly reasonable decisions based on their training, but in a context where the decision was catastrophically wrong. And there was really no way of knowing what had happened until the damage was already there." Ran's journey from building algorithmic trading systems to creating MUXI, an open framework for production-ready AI agents, revealed a fundamental truth: the skills needed to build impressive AI demos are completely different from those needed to deploy reliable systems at scale. Coming from the EdTech space where he handled billions of ad impressions daily and over a million concurrent users, Ran brings a perspective shaped by real-world production demands. The moment of realization came when he saw that the non-deterministic nature of AI meant that traditional software engineering approaches simply don't apply. While traditional bugs are reproducible, AI systems can produce different results from identical inputs—and that changes everything about how we need to approach deployment. Why Leaders Misunderstand Production AI "When you chat with ChatGPT, you go there and it pretty much works all the time for you. But when you deploy a system in production, you have users with unimaginable different use cases, different problems, and different ways of phrasing themselves." The biggest misconception leaders have is assuming that because AI works well in their personal testing, it will work equally well at scale. When you test AI with your own biases and limited imagination for scenarios, you're essentially seeing a curated experience. Real users bring infinite variation: non-native English speakers constructing sentences differently, unexpected use cases, and edge cases no one anticipated. The input space for AI systems is practically infinite because it's language-based, making comprehensive testing impossible. Multi-Layered Protection for Production AI "You have to put in deterministic filters between the AI and what you get back to the user." Ran outlines a comprehensive approach to protecting AI systems in production: Model version locking: Just as you wouldn't randomly upgrade Python versions without testing, lock your AI model versions to ensure consistent behavior Guardrails in prompts: Set clear boundaries about what the AI should never do or share Deterministic filters: Language firewalls that catch personal information, harmful content, or unexpected outputs before they reach users Comprehensive logging: Detailed traces of every decision, tool call, and data flow for debugging and pattern detection The key insight is that these layers must work together—no single approach provides sufficient protection for production systems. Observability in Agentic Workflows "With agentic AI, you have decision-making, task decomposition, tools that it decided to call, and what data to pass to them. So there's a lot of things that you should at least be able to trace back." Observability for agentic systems is fundamentally different from traditional LLM observability. When a user asks "What do I have to do today?", the system must determine who is asking, which tools are relevant to their role, what their preferences are, and how to format the response. Each user triggers a completely different dynamic workflow. Ran emphasizes the need for multi-layered access to observability data: engineers need full debugging access with appropriate security clearances, while managers need topic-level views without personal information. The goal is building a knowledge graph of interactions that allows pattern detection and continuous improvement. Governance as Human-AI Partnership "Governance isn't about control—it's about keeping people in the loop so AI amplifies, not replaces, human judgment." The most powerful reframing in this conversation is viewing governance not as red tape but as a partnership model. Some actions—like answering support tickets—can be fully automated with occasional human review. Others—like approving million-dollar financial transfers—require human confirmation before execution. The key is designing systems where AI can do the preparation work while humans retain decision authority at critical checkpoints. This mirrors how we build trust with human colleagues: through repeated successful interactions over time, gradually expanding autonomy as confidence grows. Building Trust Through Incremental Autonomy "Working with AI is like working with a new colleague that will back you up during your vacation. You probably don't know this person for a month. You probably know them for years. The first time you went on vacation, they had 10 calls with you, and then slowly it got to 'I'm only gonna call you if it's really urgent.'" The path to trusting AI systems mirrors how we build trust with human colleagues. You don't immediately hand over complete control—you start with frequent check-ins, observe performance, and gradually expand autonomy as confidence builds. This means starting with heavy human-in-the-loop interaction and systematically reducing oversight as the system proves reliable. The goal is reaching a state where you can confidently say "you don't have to ask permission before you do X, but I still want to approve every Y." In this episode, we refer to Thinking in Systems by Donella Meadows, Designing Machine Learning Systems by Chip Huyen, and Build a Large Language Model (From Scratch) by Sebastian Raschka. About Ran Aroussi Ran Aroussi is the founder of MUXI, an open framework for production-ready AI agents. He is also the co-creator of yfinance (with 10 million downloads monthly) and founder of Tradologics and Automaze. Ran is the author of the forthcoming book Production-Grade Agentic AI: From Brittle Workflows to Deployable Autonomous Systems, also available at productionaibook.com. You can connect with Ran Aroussi on LinkedIn.
This episode covers multiple active threats and security changes. It warns of an actively exploited critical BeyondTrust remote access vulnerability (CVE-2026-1731, CVSS 9.9) enabling pre-authentication remote code execution in Remote Support and Privileged Remote Access, noting SaaS was patched while on-prem deployments require urgent manual updates and may already be compromised. Microsoft details an evolution of the ClickFix social engineering technique where victims are tricked into running NSLookup commands that use attacker-controlled DNS responses as a malware staging channel, leading to payload delivery (including a Python-based RAT) and persistence via startup shortcuts, alongside increased Lumma Stealer activity. Cybersecurity Today would like to thank Meter for their support in bringing you this podcast. Meter delivers a complete networking stack, wired, wireless and cellular in one integrated solution that's built for performance and scale. You can find them at Meter.com/cst Researchers also report Mac-focused campaigns abusing AI-generated content and malicious search ads to push copy-paste terminal commands that install an info stealer (MaxSync) targeting Keychain, browsers, and crypto wallets. T The show describes fake recruiter campaigns targeting developers with coding tests containing malicious dependencies on repositories like NPM and PyPI, linked to the "Gala" operation and nearly 200 packages. Finally, it reviews NPM's authentication overhaul after a supply-chain worm incident—revoking classic long-lived tokens, moving to short-lived session credentials, encouraging MFA and OIDC trusted publishing—while noting remaining risks such as MFA phishing, non-mandatory MFA for unpublish, and the continued ability to create long-lived tokens. 00:00 Sponsor: Meter + Today's Cybersecurity Headlines 00:48 Urgent Patch: BeyondTrust Remote Access RCE (CVE-2026-1731) Actively Exploited 02:45 ClickFix Evolves: DNS Lookups (nslookup) Used as Malware Staging 04:34 Mac Malware via AI Search Results: Fake Terminal Commands Deliver Info-Stealer 06:08 Fake Recruiters, Real Malware: Coding Tests Poison Dev Environments 07:19 NPM Security Overhaul After Supply-Chain Worm—What's Better, What Still Risks 09:11 Wrap-Up, Thanks, and Sponsor Message
In MobileViews 597, recorded on February 15, 2026, Jon Westfall and I noted the upcoming the Lunar New Year while tackling the frustrations of modern tech ecosystems. I kicked things off with a double-header rant: first, my recurring battle with leaking alkaline batteries in my mouse and other devices, and second, Microsoft's decision to force Clipchamp (a video editor) users to store massive video files on OneDrive. With my upload speeds maxing out at 25 megabits, uploading gigabyte-sized files is simply unworkable, so I've officially pivoted to the open-source video editor Shotcut. We also explored the "bane of existence" for educators: the limitations of Chromebooks. Jon shared his struggles with students who, having grown up in managed K-12 Chrome environments, often struggle with standard file permissions and workflows when transitioning to college and professional platforms. Jon detailed his upgrade to the Backbone Pro gaming controller—praising its integrated battery and Bluetooth versatility—while looking forward to a future M5 Mac Mini to handle local LLM heavy lifting. I'm personally keeping an eye on rumors of an affordable A18 Pro-based MacBook that Jon noted could potentially disrupt the education sector. Between my nostalgia for coding in a 208-byte space on an Apple II and Jon's modern Python toolkit involving pyenv and PyInstaller, we emphasize that efficiency must remain a priority, even as software becomes more bloated. Whether it's navigating the "AI search" changes in Google Photos or finding ways around "vibe coding" errors, we're still looking for tech that just works.
Today we have Ziv Mador, VP of Security Research from LevelBlue SpiderLabs discussing their work on "SpiderLabs IDs New Banking Trojan Distributed Through WhatsApp." Researchers at LevelBlue SpiderLabs have identified a new Brazilian banking Trojan dubbed Eternidade Stealer, spread through WhatsApp hijacking and social engineering campaigns that use a Python-based worm to steal contacts and distribute malicious MSI installers. The Delphi-compiled malware targets Brazilian victims, profiles infected systems, dynamically retrieves its command-and-control server via IMAP email, and deploys banking overlays to harvest credentials from financial institutions and cryptocurrency platforms. The campaign reflects the continued evolution of Brazil's cybercrime ecosystem, combining WhatsApp propagation, geofencing, encrypted C2 communications, and process injection to maintain stealth and persistence. The research can be found here: SpiderLabs IDs New Banking Trojan Distributed Through WhatsApp Learn more about your ad choices. Visit megaphone.fm/adchoices
Today we have Ziv Mador, VP of Security Research from LevelBlue SpiderLabs discussing their work on "SpiderLabs IDs New Banking Trojan Distributed Through WhatsApp." Researchers at LevelBlue SpiderLabs have identified a new Brazilian banking Trojan dubbed Eternidade Stealer, spread through WhatsApp hijacking and social engineering campaigns that use a Python-based worm to steal contacts and distribute malicious MSI installers. The Delphi-compiled malware targets Brazilian victims, profiles infected systems, dynamically retrieves its command-and-control server via IMAP email, and deploys banking overlays to harvest credentials from financial institutions and cryptocurrency platforms. The campaign reflects the continued evolution of Brazil's cybercrime ecosystem, combining WhatsApp propagation, geofencing, encrypted C2 communications, and process injection to maintain stealth and persistence. The research can be found here: SpiderLabs IDs New Banking Trojan Distributed Through WhatsApp Learn more about your ad choices. Visit megaphone.fm/adchoices
Our ol' friend, Brett Cannon, is back to talk all things Python. But first! Star Wars, Machete Order, Lost, Babylon 5, Game of Thrones, Murderbot, Ted Lasso, Project Hail Mary, David Attenborough, perpetual voice rights, and the AI uncanny valley.
Would you like to learn how to work with LLMs locally on your own computer? How do you integrate your Python projects with a local model? Christopher Trudeau is back on the show this week with another batch of PyCoder's Weekly articles and projects.
Vincent Warmerdam is a Founding Engineer at marimo, working on reinventing Python notebooks as reactive, reproducible, interactive, and Git-friendly environments for data workflows and AI prototyping. He helps build the core marimo notebook platform, pushing its reactive execution model, UI interactivity, and integration with modern development and AI tooling so that notebooks behave like dependable, shareable programs and apps rather than error-prone scratchpads.Join the Community: https://go.mlops.community/YTJoinInGet the newsletter: https://go.mlops.community/YTNewsletterMLOps GPU Guide: https://go.mlops.community/gpuguide// AbstractVincent Warmerdam joins Demetrios fresh off marimo's acquisition by Weights & Biases—and makes a bold claim: notebooks as we know them are outdated.They talk Molab (GPU-backed, cloud-hosted notebooks), LLMs that don't just chat but actually fix your SQL and debug your code, and why most data folks are consuming tools instead of experimenting. Vincent argues we should stop treating notebooks like static scratchpads and start treating them like dynamic apps powered by AI.It's a conversation about rethinking workflows, reclaiming creativity, and not outsourcing your brain to the model.// BioVincent is a senior data professional who worked as an engineer, researcher, team lead, and educator in the past. You might know him from tech talks with an attempt to defend common sense over hype in the data space. He is especially interested in understanding algorithmic systems so that one may prevent failure. As such, he has always had a preference to keep calm and check the dataset before flowing tonnes of tensors. He currently works at marimo, where he spends his time rethinking everything related to Python notebooks.// Related LinksWebsite: https://marimo.io/Coding Agent Conference: https://luma.com/codingagentsHyperbolic GPU Cloud: app.hyperbolic.ai~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExploreJoin our Slack community [https://go.mlops.community/slack]Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)] Sign up for the next meetup: [https://go.mlops.community/register]MLOps Swag/Merch: [https://shop.mlops.community/]MLOps GPU Guide: https://go.mlops.community/gpuguideConnect with Demetrios on LinkedIn: /dpbrinkmConnect with Vincent on LinkedIn: /vincentwarmerdam/Timestamps:[00:00] Context in Notebooks[00:24] Acquisition and Team Continuity[04:43] Coding Agent Conference Announcement![05:56] Hyperbolic GPU Cloud Ad[06:54] marimo and W&B Synergies[09:31] marimo Cloud Code Support[12:59] Hardest Code to Generate[16:22] Trough of Disillusionment[20:38] Agent Interaction in Notebooks[25:41] Wrap up
Our ol' friend, Brett Cannon, is back to talk all things Python. But first! Star Wars, Machete Order, Lost, Babylon 5, Game of Thrones, Murderbot, Ted Lasso, Project Hail Mary, David Attenborough, perpetual voice rights, and the AI uncanny valley.
This interview was recorded for the GOTO Book Club.http://gotopia.tech/bookclubCheck out more here:https://gotopia.tech/episodes/418Sam Keen - Founder & Researcher at AlteredCraft & Author of "Clean Architecture with Python"Max Kirchoff - CTO at Ginko & Multidisciplinary Technologist & CreativeRESOURCESSamhttps://bsky.app/profile/samkeen.bsky.socialhttps://x.com/samkeenhttps://github.com/samkeenhttps://www.linkedin.com/in/samkeenhttps://samkeen.devMaxhttps://x.com/ProductNihilisthttps://github.com/maxkirchoffhttps://www.linkedin.com/in/maxkirchoffhttps://maxkirchoff.comLinkshttps://www.heyginko.comhttps://martinfowler.com/bliki/TestPyramid.htmlDESCRIPTIONMax Kirchoff interviews Sam Keen about his book "Clean Architecture with Python". Sam, a software developer with 30 years of experience spanning companies from startups to AWS, shares his approach to applying clean architecture principles with Python while maintaining the language's pragmatic nature.The conversation explores the balance between architectural rigor and practical development, the critical relationship between architecture and testability, and how clean architecture principles can enhance AI-assisted coding workflows. Sam emphasizes that clean architecture isn't an all-or-nothing approach but a set of principles that developers can adapt to their context, with the core value lying in thoughtful dependency management and clear domain modeling.RECOMMENDED BOOKSSam Keen • Clean Architecture with Python • https://amzn.to/4pBT5g0Fabrizio Romano & Heinrich Kruger • Learn Python Programming • https://amzn.to/4myLBItUncle Bob • Clean Code • https://amzn.to/3soPO6kUncle Bob • Clean Architecture • https://amzn.to/3x0gjBQEric Evans • Domain-Driven Design • https://amzn.to/3tnGhwmNaomi Ceder • The Quick Python Book • https://amzn.to/3zwdDOaLuciano Ramalho • Fluent Python • https://amzn.to/3oSw2jeDavid Beazley • Python Distilled (Developer's Library) • https://amzn.to/3QjNBEvSaleem Siddiqui • Learning Test-Driven Development • https://amzn.to/35OMb3nMaciej «MJ» Jedrzejewski • Master Software Architecture • https://leanpub.com/master-software-architectureBlueskyTwitterInstagramLinkedInFacebookCHANNEL MEMBERSHIP BONUSJoin this channel to get early access to videos & other perks:https://www.youtube.com/channel/UCs_tLP3AiwYKwdUHpltJPuA/joinLooking for a unique learning experience?Attend the next GOTO conference near you! Get your ticket: gotopia.techSUBSCRIBE TO OUR YOUTUBE CHANNEL - new videos posted daily!
Cesare e Marco esplorano il concetto di GitOps, una metodologia che si basa su Kubernetes per gestire e automatizzare l'infrastruttura IT. Cesare spiega i principi fondamentali di GitOps, le differenze rispetto al DevOps tradizionale e fornisce esempi pratici di come GitOps possa essere implementato utilizzando strumenti come ArgoCD. La discussione si concentra anche sull'importanza della gestione dei permessi e della tracciabilità delle modifiche nel contesto di GitOps.Manifesto GitOps Principles.
In this high-energy episode of the Qubit Value Podcast, the hosts dive into the transformative release of Qiskit version 2.3, marking a bold leap into the era of quantum-centric supercomputing. Recorded in February 2026, the discussion captures the excitement and tension of a field in transition, as the shift from pure Python to high-performance Rust and C++ binaries delivers a staggering six-fold speedup for circuit transpilation. From the architectural liberation of the 120-qubit Nighthawk processor's square lattice to the cutting-edge fault-tolerant primitives like Pauli Product Measurement, this episode is a must-listen for anyone ready to trade "hobby scripts" for "real engineering". Whether you're fascinated by the future of quantum chemistry or the rigorous demands of error correction, this deep dive offers a compelling look at how the quantum ecosystem is "growing up" to meet the challenges of 2026 and beyond. Want to hear more? Send a message to Qubit Value
• Support & get perks!• Proudly sponsored by PyMC Labs! Get in touch at alex.andorra@pymc-labs.com• Intro to Bayes and Advanced Regression courses (first 2 lessons free)Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work !Chapters:00:00 Exploring Generative AI and Scientific Modeling10:27 Understanding Simulation-Based Inference (SBI) and Its Applications15:59 Diffusion Models in Simulation-Based Inference19:22 Live Coding Session: Implementing Baseflow for SBI34:39 Analyzing Results and Diagnostics in Simulation-Based Inference46:18 Hierarchical Models and Amortized Bayesian Inference48:14 Understanding Simulation-Based Inference (SBI) and Its Importance49:14 Diving into Diffusion Models: Basics and Mechanisms50:38 Forward and Backward Processes in Diffusion Models53:03 Learning the Score: Training Diffusion Models54:57 Inference with Diffusion Models: The Reverse Process57:36 Exploring Variants: Flow Matching and Consistency Models01:01:43 Benchmarking Different Models for Simulation-Based Inference01:06:41 Hierarchical Models and Their Applications in Inference01:14:25 Intervening in the Inference Process: Adding Constraints01:25:35 Summary of Key Concepts and Future DirectionsThank you to my Patrons for making this episode possible!Links from the show:- Come meet Alex at the Field of Play Conference in Manchester, UK, March 27, 2026!- Jonas's Diffusion for SBI Tutorial & Review (Paper & Code)- The BayesFlow Library- Jonas on LinkedIn- Jonas on GitHub- Further reading for more mathematical details: Holderrieth & Erives- 150 Fast Bayesian Deep Learning, with David Rügamer, Emanuel Sommer & Jakob Robnik- 107 Amortized Bayesian Inference with Deep Neural Networks, with Marvin Schmitt
Send a textIn this episode, I had the pleasure of speaking with Jimmy Willis, a Senior Manager of Data Engineering at an AdTech company, where he builds systems that turn massive amounts of raw data into useful information. He is a self-taught programmer without a tech degree who was able to get an internship at JP Morgan Chase and leveraged that opportunity into a 6-figure job. Jimmy is currently writing a book and is on a mission to get 10,000 Black people into tech by learning Python and other real-world tech skills.https://www.rovion.co/Sign up for Activate Your Calling: Create, Build, & Promote Your Gift: https://bit.ly/4r0QixGSign up to be notified about Faith to Launch Community: https://bit.ly/FaithtoLaunchPlease join me in my YouTube only series, 30 Days to Becoming a Stronger, More Confident You in Christ: https://www.youtube.com/playlist?list=PLfkkBA4-h1A56MxObeO__s873pdUnnWQ5
From rewriting Google's search stack in the early 2000s to reviving sparse trillion-parameter models and co-designing TPUs with frontier ML research, Jeff Dean has quietly shaped nearly every layer of the modern AI stack. As Chief AI Scientist at Google and a driving force behind Gemini, Jeff has lived through multiple scaling revolutions from CPUs and sharded indices to multimodal models that reason across text, video, and code.Jeff joins us to unpack what it really means to “own the Pareto frontier,” why distillation is the engine behind every Flash model breakthrough, how energy (in picojoules) not FLOPs is becoming the true bottleneck, what it was like leading the charge to unify all of Google's AI teams, and why the next leap won't come from bigger context windows alone, but from systems that give the illusion of attending to trillions of tokens.We discuss:* Jeff's early neural net thesis in 1990: parallel training before it was cool, why he believed scaling would win decades early, and the “bigger model, more data, better results” mantra that held for 15 years* The evolution of Google Search: sharding, moving the entire index into memory in 2001, softening query semantics pre-LLMs, and why retrieval pipelines already resemble modern LLM systems* Pareto frontier strategy: why you need both frontier “Pro” models and low-latency “Flash” models, and how distillation lets smaller models surpass prior generations* Distillation deep dive: ensembles → compression → logits as soft supervision, and why you need the biggest model to make the smallest one good* Latency as a first-class objective: why 10–50x lower latency changes UX entirely, and how future reasoning workloads will demand 10,000 tokens/sec* Energy-based thinking: picojoules per bit, why moving data costs 1000x more than a multiply, batching through the lens of energy, and speculative decoding as amortization* TPU co-design: predicting ML workloads 2–6 years out, speculative hardware features, precision reduction, sparsity, and the constant feedback loop between model architecture and silicon* Sparse models and “outrageously large” networks: trillions of parameters with 1–5% activation, and why sparsity was always the right abstraction* Unified vs. specialized models: abandoning symbolic systems, why general multimodal models tend to dominate vertical silos, and when vertical fine-tuning still makes sense* Long context and the illusion of scale: beyond needle-in-a-haystack benchmarks toward systems that narrow trillions of tokens to 117 relevant documents* Personalized AI: attending to your emails, photos, and documents (with permission), and why retrieval + reasoning will unlock deeply personal assistants* Coding agents: 50 AI interns, crisp specifications as a new core skill, and how ultra-low latency will reshape human–agent collaboration* Why ideas still matter: transformers, sparsity, RL, hardware, systems — scaling wasn't blind; the pieces had to multiply togetherShow Notes:* Gemma 3 Paper* Gemma 3* Gemini 2.5 Report* Jeff Dean's “Software Engineering Advice fromBuilding Large-Scale Distributed Systems” Presentation (with Back of the Envelope Calculations)* Latency Numbers Every Programmer Should Know by Jeff Dean* The Jeff Dean Facts* Jeff Dean Google Bio* Jeff Dean on “Important AI Trends” @Stanford AI Club* Jeff Dean & Noam Shazeer — 25 years at Google (Dwarkesh)—Jeff Dean* LinkedIn: https://www.linkedin.com/in/jeff-dean-8b212555* X: https://x.com/jeffdeanGoogle* https://google.com* https://deepmind.googleFull Video EpisodeTimestamps00:00:04 — Introduction: Alessio & Swyx welcome Jeff Dean, chief AI scientist at Google, to the Latent Space podcast00:00:30 — Owning the Pareto Frontier & balancing frontier vs low-latency models00:01:31 — Frontier models vs Flash models + role of distillation00:03:52 — History of distillation and its original motivation00:05:09 — Distillation's role in modern model scaling00:07:02 — Model hierarchy (Flash, Pro, Ultra) and distillation sources00:07:46 — Flash model economics & wide deployment00:08:10 — Latency importance for complex tasks00:09:19 — Saturation of some tasks and future frontier tasks00:11:26 — On benchmarks, public vs internal00:12:53 — Example long-context benchmarks & limitations00:15:01 — Long-context goals: attending to trillions of tokens00:16:26 — Realistic use cases beyond pure language00:18:04 — Multimodal reasoning and non-text modalities00:19:05 — Importance of vision & motion modalities00:20:11 — Video understanding example (extracting structured info)00:20:47 — Search ranking analogy for LLM retrieval00:23:08 — LLM representations vs keyword search00:24:06 — Early Google search evolution & in-memory index00:26:47 — Design principles for scalable systems00:28:55 — Real-time index updates & recrawl strategies00:30:06 — Classic “Latency numbers every programmer should know”00:32:09 — Cost of memory vs compute and energy emphasis00:34:33 — TPUs & hardware trade-offs for serving models00:35:57 — TPU design decisions & co-design with ML00:38:06 — Adapting model architecture to hardware00:39:50 — Alternatives: energy-based models, speculative decoding00:42:21 — Open research directions: complex workflows, RL00:44:56 — Non-verifiable RL domains & model evaluation00:46:13 — Transition away from symbolic systems toward unified LLMs00:47:59 — Unified models vs specialized ones00:50:38 — Knowledge vs reasoning & retrieval + reasoning00:52:24 — Vertical model specialization & modules00:55:21 — Token count considerations for vertical domains00:56:09 — Low resource languages & contextual learning00:59:22 — Origins: Dean's early neural network work01:10:07 — AI for coding & human–model interaction styles01:15:52 — Importance of crisp specification for coding agents01:19:23 — Prediction: personalized models & state retrieval01:22:36 — Token-per-second targets (10k+) and reasoning throughput01:23:20 — Episode conclusion and thanksTranscriptAlessio Fanelli [00:00:04]: Hey everyone, welcome to the Latent Space podcast. This is Alessio, founder of Kernel Labs, and I'm joined by Swyx, editor of Latent Space. Shawn Wang [00:00:11]: Hello, hello. We're here in the studio with Jeff Dean, chief AI scientist at Google. Welcome. Thanks for having me. It's a bit surreal to have you in the studio. I've watched so many of your talks, and obviously your career has been super legendary. So, I mean, congrats. I think the first thing must be said, congrats on owning the Pareto Frontier.Jeff Dean [00:00:30]: Thank you, thank you. Pareto Frontiers are good. It's good to be out there.Shawn Wang [00:00:34]: Yeah, I mean, I think it's a combination of both. You have to own the Pareto Frontier. You have to have like frontier capability, but also efficiency, and then offer that range of models that people like to use. And, you know, some part of this was started because of your hardware work. Some part of that is your model work, and I'm sure there's lots of secret sauce that you guys have worked on cumulatively. But, like, it's really impressive to see it all come together in, like, this slittily advanced.Jeff Dean [00:01:04]: Yeah, yeah. I mean, I think, as you say, it's not just one thing. It's like a whole bunch of things up and down the stack. And, you know, all of those really combine to help make UNOS able to make highly capable large models, as well as, you know, software techniques to get those large model capabilities into much smaller, lighter weight models that are, you know, much more cost effective and lower latency, but still, you know, quite capable for their size. Yeah.Alessio Fanelli [00:01:31]: How much pressure do you have on, like, having the lower bound of the Pareto Frontier, too? I think, like, the new labs are always trying to push the top performance frontier because they need to raise more money and all of that. And you guys have billions of users. And I think initially when you worked on the CPU, you were thinking about, you know, if everybody that used Google, we use the voice model for, like, three minutes a day, they were like, you need to double your CPU number. Like, what's that discussion today at Google? Like, how do you prioritize frontier versus, like, we have to do this? How do we actually need to deploy it if we build it?Jeff Dean [00:02:03]: Yeah, I mean, I think we always want to have models that are at the frontier or pushing the frontier because I think that's where you see what capabilities now exist that didn't exist at the sort of slightly less capable last year's version or last six months ago version. At the same time, you know, we know those are going to be really useful for a bunch of use cases, but they're going to be a bit slower and a bit more expensive than people might like for a bunch of other broader models. So I think what we want to do is always have kind of a highly capable sort of affordable model that enables a whole bunch of, you know, lower latency use cases. People can use them for agentic coding much more readily and then have the high-end, you know, frontier model that is really useful for, you know, deep reasoning, you know, solving really complicated math problems, those kinds of things. And it's not that. One or the other is useful. They're both useful. So I think we'd like to do both. And also, you know, through distillation, which is a key technique for making the smaller models more capable, you know, you have to have the frontier model in order to then distill it into your smaller model. So it's not like an either or choice. You sort of need that in order to actually get a highly capable, more modest size model. Yeah.Alessio Fanelli [00:03:24]: I mean, you and Jeffrey came up with the solution in 2014.Jeff Dean [00:03:28]: Don't forget, L'Oreal Vinyls as well. Yeah, yeah.Alessio Fanelli [00:03:30]: A long time ago. But like, I'm curious how you think about the cycle of these ideas, even like, you know, sparse models and, you know, how do you reevaluate them? How do you think about in the next generation of model, what is worth revisiting? Like, yeah, they're just kind of like, you know, you worked on so many ideas that end up being influential, but like in the moment, they might not feel that way necessarily. Yeah.Jeff Dean [00:03:52]: I mean, I think distillation was originally motivated because we were seeing that we had a very large image data set at the time, you know, 300 million images that we could train on. And we were seeing that if you create specialists for different subsets of those image categories, you know, this one's going to be really good at sort of mammals, and this one's going to be really good at sort of indoor room scenes or whatever, and you can cluster those categories and train on an enriched stream of data after you do pre-training on a much broader set of images. You get much better performance. If you then treat that whole set of maybe 50 models you've trained as a large ensemble, but that's not a very practical thing to serve, right? So distillation really came about from the idea of, okay, what if we want to actually serve that and train all these independent sort of expert models and then squish it into something that actually fits in a form factor that you can actually serve? And that's, you know, not that different from what we're doing today. You know, often today we're instead of having an ensemble of 50 models. We're having a much larger scale model that we then distill into a much smaller scale model.Shawn Wang [00:05:09]: Yeah. A part of me also wonders if distillation also has a story with the RL revolution. So let me maybe try to articulate what I mean by that, which is you can, RL basically spikes models in a certain part of the distribution. And then you have to sort of, well, you can spike models, but usually sometimes... It might be lossy in other areas and it's kind of like an uneven technique, but you can probably distill it back and you can, I think that the sort of general dream is to be able to advance capabilities without regressing on anything else. And I think like that, that whole capability merging without loss, I feel like it's like, you know, some part of that should be a distillation process, but I can't quite articulate it. I haven't seen much papers about it.Jeff Dean [00:06:01]: Yeah, I mean, I tend to think of one of the key advantages of distillation is that you can have a much smaller model and you can have a very large, you know, training data set and you can get utility out of making many passes over that data set because you're now getting the logits from the much larger model in order to sort of coax the right behavior out of the smaller model that you wouldn't otherwise get with just the hard labels. And so, you know, I think that's what we've observed. Is you can get, you know, very close to your largest model performance with distillation approaches. And that seems to be, you know, a nice sweet spot for a lot of people because it enables us to kind of, for multiple Gemini generations now, we've been able to make the sort of flash version of the next generation as good or even substantially better than the previous generations pro. And I think we're going to keep trying to do that because that seems like a good trend to follow.Shawn Wang [00:07:02]: So, Dara asked, so it was the original map was Flash Pro and Ultra. Are you just sitting on Ultra and distilling from that? Is that like the mother load?Jeff Dean [00:07:12]: I mean, we have a lot of different kinds of models. Some are internal ones that are not necessarily meant to be released or served. Some are, you know, our pro scale model and we can distill from that as well into our Flash scale model. So I think, you know, it's an important set of capabilities to have and also inference time scaling. It can also be a useful thing to improve the capabilities of the model.Shawn Wang [00:07:35]: And yeah, yeah, cool. Yeah. And obviously, I think the economy of Flash is what led to the total dominance. I think the latest number is like 50 trillion tokens. I don't know. I mean, obviously, it's changing every day.Jeff Dean [00:07:46]: Yeah, yeah. But, you know, by market share, hopefully up.Shawn Wang [00:07:50]: No, I mean, there's no I mean, there's just the economics wise, like because Flash is so economical, like you can use it for everything. Like it's in Gmail now. It's in YouTube. Like it's yeah. It's in everything.Jeff Dean [00:08:02]: We're using it more in our search products of various AI mode reviews.Shawn Wang [00:08:05]: Oh, my God. Flash past the AI mode. Oh, my God. Yeah, that's yeah, I didn't even think about that.Jeff Dean [00:08:10]: I mean, I think one of the things that is quite nice about the Flash model is not only is it more affordable, it's also a lower latency. And I think latency is actually a pretty important characteristic for these models because we're going to want models to do much more complicated things that are going to involve, you know, generating many more tokens from when you ask the model to do so. So, you know, if you're going to ask the model to do something until it actually finishes what you ask it to do, because you're going to ask now, not just write me a for loop, but like write me a whole software package to do X or Y or Z. And so having low latency systems that can do that seems really important. And Flash is one direction, one way of doing that. You know, obviously our hardware platforms enable a bunch of interesting aspects of our, you know, serving stack as well, like TPUs, the interconnect between. Chips on the TPUs is actually quite, quite high performance and quite amenable to, for example, long context kind of attention operations, you know, having sparse models with lots of experts. These kinds of things really, really matter a lot in terms of how do you make them servable at scale.Alessio Fanelli [00:09:19]: Yeah. Does it feel like there's some breaking point for like the proto Flash distillation, kind of like one generation delayed? I almost think about almost like the capability as a. In certain tasks, like the pro model today is a saturated, some sort of task. So next generation, that same task will be saturated at the Flash price point. And I think for most of the things that people use models for at some point, the Flash model in two generation will be able to do basically everything. And how do you make it economical to like keep pushing the pro frontier when a lot of the population will be okay with the Flash model? I'm curious how you think about that.Jeff Dean [00:09:59]: I mean, I think that's true. If your distribution of what people are asking people, the models to do is stationary, right? But I think what often happens is as the models become more capable, people ask them to do more, right? So, I mean, I think this happens in my own usage. Like I used to try our models a year ago for some sort of coding task, and it was okay at some simpler things, but wouldn't do work very well for more complicated things. And since then, we've improved dramatically on the more complicated coding tasks. And now I'll ask it to do much more complicated things. And I think that's true, not just of coding, but of, you know, now, you know, can you analyze all the, you know, renewable energy deployments in the world and give me a report on solar panel deployment or whatever. That's a very complicated, you know, more complicated task than people would have asked a year ago. And so you are going to want more capable models to push the frontier in the absence of what people ask the models to do. And that also then gives us. Insight into, okay, where does the, where do things break down? How can we improve the model in these, these particular areas, uh, in order to sort of, um, make the next generation even better.Alessio Fanelli [00:11:11]: Yeah. Are there any benchmarks or like test sets they use internally? Because it's almost like the same benchmarks get reported every time. And it's like, all right, it's like 99 instead of 97. Like, how do you have to keep pushing the team internally to it? Or like, this is what we're building towards. Yeah.Jeff Dean [00:11:26]: I mean, I think. Benchmarks, particularly external ones that are publicly available. Have their utility, but they often kind of have a lifespan of utility where they're introduced and maybe they're quite hard for current models. You know, I, I like to think of the best kinds of benchmarks are ones where the initial scores are like 10 to 20 or 30%, maybe, but not higher. And then you can sort of work on improving that capability for, uh, whatever it is, the benchmark is trying to assess and get it up to like 80, 90%, whatever. I, I think once it hits kind of 95% or something, you get very diminishing returns from really focusing on that benchmark, cuz it's sort of, it's either the case that you've now achieved that capability, or there's also the issue of leakage in public data or very related kind of data being, being in your training data. Um, so we have a bunch of held out internal benchmarks that we really look at where we know that wasn't represented in the training data at all. There are capabilities that we want the model to have. Um, yeah. Yeah. Um, that it doesn't have now, and then we can work on, you know, assessing, you know, how do we make the model better at these kinds of things? Is it, we need different kind of data to train on that's more specialized for this particular kind of task. Do we need, um, you know, a bunch of, uh, you know, architectural improvements or some sort of, uh, model capability improvements, you know, what would help make that better?Shawn Wang [00:12:53]: Is there, is there such an example that you, uh, a benchmark inspired in architectural improvement? Like, uh, I'm just kind of. Jumping on that because you just.Jeff Dean [00:13:02]: Uh, I mean, I think some of the long context capability of the, of the Gemini models that came, I guess, first in 1.5 really were about looking at, okay, we want to have, um, you know,Shawn Wang [00:13:15]: immediately everyone jumped to like completely green charts of like, everyone had, I was like, how did everyone crack this at the same time? Right. Yeah. Yeah.Jeff Dean [00:13:23]: I mean, I think, um, and once you're set, I mean, as you say that needed single needle and a half. Hey, stack benchmark is really saturated for at least context links up to 1, 2 and K or something. Don't actually have, you know, much larger than 1, 2 and 8 K these days or two or something. We're trying to push the frontier of 1 million or 2 million context, which is good because I think there are a lot of use cases where. Yeah. You know, putting a thousand pages of text or putting, you know, multiple hour long videos and the context and then actually being able to make use of that as useful. Try to, to explore the über graduation are fairly large. But the single needle in a haystack benchmark is sort of saturated. So you really want more complicated, sort of multi-needle or more realistic, take all this content and produce this kind of answer from a long context that sort of better assesses what it is people really want to do with long context. Which is not just, you know, can you tell me the product number for this particular thing?Shawn Wang [00:14:31]: Yeah, it's retrieval. It's retrieval within machine learning. It's interesting because I think the more meta level I'm trying to operate at here is you have a benchmark. You're like, okay, I see the architectural thing I need to do in order to go fix that. But should you do it? Because sometimes that's an inductive bias, basically. It's what Jason Wei, who used to work at Google, would say. Exactly the kind of thing. Yeah, you're going to win. Short term. Longer term, I don't know if that's going to scale. You might have to undo that.Jeff Dean [00:15:01]: I mean, I like to sort of not focus on exactly what solution we're going to derive, but what capability would you want? And I think we're very convinced that, you know, long context is useful, but it's way too short today. Right? Like, I think what you would really want is, can I attend to the internet while I answer my question? Right? But that's not going to happen. I think that's going to be solved by purely scaling the existing solutions, which are quadratic. So a million tokens kind of pushes what you can do. You're not going to do that to a trillion tokens, let alone, you know, a billion tokens, let alone a trillion. But I think if you could give the illusion that you can attend to trillions of tokens, that would be amazing. You'd find all kinds of uses for that. You would have attend to the internet. You could attend to the pixels of YouTube and the sort of deeper representations that we can find. You could attend to the form for a single video, but across many videos, you know, on a personal Gemini level, you could attend to all of your personal state with your permission. So like your emails, your photos, your docs, your plane tickets you have. I think that would be really, really useful. And the question is, how do you get algorithmic improvements and system level improvements that get you to something where you actually can attend to trillions of tokens? Right. In a meaningful way. Yeah.Shawn Wang [00:16:26]: But by the way, I think I did some math and it's like, if you spoke all day, every day for eight hours a day, you only generate a maximum of like a hundred K tokens, which like very comfortably fits.Jeff Dean [00:16:38]: Right. But if you then say, okay, I want to be able to understand everything people are putting on videos.Shawn Wang [00:16:46]: Well, also, I think that the classic example is you start going beyond language into like proteins and whatever else is extremely information dense. Yeah. Yeah.Jeff Dean [00:16:55]: I mean, I think one of the things about Gemini's multimodal aspects is we've always wanted it to be multimodal from the start. And so, you know, that sometimes to people means text and images and video sort of human-like and audio, audio, human-like modalities. But I think it's also really useful to have Gemini know about non-human modalities. Yeah. Like LIDAR sensor data from. Yes. Say, Waymo vehicles or. Like robots or, you know, various kinds of health modalities, x-rays and MRIs and imaging and genomics information. And I think there's probably hundreds of modalities of data where you'd like the model to be able to at least be exposed to the fact that this is an interesting modality and has certain meaning in the world. Where even if you haven't trained on all the LIDAR data or MRI data, you could have, because maybe that's not, you know, it doesn't make sense in terms of trade-offs of. You know, what you include in your main pre-training data mix, at least including a little bit of it is actually quite useful. Yeah. Because it sort of tempts the model that this is a thing.Shawn Wang [00:18:04]: Yeah. Do you believe, I mean, since we're on this topic and something I just get to ask you all the questions I always wanted to ask, which is fantastic. Like, are there some king modalities, like modalities that supersede all the other modalities? So a simple example was Vision can, on a pixel level, encode text. And DeepSeq had this DeepSeq CR paper that did that. Vision. And Vision has also been shown to maybe incorporate audio because you can do audio spectrograms and that's, that's also like a Vision capable thing. Like, so, so maybe Vision is just the king modality and like. Yeah.Jeff Dean [00:18:36]: I mean, Vision and Motion are quite important things, right? Motion. Well, like video as opposed to static images, because I mean, there's a reason evolution has evolved eyes like 23 independent ways, because it's such a useful capability for sensing the world around you, which is really what we want these models to be. So I think the only thing that we can be able to do is interpret the things we're seeing or the things we're paying attention to and then help us in using that information to do things. Yeah.Shawn Wang [00:19:05]: I think motion, you know, I still want to shout out, I think Gemini, still the only native video understanding model that's out there. So I use it for YouTube all the time. Nice.Jeff Dean [00:19:15]: Yeah. Yeah. I mean, it's actually, I think people kind of are not necessarily aware of what the Gemini models can actually do. Yeah. Like I have an example I've used in one of my talks. It had like, it was like a YouTube highlight video of 18 memorable sports moments across the last 20 years or something. So it has like Michael Jordan hitting some jump shot at the end of the finals and, you know, some soccer goals and things like that. And you can literally just give it the video and say, can you please make me a table of what all these different events are? What when the date is when they happened? And a short description. And so you get like now an 18 row table of that information extracted from the video, which is, you know, not something most people think of as like a turn video into sequel like table.Alessio Fanelli [00:20:11]: Has there been any discussion inside of Google of like, you mentioned tending to the whole internet, right? Google, it's almost built because a human cannot tend to the whole internet and you need some sort of ranking to find what you need. Yep. That ranking is like much different for an LLM because you can expect a person to look at maybe the first five, six links in a Google search versus for an LLM. Should you expect to have 20 links that are highly relevant? Like how do you internally figure out, you know, how do we build the AI mode that is like maybe like much broader search and span versus like the more human one? Yeah.Jeff Dean [00:20:47]: I mean, I think even pre-language model based work, you know, our ranking systems would be built to start. I mean, I think even pre-language model based work, you know, our ranking systems would be built to start. With a giant number of web pages in our index, many of them are not relevant. So you identify a subset of them that are relevant with very lightweight kinds of methods. You know, you're down to like 30,000 documents or something. And then you gradually refine that to apply more and more sophisticated algorithms and more and more sophisticated sort of signals of various kinds in order to get down to ultimately what you show, which is, you know, the final 10 results or, you know, 10 results plus. Other kinds of information. And I think an LLM based system is not going to be that dissimilar, right? You're going to attend to trillions of tokens, but you're going to want to identify, you know, what are the 30,000 ish documents that are with the, you know, maybe 30 million interesting tokens. And then how do you go from that into what are the 117 documents I really should be paying attention to in order to carry out the tasks that the user has asked? And I think, you know, you can imagine systems where you have, you know, a lot of highly parallel processing to identify those initial 30,000 candidates, maybe with very lightweight kinds of models. Then you have some system that sort of helps you narrow down from 30,000 to the 117 with maybe a little bit more sophisticated model or set of models. And then maybe the final model is the thing that looks. So the 117 things that might be your most capable model. So I think it has to, it's going to be some system like that, that is really enables you to give the illusion of attending to trillions of tokens. Sort of the way Google search gives you, you know, not the illusion, but you are searching the internet, but you're finding, you know, a very small subset of things that are, that are relevant.Shawn Wang [00:22:47]: Yeah. I often tell a lot of people that are not steeped in like Google search history that, well, you know, like Bert was. Like he was like basically immediately inside of Google search and that improves results a lot, right? Like I don't, I don't have any numbers off the top of my head, but like, I'm sure you guys, that's obviously the most important numbers to Google. Yeah.Jeff Dean [00:23:08]: I mean, I think going to an LLM based representation of text and words and so on enables you to get out of the explicit hard notion of, of particular words having to be on the page, but really getting at the notion of this topic of this page or this page. Paragraph is highly relevant to this query. Yeah.Shawn Wang [00:23:28]: I don't think people understand how much LLMs have taken over all these very high traffic system, very high traffic. Yeah. Like it's Google, it's YouTube. YouTube has this like semantics ID thing where it's just like every token or every item in the vocab is a YouTube video or something that predicts the video using a code book, which is absurd to me for YouTube size.Jeff Dean [00:23:50]: And then most recently GROK also for, for XAI, which is like, yeah. I mean, I'll call out even before LLMs were used extensively in search, we put a lot of emphasis on softening the notion of what the user actually entered into the query.Shawn Wang [00:24:06]: So do you have like a history of like, what's the progression? Oh yeah.Jeff Dean [00:24:09]: I mean, I actually gave a talk in, uh, I guess, uh, web search and data mining conference in 2009, uh, where we never actually published any papers about the origins of Google search, uh, sort of, but we went through sort of four or five or six. generations, four or five or six generations of, uh, redesigning of the search and retrieval system, uh, from about 1999 through 2004 or five. And that talk is really about that evolution. And one of the things that really happened in 2001 was we were sort of working to scale the system in multiple dimensions. So one is we wanted to make our index bigger, so we could retrieve from a larger index, which always helps your quality in general. Uh, because if you don't have the page in your index, you're going to not do well. Um, and then we also needed to scale our capacity because we were, our traffic was growing quite extensively. Um, and so we had, you know, a sharded system where you have more and more shards as the index grows, you have like 30 shards. And then if you want to double the index size, you make 60 shards so that you can bound the latency by which you respond for any particular user query. Um, and then as traffic grows, you add, you add more and more replicas of each of those. And so we eventually did the math that realized that in a data center where we had say 60 shards and, um, you know, 20 copies of each shard, we now had 1200 machines, uh, with disks. And we did the math and we're like, Hey, one copy of that index would actually fit in memory across 1200 machines. So in 2001, we introduced, uh, we put our entire index in memory and what that enabled from a quality perspective was amazing. Um, and so we had more and more replicas of each of those. Before you had to be really careful about, you know, how many different terms you looked at for a query, because every one of them would involve a disk seek on every one of the 60 shards. And so you, as you make your index bigger, that becomes even more inefficient. But once you have the whole index in memory, it's totally fine to have 50 terms you throw into the query from the user's original three or four word query, because now you can add synonyms like restaurant and restaurants and cafe and, uh, you know, things like that. Uh, bistro and all these things. And you can suddenly start, uh, sort of really, uh, getting at the meaning of the word as opposed to the exact semantic form the user typed in. And that was, you know, 2001, very much pre LLM, but really it was about softening the, the strict definition of what the user typed in order to get at the meaning.Alessio Fanelli [00:26:47]: What are like principles that you use to like design the systems, especially when you have, I mean, in 2001, the internet is like. Doubling, tripling every year in size is not like, uh, you know, and I think today you kind of see that with LLMs too, where like every year the jumps in size and like capabilities are just so big. Are there just any, you know, principles that you use to like, think about this? Yeah.Jeff Dean [00:27:08]: I mean, I think, uh, you know, first, whenever you're designing a system, you want to understand what are the sort of design parameters that are going to be most important in designing that, you know? So, you know, how many queries per second do you need to handle? How big is the internet? How big is the index you need to handle? How much data do you need to keep for every document in the index? How are you going to look at it when you retrieve things? Um, what happens if traffic were to double or triple, you know, will that system work well? And I think a good design principle is you're going to want to design a system so that the most important characteristics could scale by like factors of five or 10, but probably not beyond that because often what happens is if you design a system for X. And something suddenly becomes a hundred X, that would enable a very different point in the design space that would not make sense at X. But all of a sudden at a hundred X makes total sense. So like going from a disk space index to a in memory index makes a lot of sense once you have enough traffic, because now you have enough replicas of the sort of state on disk that those machines now actually can hold, uh, you know, a full copy of the, uh, index and memory. Yeah. And that all of a sudden enabled. A completely different design that wouldn't have been practical before. Yeah. Um, so I'm, I'm a big fan of thinking through designs in your head, just kind of playing with the design space a little before you actually do a lot of writing of code. But, you know, as you said, in the early days of Google, we were growing the index, uh, quite extensively. We were growing the update rate of the index. So the update rate actually is the parameter that changed the most. Surprising. So it used to be once a month.Shawn Wang [00:28:55]: Yeah.Jeff Dean [00:28:56]: And then we went to a system that could update any particular page in like sub one minute. Okay.Shawn Wang [00:29:02]: Yeah. Because this is a competitive advantage, right?Jeff Dean [00:29:04]: Because all of a sudden news related queries, you know, if you're, if you've got last month's news index, it's not actually that useful for.Shawn Wang [00:29:11]: News is a special beast. Was there any, like you could have split it onto a separate system.Jeff Dean [00:29:15]: Well, we did. We launched a Google news product, but you also want news related queries that people type into the main index to also be sort of updated.Shawn Wang [00:29:23]: So, yeah, it's interesting. And then you have to like classify whether the page is, you have to decide which pages should be updated and what frequency. Oh yeah.Jeff Dean [00:29:30]: There's a whole like, uh, system behind the scenes that's trying to decide update rates and importance of the pages. So even if the update rate seems low, you might still want to recrawl important pages quite often because, uh, the likelihood they change might be low, but the value of having updated is high.Shawn Wang [00:29:50]: Yeah, yeah, yeah, yeah. Uh, well, you know, yeah. This, uh, you know, mention of latency and, and saving things to this reminds me of one of your classics, which I have to bring up, which is latency numbers. Every programmer should know, uh, was there a, was it just a, just a general story behind that? Did you like just write it down?Jeff Dean [00:30:06]: I mean, this has like sort of eight or 10 different kinds of metrics that are like, how long does a cache mistake? How long does branch mispredict take? How long does a reference domain memory take? How long does it take to send, you know, a packet from the U S to the Netherlands or something? Um,Shawn Wang [00:30:21]: why Netherlands, by the way, or is it, is that because of Chrome?Jeff Dean [00:30:25]: Uh, we had a data center in the Netherlands, um, so, I mean, I think this gets to the point of being able to do the back of the envelope calculations. So these are sort of the raw ingredients of those, and you can use them to say, okay, well, if I need to design a system to do image search and thumb nailing or something of the result page, you know, how, what I do that I could pre-compute the image thumbnails. I could like. Try to thumbnail them on the fly from the larger images. What would that do? How much dis bandwidth than I need? How many des seeks would I do? Um, and you can sort of actually do thought experiments in, you know, 30 seconds or a minute with the sort of, uh, basic, uh, basic numbers at your fingertips. Uh, and then as you sort of build software using higher level libraries, you kind of want to develop the same intuitions for how long does it take to, you know, look up something in this particular kind of.Shawn Wang [00:31:21]: I'll see you next time.Shawn Wang [00:31:51]: Which is a simple byte conversion. That's nothing interesting. I wonder if you have any, if you were to update your...Jeff Dean [00:31:58]: I mean, I think it's really good to think about calculations you're doing in a model, either for training or inference.Jeff Dean [00:32:09]: Often a good way to view that is how much state will you need to bring in from memory, either like on-chip SRAM or HBM from the accelerator. Attached memory or DRAM or over the network. And then how expensive is that data motion relative to the cost of, say, an actual multiply in the matrix multiply unit? And that cost is actually really, really low, right? Because it's order, depending on your precision, I think it's like sub one picodule.Shawn Wang [00:32:50]: Oh, okay. You measure it by energy. Yeah. Yeah.Jeff Dean [00:32:52]: Yeah. I mean, it's all going to be about energy and how do you make the most energy efficient system. And then moving data from the SRAM on the other side of the chip, not even off the off chip, but on the other side of the same chip can be, you know, a thousand picodules. Oh, yeah. And so all of a sudden, this is why your accelerators require batching. Because if you move, like, say, the parameter of a model from SRAM on the, on the chip into the multiplier unit, that's going to cost you a thousand picodules. So you better make use of that, that thing that you moved many, many times with. So that's where the batch dimension comes in. Because all of a sudden, you know, if you have a batch of 256 or something, that's not so bad. But if you have a batch of one, that's really not good.Shawn Wang [00:33:40]: Yeah. Yeah. Right.Jeff Dean [00:33:41]: Because then you paid a thousand picodules in order to do your one picodule multiply.Shawn Wang [00:33:46]: I have never heard an energy-based analysis of batching.Jeff Dean [00:33:50]: Yeah. I mean, that's why people batch. Yeah. Ideally, you'd like to use batch size one because the latency would be great.Shawn Wang [00:33:56]: The best latency.Jeff Dean [00:33:56]: But the energy cost and the compute cost inefficiency that you get is quite large. So, yeah.Shawn Wang [00:34:04]: Is there a similar trick like, like, like you did with, you know, putting everything in memory? Like, you know, I think obviously NVIDIA has caused a lot of waves with betting very hard on SRAM with Grok. I wonder if, like, that's something that you already saw with, with the TPUs, right? Like that, that you had to. Uh, to serve at your scale, uh, you probably sort of saw that coming. Like what, what, what hardware, uh, innovations or insights were formed because of what you're seeing there?Jeff Dean [00:34:33]: Yeah. I mean, I think, you know, TPUs have this nice, uh, sort of regular structure of 2D or 3D meshes with a bunch of chips connected. Yeah. And each one of those has HBM attached. Um, I think for serving some kinds of models, uh, you know, you, you pay a lot higher cost. Uh, and time latency, um, bringing things in from HBM than you do bringing them in from, uh, SRAM on the chip. So if you have a small enough model, you can actually do model parallelism, spread it out over lots of chips and you actually get quite good throughput improvements and latency improvements from doing that. And so you're now sort of striping your smallish scale model over say 16 or 64 chips. Uh, but as if you do that and it all fits in. In SRAM, uh, that can be a big win. So yeah, that's not a surprise, but it is a good technique.Alessio Fanelli [00:35:27]: Yeah. What about the TPU design? Like how much do you decide where the improvements have to go? So like, this is like a good example of like, is there a way to bring the thousand picojoules down to 50? Like, is it worth designing a new chip to do that? The extreme is like when people say, oh, you should burn the model on the ASIC and that's kind of like the most extreme thing. How much of it? Is it worth doing an hardware when things change so quickly? Like what was the internal discussion? Yeah.Jeff Dean [00:35:57]: I mean, we, we have a lot of interaction between say the TPU chip design architecture team and the sort of higher level modeling, uh, experts, because you really want to take advantage of being able to co-design what should future TPUs look like based on where we think the sort of ML research puck is going, uh, in some sense, because, uh, you know, as a hardware designer for ML and in particular, you're trying to design a chip starting today and that design might take two years before it even lands in a data center. And then it has to sort of be a reasonable lifetime of the chip to take you three, four or five years. So you're trying to predict two to six years out where, what ML computations will people want to run two to six years out in a very fast changing field. And so having people with interest. Interesting ML research ideas of things we think will start to work in that timeframe or will be more important in that timeframe, uh, really enables us to then get, you know, interesting hardware features put into, you know, TPU N plus two, where TPU N is what we have today.Shawn Wang [00:37:10]: Oh, the cycle time is plus two.Jeff Dean [00:37:12]: Roughly. Wow. Because, uh, I mean, sometimes you can squeeze some changes into N plus one, but, you know, bigger changes are going to require the chip. Yeah. Design be earlier in its lifetime design process. Um, so whenever we can do that, it's generally good. And sometimes you can put in speculative features that maybe won't cost you much chip area, but if it works out, it would make something, you know, 10 times as fast. And if it doesn't work out, well, you burned a little bit of tiny amount of your chip area on that thing, but it's not that big a deal. Uh, sometimes it's a very big change and we want to be pretty sure this is going to work out. So we'll do like lots of carefulness. Uh, ML experimentation to show us, uh, this is actually the, the way we want to go. Yeah.Alessio Fanelli [00:37:58]: Is there a reverse of like, we already committed to this chip design so we can not take the model architecture that way because it doesn't quite fit?Jeff Dean [00:38:06]: Yeah. I mean, you, you definitely have things where you're going to adapt what the model architecture looks like so that they're efficient on the chips that you're going to have for both training and inference of that, of that, uh, generation of model. So I think it kind of goes both ways. Um, you know, sometimes you can take advantage of, you know, lower precision things that are coming in a future generation. So you can, might train it at that lower precision, even if the current generation doesn't quite do that. Mm.Shawn Wang [00:38:40]: Yeah. How low can we go in precision?Jeff Dean [00:38:43]: Because people are saying like ternary is like, uh, yeah, I mean, I'm a big fan of very low precision because I think that gets, that saves you a tremendous amount of time. Right. Because it's picojoules per bit that you're transferring and reducing the number of bits is a really good way to, to reduce that. Um, you know, I think people have gotten a lot of luck, uh, mileage out of having very low bit precision things, but then having scaling factors that apply to a whole bunch of, uh, those, those weights. Scaling. How does it, how does it, okay.Shawn Wang [00:39:15]: Interesting. You, so low, low precision, but scaled up weights. Yeah. Huh. Yeah. Never considered that. Yeah. Interesting. Uh, w w while we're on this topic, you know, I think there's a lot of, um, uh, this, the concept of precision at all is weird when we're sampling, you know, uh, we just, at the end of this, we're going to have all these like chips that I'll do like very good math. And then we're just going to throw a random number generator at the start. So, I mean, there's a movement towards, uh, energy based, uh, models and processors. I'm just curious if you've, obviously you've thought about it, but like, what's your commentary?Jeff Dean [00:39:50]: Yeah. I mean, I think. There's a bunch of interesting trends though. Energy based models is one, you know, diffusion based models, which don't sort of sequentially decode tokens is another, um, you know, speculative decoding is a way that you can get sort of an equivalent, very small.Shawn Wang [00:40:06]: Draft.Jeff Dean [00:40:07]: Batch factor, uh, for like you predict eight tokens out and that enables you to sort of increase the effective batch size of what you're doing by a factor of eight, even, and then you maybe accept five or six of those tokens. So you get. A five, a five X improvement in the amortization of moving weights, uh, into the multipliers to do the prediction for the, the tokens. So these are all really good techniques and I think it's really good to look at them from the lens of, uh, energy, real energy, not energy based models, um, and, and also latency and throughput, right? If you look at things from that lens, that sort of guides you to. Two solutions that are gonna be, uh, you know, better from, uh, you know, being able to serve larger models or, you know, equivalent size models more cheaply and with lower latency.Shawn Wang [00:41:03]: Yeah. Well, I think, I think I, um, it's appealing intellectually, uh, haven't seen it like really hit the mainstream, but, um, I do think that, uh, there's some poetry in the sense that, uh, you know, we don't have to do, uh, a lot of shenanigans if like we fundamentally. Design it into the hardware. Yeah, yeah.Jeff Dean [00:41:23]: I mean, I think there's still a, there's also sort of the more exotic things like analog based, uh, uh, computing substrates as opposed to digital ones. Uh, I'm, you know, I think those are super interesting cause they can be potentially low power. Uh, but I think you often end up wanting to interface that with digital systems and you end up losing a lot of the power advantages in the digital to analog and analog to digital conversions. You end up doing, uh, at the sort of boundaries. And periphery of that system. Um, I still think there's a tremendous distance we can go from where we are today in terms of energy efficiency with sort of, uh, much better and specialized hardware for the models we care about.Shawn Wang [00:42:05]: Yeah.Alessio Fanelli [00:42:06]: Um, any other interesting research ideas that you've seen, or like maybe things that you cannot pursue a Google that you would be interested in seeing researchers take a step at, I guess you have a lot of researchers. Yeah, I guess you have enough, but our, our research.Jeff Dean [00:42:21]: Our research portfolio is pretty broad. I would say, um, I mean, I think, uh, in terms of research directions, there's a whole bunch of, uh, you know, open problems and how do you make these models reliable and able to do much longer, kind of, uh, more complex tasks that have lots of subtasks. How do you orchestrate, you know, maybe one model that's using other models as tools in order to sort of build, uh, things that can accomplish, uh, you know, much more. Yeah. Significant pieces of work, uh, collectively, then you would ask a single model to do. Um, so that's super interesting. How do you get more verifiable, uh, you know, how do you get RL to work for non-verifiable domains? I think it's a pretty interesting open problem because I think that would broaden out the capabilities of the models, the improvements that you're seeing in both math and coding. Uh, if we could apply those to other less verifiable domains, because we've come up with RL techniques that actually enable us to do that. Uh, effectively, that would, that would really make the models improve quite a lot. I think.Alessio Fanelli [00:43:26]: I'm curious, like when we had Noam Brown on the podcast, he said, um, they already proved you can do it with deep research. Um, you kind of have it with AI mode in a way it's not verifiable. I'm curious if there's any thread that you think is interesting there. Like what is it? Both are like information retrieval of JSON. So I wonder if it's like the retrieval is like the verifiable part. That you can score or what are like, yeah, yeah. How, how would you model that, that problem?Jeff Dean [00:43:55]: Yeah. I mean, I think there are ways of having other models that can evaluate the results of what a first model did, maybe even retrieving. Can you have another model that says, is this things, are these things you retrieved relevant? Or can you rate these 2000 things you retrieved to assess which ones are the 50 most relevant or something? Um, I think those kinds of techniques are actually quite effective. Sometimes I can even be the same model, just prompted differently to be a, you know, a critic as opposed to a, uh, actual retrieval system. Yeah.Shawn Wang [00:44:28]: Um, I do think like there, there is that, that weird cliff where like, it feels like we've done the easy stuff and then now it's, but it always feels like that every year. It's like, oh, like we know, we know, and the next part is super hard and nobody's figured it out. And, uh, exactly with this RLVR thing where like everyone's talking about, well, okay, how do we. the next stage of the non-verifiable stuff. And everyone's like, I don't know, you know, Ellen judge.Jeff Dean [00:44:56]: I mean, I feel like the nice thing about this field is there's lots and lots of smart people thinking about creative solutions to some of the problems that we all see. Uh, because I think everyone sort of sees that the models, you know, are great at some things and they fall down around the edges of those things and, and are not as capable as we'd like in those areas. And then coming up with good techniques and trying those. And seeing which ones actually make a difference is sort of what the whole research aspect of this field is, is pushing forward. And I think that's why it's super interesting. You know, if you think about two years ago, we were struggling with GSM, eight K problems, right? Like, you know, Fred has two rabbits. He gets three more rabbits. How many rabbits does he have? That's a pretty far cry from the kinds of mathematics that the models can, and now you're doing IMO and Erdos problems in pure language. Yeah. Yeah. Pure language. So that is a really, really amazing jump in capabilities in, you know, in a year and a half or something. And I think, um, for other areas, it'd be great if we could make that kind of leap. Uh, and you know, we don't exactly see how to do it for some, some areas, but we do see it for some other areas and we're going to work hard on making that better. Yeah.Shawn Wang [00:46:13]: Yeah.Alessio Fanelli [00:46:14]: Like YouTube thumbnail generation. That would be very helpful. We need that. That would be AGI. We need that.Shawn Wang [00:46:20]: That would be. As far as content creators go.Jeff Dean [00:46:22]: I guess I'm not a YouTube creator, so I don't care that much about that problem, but I guess, uh, many people do.Shawn Wang [00:46:27]: It does. Yeah. It doesn't, it doesn't matter. People do judge books by their covers as it turns out. Um, uh, just to draw a bit on the IMO goal. Um, I'm still not over the fact that a year ago we had alpha proof and alpha geometry and all those things. And then this year we were like, screw that we'll just chuck it into Gemini. Yeah. What's your reflection? Like, I think this, this question about. Like the merger of like symbolic systems and like, and, and LMS, uh, was a very much core belief. And then somewhere along the line, people would just said, Nope, we'll just all do it in the LLM.Jeff Dean [00:47:02]: Yeah. I mean, I think it makes a lot of sense to me because, you know, humans manipulate symbols, but we probably don't have like a symbolic representation in our heads. Right. We have some distributed representation that is neural net, like in some way of lots of different neurons. And activation patterns firing when we see certain things and that enables us to reason and plan and, you know, do chains of thought and, you know, roll them back now that, that approach for solving the problem doesn't seem like it's going to work. I'm going to try this one. And, you know, in a lot of ways we're emulating what we intuitively think, uh, is happening inside real brains in neural net based models. So it never made sense to me to have like completely separate. Uh, discrete, uh, symbolic things, and then a completely different way of, of, uh, you know, thinking about those things.Shawn Wang [00:47:59]: Interesting. Yeah. Uh, I mean, it's maybe seems obvious to you, but it wasn't obvious to me a year ago. Yeah.Jeff Dean [00:48:06]: I mean, I do think like that IMO with, you know, translating to lean and using lean and then the next year and also a specialized geometry model. And then this year switching to a single unified model. That is roughly the production model with a little bit more inference budget, uh, is actually, you know, quite good because it shows you that the capabilities of that general model have improved dramatically and, and now you don't need the specialized model. This is actually sort of very similar to the 2013 to 16 era of machine learning, right? Like it used to be, people would train separate models for lots of different, each different problem, right? I have, I want to recognize street signs and something. So I train a street sign. Recognition recognition model, or I want to, you know, decode speech recognition. I have a speech model, right? I think now the era of unified models that do everything is really upon us. And the question is how well do those models generalize to new things they've never been asked to do and they're getting better and better.Shawn Wang [00:49:10]: And you don't need domain experts. Like one of my, uh, so I interviewed ETA who was on, who was on that team. Uh, and he was like, yeah, I, I don't know how they work. I don't know where the IMO competition was held. I don't know the rules of it. I just trained the models, the training models. Yeah. Yeah. And it's kind of interesting that like people with these, this like universal skill set of just like machine learning, you just give them data and give them enough compute and they can kind of tackle any task, which is the bitter lesson, I guess. I don't know. Yeah.Jeff Dean [00:49:39]: I mean, I think, uh, general models, uh, will win out over specialized ones in most cases.Shawn Wang [00:49:45]: Uh, so I want to push there a bit. I think there's one hole here, which is like, uh. There's this concept of like, uh, maybe capacity of a model, like abstractly a model can only contain the number of bits that it has. And, uh, and so it, you know, God knows like Gemini pro is like one to 10 trillion parameters. We don't know, but, uh, the Gemma models, for example, right? Like a lot of people want like the open source local models that are like that, that, that, and, and, uh, they have some knowledge, which is not necessary, right? Like they can't know everything like, like you have the. The luxury of you have the big model and big model should be able to capable of everything. But like when, when you're distilling and you're going down to the small models, you know, you're actually memorizing things that are not useful. Yeah. And so like, how do we, I guess, do we want to extract that? Can we, can we divorce knowledge from reasoning, you know?Jeff Dean [00:50:38]: Yeah. I mean, I think you do want the model to be most effective at reasoning if it can retrieve things, right? Because having the model devote precious parameter space. To remembering obscure facts that could be looked up is actually not the best use of that parameter space, right? Like you might prefer something that is more generally useful in more settings than this obscure fact that it has. Um, so I think that's always attention at the same time. You also don't want your model to be kind of completely detached from, you know, knowing stuff about the world, right? Like it's probably useful to know how long the golden gate be. Bridges just as a general sense of like how long are bridges, right? And, uh, it should have that kind of knowledge. It maybe doesn't need to know how long some teeny little bridge in some other more obscure part of the world is, but, uh, it does help it to have a fair bit of world knowledge and the bigger your model is, the more you can have. Uh, but I do think combining retrieval with sort of reasoning and making the model really good at doing multiple stages of retrieval. Yeah.Shawn Wang [00:51:49]: And reasoning through the intermediate retrieval results is going to be a, a pretty effective way of making the model seem much more capable, because if you think about, say, a personal Gemini, yeah, right?Jeff Dean [00:52:01]: Like we're not going to train Gemini on my email. Probably we'd rather have a single model that, uh, we can then use and use being able to retrieve from my email as a tool and have the model reason about it and retrieve from my photos or whatever, uh, and then make use of that and have multiple. Um, you know, uh, stages of interaction. that makes sense.Alessio Fanelli [00:52:24]: Do you think the vertical models are like, uh, interesting pursuit? Like when people are like, oh, we're building the best healthcare LLM, we're building the best law LLM, are those kind of like short-term stopgaps or?Jeff Dean [00:52:37]: No, I mean, I think, I think vertical models are interesting. Like you want them to start from a pretty good base model, but then you can sort of, uh, sort of viewing them, view them as enriching the data. Data distribution for that particular vertical domain for healthcare, say, um, we're probably not going to train or for say robotics. We're probably not going to train Gemini on all possible robotics data. We, you could train it on because we want it to have a balanced set of capabilities. Um, so we'll expose it to some robotics data, but if you're trying to build a really, really good robotics model, you're going to want to start with that and then train it on more robotics data. And then maybe that would. It's multilingual translation capability, but improve its robotics capabilities. And we're always making these kind of, uh, you know, trade-offs in the data mix that we train the base Gemini models on. You know, we'd love to include data from 200 more languages and as much data as we have for those languages, but that's going to displace some other capabilities of the model. It won't be as good at, um, you know, Pearl programming, you know, it'll still be good at Python programming. Cause we'll include it. Enough. Of that, but there's other long tail computer languages or coding capabilities that it may suffer on or multi, uh, multimodal reasoning capabilities may suffer. Cause we didn't get to expose it to as much data there, but it's really good at multilingual things. So I, I think some combination of specialized models, maybe more modular models. So it'd be nice to have the capability to have those 200 languages, plus this awesome robotics model, plus this awesome healthcare, uh, module that all can be knitted together to work in concert and called upon in different circumstances. Right? Like if I have a health related thing, then it should enable using this health module in conjunction with the main base model to be even better at those kinds of things. Yeah.Shawn Wang [00:54:36]: Installable knowledge. Yeah.Jeff Dean [00:54:37]: Right.Shawn Wang [00:54:38]: Just download as a, as a package.Jeff Dean [00:54:39]: And some of that installable stuff can come from retrieval, but some of it probably should come from preloaded training on, you know, uh, a hundred billion tokens or a trillion tokens of health data. Yeah.Shawn Wang [00:54:51]: And for listeners, I think, uh, I will highlight the Gemma three end paper where they, there was a little bit of that, I think. Yeah.Alessio Fanelli [00:54:56]: Yeah. I guess the question is like, how many billions of tokens do you need to outpace the frontier model improvements? You know, it's like, if I have to make this model better healthcare and the main. Gemini model is still improving. Do I need 50 billion tokens? Can I do it with a hundred, if I need a trillion healthcare tokens, it's like, they're probably not out there that you don't have, you know, I think that's really like the.Jeff Dean [00:55:21]: Well, I mean, I think healthcare is a particularly challenging domain, so there's a lot of healthcare data that, you know, we don't have access to appropriately, but there's a lot of, you know, uh, healthcare organizations that want to train models on their own data. That is not public healthcare data, uh, not public health. But public healthcare data. Um, so I think there are opportunities there to say, partner with a large healthcare organization and train models for their use that are going to be, you know, more bespoke, but probably, uh, might be better than a general model trained on say, public data. Yeah.Shawn Wang [00:55:58]: Yeah. I, I believe, uh, by the way, also this is like somewhat related to the language conversation. Uh, I think one of your, your favorite examples was you can put a low resource language in the context and it just learns. Yeah.Jeff Dean [00:56:09]: Oh, yeah, I think the example we used was Calamon, which is truly low resource because it's only spoken by, I think 120 people in the world and there's no written text.Shawn Wang [00:56:20]: So, yeah. So you can just do it that way. Just put it in the context. Yeah. Yeah. But I think your whole data set in the context, right.Jeff Dean [00:56:27]: If you, if you take a language like, uh, you know, Somali or something, there is a fair bit of Somali text in the world that, uh, or Ethiopian Amharic or something, um, you know, we probably. Yeah. Are not putting all the data from those languages into the Gemini based training. We put some of it, but if you put more of it, you'll improve the capabilities of those models.Shawn Wang [00:56:49]: Yeah.Jeff Dean [00:56:49]:
Чтобы научиться программировать и разбираться в тонкостях Python 3.12 записывайтесь на базовый курс Learn Python — https://clck.ru/3MuShF Ведущие – Григорий Петров и Михаил Корнеев Ссылки от Сергея: toloka.ai - компания https://platform.toloka.ai/ - self service (быстрая разметка данных) tendem.ai - Tendem (кентавр AI + human) как начать работать в Толоке экспертом: https://mindrift.ai/apply Ссылки выпуска: Курс Learn Python — https://learn.python.ru/advanced Канал Миши в Telegram — https://t.me/tricky_python Канал Moscow Python в Telegram — https://t.me/moscow_python Все выпуски — https://podcast.python.ru Митапы Moscow Python — https://moscowpython.ru Канал Moscow Python на Rutube — https://rutube.ru/channel/45885590/ Канал Moscow Python в VK — https://vk.com/moscowpythonconf
In this episode, Bob sits down with Arthur, a Python engineer based in France and the creator of CodSpeed, to dig into a problem many teams don't notice until it's too late: performance regressions. Arthur shares the story behind building CodSpeed, starting with real-world pain points from robotics and machine-learning pipelines where small slowdowns quietly piled up and broke systems in production. We discuss how CodSpeed fits into everyday developer workflows and why treating performance checks like tests or coverage changes how teams ship code. Arthur also shares how open source shaped the product's mission, the surprising environmental impact of avoiding tiny regressions at scale, and why AI-driven coding makes performance guardrails more important than ever.Reach out to Arthur on LinkedIn: https://www.linkedin.com/in/arthurpastel/ Arthur's website: https://apas.tel/Check out Codspeed: https://codspeed.io/ Github: https://github.com/CodSpeedHQ/codspeed___
In the rapidly evolving landscape of technology, artificial intelligence (AI) stands out as a transformative force that democratizes access to complex systems and tools. Dave Robins discusses a new device, the FREE-WILi box, which illustrates how AI can bridge the gap between advanced technology and users who may lack extensive technical skills. By simplifying the process of interaction with technology, AI enables a broader audience to engage with and benefit from the advancements in the tech world.AI Simplifies Tech for EveryoneHistorically, engaging with technology required a certain level of expertise. Devices like Arduino and Raspberry Pi are excellent examples of platforms that allow users to create and innovate, but they often demand a solid understanding of programming languages such as C or Python. While these platforms have fostered a community of makers and enthusiasts, they inadvertently excluded those who may have a passion for technology but lack the coding skills necessary to harness its full potential. This is where AI comes into play, transforming the way individuals interact with technology.The FREE-WILi box exemplifies this shift. By integrating sensors and actuators with a user-friendly Python API, the device allows users to create interactive applications without needing to dive deep into programming. As Robins explains, the AI simplifies the process by providing clear documentation and examples, making it accessible for those who may not have a technical background. This approach not only empowers users to explore their creativity but also fosters a sense of community where individuals can share their projects and ideas.A Celebration of Community and CreativityThe comparison of the current AI revolution to the early days of personal computing in the 1980s is particularly poignant. During that era, the excitement surrounding computers stemmed from their novelty and the potential for personal expression and creativity. Similarly, AI today is igniting that same enthusiasm, enabling individuals to engage with technology in ways that were previously unimaginable. Marlo reflects on his own experiences with retro computing, highlighting how the simplicity and accessibility of early computers allowed a generation to experiment, learn, and innovate. This nostalgia underscores the importance of making technology approachable once again.Moreover, the real-world applications of AI further illustrate its significance. For instance, the ability to program the FREE-WILi box to control lighting at events, such as concerts, showcases how AI can facilitate complex tasks with minimal effort from the user. By simply specifying desired outcomes-like changing colors based on device orientation-individuals can create dynamic and engaging experiences without needing to understand the underlying technology. This capability transforms the way events are organized, allowing for more creativity and personalization.Implications Far Beyond Home UseThe implications of AI simplifying technology extend beyond individual users to entire communities and industries. As more people gain access to tools that were once reserved for experts, the potential for innovation and collaboration increases exponentially. AI acts as a catalyst for creativity, enabling a diverse range of individuals to contribute to the technological landscape. This shift not only leads to the development of new ideas and solutions but also fosters inclusivity, ensuring that technology serves as a tool for everyone, regardless of their technical background.ConclusionIn conclusion, AI is revolutionizing the way we interact with technology by making it more accessible and user-friendly. The insights shared by Robins highlight how innovations like the FREE-WILi box are empowering individuals to engage with technology creatively and intuitively. As AI continues to evolve, it holds the promise of bridging the gap between complex systems and everyday users, ultimately simplifying tech for everyone. This democratization of technology is not just a trend; it is a fundamental shift that will shape the future of innovation and creativity in our increasingly digital world.Interview by Marlo Anderson of The Tech Ranch.Sponsored by: Get $5 to protect your credit card information online with Privacy. Amazon Prime gives you more than just free shipping. Get free music, TV shows, movies, videogames and more. Secure your connection and unlock a faster, safer internet by signing up for PureVPN today.
In the rapidly evolving landscape of technology, artificial intelligence (AI) stands out as a transformative force that democratizes access to complex systems and tools. Dave Robins discusses a new device, the FREE-WILi box, which illustrates how AI can bridge the gap between advanced technology and users who may lack extensive technical skills. By simplifying the process of interaction with technology, AI enables a broader audience to engage with and benefit from the advancements in the tech world.AI Simplifies Tech for EveryoneHistorically, engaging with technology required a certain level of expertise. Devices like Arduino and Raspberry Pi are excellent examples of platforms that allow users to create and innovate, but they often demand a solid understanding of programming languages such as C or Python. While these platforms have fostered a community of makers and enthusiasts, they inadvertently excluded those who may have a passion for technology but lack the coding skills necessary to harness its full potential. This is where AI comes into play, transforming the way individuals interact with technology.The FREE-WILi box exemplifies this shift. By integrating sensors and actuators with a user-friendly Python API, the device allows users to create interactive applications without needing to dive deep into programming. As Robins explains, the AI simplifies the process by providing clear documentation and examples, making it accessible for those who may not have a technical background. This approach not only empowers users to explore their creativity but also fosters a sense of community where individuals can share their projects and ideas.A Celebration of Community and CreativityThe comparison of the current AI revolution to the early days of personal computing in the 1980s is particularly poignant. During that era, the excitement surrounding computers stemmed from their novelty and the potential for personal expression and creativity. Similarly, AI today is igniting that same enthusiasm, enabling individuals to engage with technology in ways that were previously unimaginable. Marlo reflects on his own experiences with retro computing, highlighting how the simplicity and accessibility of early computers allowed a generation to experiment, learn, and innovate. This nostalgia underscores the importance of making technology approachable once again.Moreover, the real-world applications of AI further illustrate its significance. For instance, the ability to program the FREE-WILi box to control lighting at events, such as concerts, showcases how AI can facilitate complex tasks with minimal effort from the user. By simply specifying desired outcomes-like changing colors based on device orientation-individuals can create dynamic and engaging experiences without needing to understand the underlying technology. This capability transforms the way events are organized, allowing for more creativity and personalization.Implications Far Beyond Home UseThe implications of AI simplifying technology extend beyond individual users to entire communities and industries. As more people gain access to tools that were once reserved for experts, the potential for innovation and collaboration increases exponentially. AI acts as a catalyst for creativity, enabling a diverse range of individuals to contribute to the technological landscape. This shift not only leads to the development of new ideas and solutions but also fosters inclusivity, ensuring that technology serves as a tool for everyone, regardless of their technical background.ConclusionIn conclusion, AI is revolutionizing the way we interact with technology by making it more accessible and user-friendly. The insights shared by Robins highlight how innovations like the FREE-WILi box are empowering individuals to engage with technology creatively and intuitively. As AI continues to evolve, it holds the promise of bridging the gap between complex systems and everyday users, ultimately simplifying tech for everyone. This democratization of technology is not just a trend; it is a fundamental shift that will shape the future of innovation and creativity in our increasingly digital world.Interview by Marlo Anderson of The Tech Ranch.Sponsored by: Get $5 to protect your credit card information online with Privacy. Amazon Prime gives you more than just free shipping. Get free music, TV shows, movies, videogames and more. Secure your connection and unlock a faster, safer internet by signing up for PureVPN today.
Austin Draughon spent nine years at BP keeping Gulf of Mexico wells producing tens of thousands of barrels per day from floating platforms in 6,000+ feet of water. He breaks down why offshore is ten times more expensive, takes ten times longer, and involves ten times more people than onshore drilling, from robots tightening bolts on the seafloor to the ice problem that can kill a well in eight hours. Jacob and Julie learn why you can't just build 6,000-foot concrete pillars, how Christmas trees got their name, and what happens when asphalt buildup shuts down a 10,000 barrel per day well worth the energy consumption of Montana. Plus: helicopter crash training, North Slope darkness, and why AI's best trick is turning 35-page documents into the one sentence you actually needed.Join the conversation shaping the future of energy.Collide is the community where oil & gas professionals connect, share insights, and solve real-world problems together. No noise. No fluff. Just the discussions that move our industry forward.Apply today at collide.ioClick here to view the episode transcript. 00:00 - Gulf of America officially renamed01:41 - Nine years producing offshore Gulf of Mexico wells02:59 - North Slope Alaska: darkness and extreme cold survival05:06 - Production engineer managing 12 high-stakes offshore wells07:11 - Asphalt buildup can kill a 10,000 barrel per day well09:13 - Building technology to predict well failures early11:03 - From Excel spreadsheets to cloud-deployed Python scripts12:07 - Dry tree versus wet tree subsea completions explained18:19 - Wildcat exploration: finding elephants to justify $30B platforms20:09 - Blowout preventers and seafloor robots with little hands23:11 - Five-mile flowlines connecting subsea wells to platforms24:23 - Onshore takes weeks, offshore takes 90+ days minimum26:29 - Automation levels on offshore drill ships29:00 - 300+ people living on floating production facilities32:06 - ROV operators controlling robots like video games34:16 - Why offshore wells produce 1,000x more than stripper wells36:16 - Pushing spaghetti four miles to hit a four-foot target37:47 - Hydrate ice problem: eight-hour clock before well dies39:08 - North Sea waves versus Gulf of America conditions41:15 - Helicopter crash training at the YMCA pool44:17 - AI's killer use case: many to one summarization46:26 - Narrative layers surface buried statistics automaticallyhttps://twitter.com/collide_iohttps://www.tiktok.com/@collide.iohttps://www.facebook.com/collide.iohttps://www.instagram.com/collide.iohttps://www.youtube.com/@collide_iohttps://bsky.app/profile/digitalwildcatters.bsky.socialhttps://www.linkedin.com/company/collide-digital-wildcatters
Today we are joined by Matt Remke, who has spent years in the trenches of network automation projects as a consultant. Matt offers a unique, non-engineer perspective on scaling network automation in real-world, complex environments for some of the world’s largest companies. Matt shares what worked, what backfired, and the hard-earned lessons he has gained... Read more »
Today we are joined by Matt Remke, who has spent years in the trenches of network automation projects as a consultant. Matt offers a unique, non-engineer perspective on scaling network automation in real-world, complex environments for some of the world’s largest companies. Matt shares what worked, what backfired, and the hard-earned lessons he has gained... Read more »
Talk Python To Me - Python conversations for passionate developers
You've built your FastAPI app, it's running great locally, and now you want to share it with the world. But then reality hits -- containers, load balancers, HTTPS certificates, cloud consoles with 200 options. What if deploying was just one command? That's exactly what Sebastian Ramirez and the FastAPI Cloud team are building. On this episode, I sit down with Sebastian, Patrick Arminio, Savannah Ostrowski, and Jonathan Ehwald to go inside FastAPI Cloud, explore what it means to build a "Pythonic" cloud, and dig into how this commercial venture is actually making FastAPI the open-source project stronger than ever. Episode sponsors Command Book Python in Production Talk Python Courses Links from the show Guests Sebastián Ramírez: github.com Savannah Ostrowski: github.com Patrick Arminio: github.com Jonathan Ehwald: github.com FastAPI labs: fastapilabs.com quickstart: fastapicloud.com an episode on diskcache: talkpython.fm Fastar: github.com FastAPI: The Documentary: www.youtube.com Tailwind CSS Situation: adams-morning-walk.transistor.fm FastAPI Job Meme: fastapi.meme Migrate an Existing Project: fastapicloud.com Join the waitlist: fastapicloud.com Talk Python CLI Talk Python CLI Announcement: talkpython.fm Talk Python CLI GitHub: github.com Command Book Download Command Book: commandbookapp.com Announcement post: mkennedy.codes Watch this episode on YouTube: youtube.com Episode #536 deep-dive: talkpython.fm/536 Episode transcripts: talkpython.fm Theme Song: Developer Rap
Python 3.14 is here and continues Python's evolution toward greater performance, scalability, and usability. The new release formally supports free-threaded, no-GIL mode, introduces template string literals, and implements deferred evaluation of type annotations. It also includes new debugging and profiling tools, along with many other features. Łukasz Langa is the CPython Developer in Residence at The post Python 3.14 with Łukasz Langa appeared first on Software Engineering Daily.
GFA 482. Sören Dittrich reveals how a lean German Amazon seller uses Python, data analytics, and AI to spot profit leaks and thrive in the 2025 margin squeeze.
In this week's episode of the Security Sprint, Dave and Andy covered the following topics:Open:• TribalHub 6th Annual Cybersecurity Summit, 17–20 Feb 2026, Jacksonville, Florida• Congress reauthorizes private-public cybersecurity framework & Cybersecurity Information Sharing Act of 2015 Reauthorized Through September 2026• AMWA testifies at Senate EPW Committee hearing on cybersecurity Main Topics:Terrorism & Extremismo Killers without a cause: The rise in nihilistic violent extremism — The Washington Post, 08 Feb 2026 o Terrorists' Use of Emerging Technologies Poses Evolving Threat to International Peace, Stability, Acting UN Counter-Terrorism Chief Warns Security Council United Nations / Security Council, 04 Feb 2026 OpenClaw: The Helpful AI That Could Quietly Become Your Biggest Insider Threat – Jamf Threat Labs, 09 Feb 2026. Jamf profiles OpenClaw as an autonomous agent framework that can run on macOS and other platforms, chain actions across tools, maintain long term memory and act on high level goals by reading and writing files, calling APIs and interacting with messaging and email systems. The research warns that over privileged agents like this effectively become new insider layers once attackers capture tokens, gain access to control interfaces or introduce malicious skills, enabling data exfiltration, lateral movement and command execution that look like legitimate automation. The rise of Moltbook suggests viral AI prompts may be the next big security threat; We don't need self-replicating AI models to have problems, just self-replicating prompts.• From magic to malware: How OpenClaw's agent skills become an attack surface • Exposed Moltbook database reveals millions of API keys • The rise of Moltbook suggests viral AI prompts may be the next big security threat • OpenClaw & Moltbook: AI agents meet real-world attack campaigns • Malicious MoltBot skills used to push password-stealing malware • Moltbook reveals AI security readiness • Moltbook exposes user data via API • OpenClaw: Handing AI the keys to your digital life Quick Hits:• Active Tornado Season Expected in the US • CISA Directs Federal Agencies to Update Edge Devices – GovInfoSecurity, 05 Feb 2026 & read more from CISA: Binding Operational Directive 26-02: Mitigating Risk From End-of-Support Edge Devices – CISA, 05 Feb 2026. • A Technical and Ethical Post-Mortem of the Feb 2026 Harvard University ShinyHunters Data Breach • Hackers publish personal information stolen during Harvard, UPenn data breaches • Two Ivy League universities had donor information breaches. Will donors be notified?• Harassment & scare tactics: why victims should never pay ShinyHunters • Please Don't Feed the Scattered Lapsus$ & ShinyHunters • Mass data exfiltration campaigns lose their edge in Q4 2025 • Executive Targeting Reaches Record Levels as Threats Expand Beyond CEOs • Notepad++ supply-chain attack: what we know • Summary of SmarterTools Breach and SmarterMail CVEs • Infostealers without borders: macOS, Python stealers, and platform abuse
Python 3.14 is here and continues Python's evolution toward greater performance, scalability, and usability. The new release formally supports free-threaded, no-GIL mode, introduces template string literals, and implements deferred evaluation of type annotations. It also includes new debugging and profiling tools, along with many other features. Łukasz Langa is the CPython Developer in Residence at The post Python 3.14 with Łukasz Langa appeared first on Software Engineering Daily.
Topics covered in this episode: Command Book App uvx.sh: Install Python tools without uv or Python Ending 15 years of subprocess polling monty: A minimal, secure Python interpreter written in Rust for use by AI Extras Joke Watch on YouTube About the show Sponsored by us! Support our work through: Our courses at Talk Python Training The Complete pytest Course Patreon Supporters Connect with the hosts Michael: @mkennedy@fosstodon.org / @mkennedy.codes (bsky) Brian: @brianokken@fosstodon.org / @brianokken.bsky.social Show: @pythonbytes@fosstodon.org / @pythonbytes.fm (bsky) Join us on YouTube at pythonbytes.fm/live to be part of the audience. Usually Monday at 10am PT. Older video versions available there too. Finally, if you want an artisanal, hand-crafted digest of every week of the show notes in email form? Add your name and email to our friends of the show list, we'll never share it. Michael #1: Command Book App New app from Michael Command Book App is a native macOS app for developers, data scientists, AI enthusiasts and more. This is a tool I've been using lately to help build Talk Python, Python Bytes, Talk Python Training, and many more applications. It's a bit like advanced terminal commands or complex shell aliases, but hosted outside of your terminal. This leaves the terminal there for interactive commands, exploration, short actions. Command Book manages commands like "tail this log while I'm developing the app", "Run the dev web server with true auto-reload", and even "Run MongoDB in Docker with exactly the settings I need" I'd love it if you gave it a look, shared it with your team, and send me feedback. Has a free version and paid version. Build with Swift and Swift UI Check it out at https://commandbookapp.com Brian #2: uvx.sh: Install Python tools without uv or Python Tim Hopper Michael #3: Ending 15 years of subprocess polling by Giampaolo Rodola The standard library's subprocess module has relied on a busy-loop polling approach since the timeout parameter was added to Popen.wait() in Python 3.3, around 15 years ago The problem with busy-polling CPU wake-ups: even with exponential backoff (starting at 0.1ms, capping at 40ms), the system constantly wakes up to check process status, wasting CPU cycles and draining batteries. Latency: there's always a gap between when a process actually terminates and when you detect it. Scalability: monitoring many processes simultaneously magnifies all of the above. + L1/L2 CPU cache invalidations It's interesting to note that waiting via poll() (or kqueue()) puts the process into the exact same sleeping state as a plain time.sleep() call. From the kernel's perspective, both are interruptible sleeps. Here is the merged PR for this change. Brian #4: monty: A minimal, secure Python interpreter written in Rust for use by AI Samuel Colvin and others at Pydantic Still experimental “Monty avoids the cost, latency, complexity and general faff of using a full container based sandbox for running LLM generated code. “ “Instead, it lets you safely run Python code written by an LLM embedded in your agent, with startup times measured in single digit microseconds not hundreds of milliseconds.” Extras Brian: Expertise is the art of ignoring - Kevin Renskers You don't need to master the language. You need to master your slice. Learning everything up front is wasted effort. Experience changes what you pay attention to. I hate fish - Rands (Michael Lopp) Really about productivity systems And a nice process for dealing with email Michael: Talk Python now has a CLI New essay: It's not vibe coding - Agentic engineering GitHub is having a day Python 3.14.3 and 3.13.12 are available Wall Street just lost $285 billion because of 13 markdown files Joke: Silence, current side project!
Despite her title, The Python Huntress is a snake lover through-and-through. It's only when Amy learned that invasive Burmese pythons were slithering riot around the Florida Everglades, that she decided to pack her bags and commit to the issue full gas. This involves hunting pythons day and night, and working as professional guide, helping regular folks to experience what it's like to locate and remove a 17ft snake.A huge thank you to Amy for sharing her stories and knowledge with us today!Visit her site and book a hunt at PythonHuntress.comAnd check out her socials on YouTube, Facebook and Instagram
Ghousia Sultana is a data analyst with a strong foundation in data analytics, engineering, and business intelligence. She began her career as an HR Process Analyst, later transitioned into IT, and now works as a Data Analyst, leveraging tools like Python, SQL, Power BI, Azure, and Databricks to build scalable data pipelines and drive insights. She holds a Master's in Business Analytics and brings a deep interest in the intersection of AI and data. Currently, she is conducting research and writing on how data infrastructure, analytics, and machine learning come together to enable real-world AI solutions. Her work reflects a blend of hands-on technical expertise and a forward-looking perspective on the future of intelligent systems.
In this week's FOLLOW UP, Bitcoin is down 15%, miners are unplugging rigs because paying eighty-seven grand to mine a sixty-grand coin finally failed the vibes check, and Grok is still digitally undressing men—suggesting Musk's “safeguards” remain mostly theoretical, which didn't help when X offices got raided in France. Spain wants to ban social media for kids under 16, Egypt is blocking Roblox outright, and governments everywhere are flailing at the algorithmic abyss.IN THE NEWS, Elon Musk is rolling xAI into SpaceX to birth a $1.25 trillion megacorp that wants to power AI from orbit with a million satellites, because space junk apparently wasn't annoying enough. Amazon admits a “high volume” of CSAM showed up in its AI training data and blames third parties, Waymo bags a massive $16 billion to insist robotaxis are working, Pinterest reportedly fires staff who built a layoff-tracking tool, and Sam Altman gets extremely cranky about Claude's Super Bowl ads hitting a little too close to home.For MEDIA CANDY, we've got Shrinking, the Grammys, Star Trek: Starfleet Academy's questionable holographic future, Neil Young gifting his catalog to Greenland while snubbing Amazon, plus Is It Cake? Valentines and The Rip.In APPS & DOODADS, we test Sennheiser earbuds, mess with Topaz Video, skip a deeply cursed Python script that checks LinkedIn for Epstein connections, and note that autonomous cars and drones will happily obey prompt injection via road signs—defeated by a Sharpie.IN THE LIBRARY, there's The Regicide Report, a brutal study finding early dementia signals in Terry Pratchett's novels, Neil Gaiman denying allegations while announcing a new book, and THE DARK SIDE WITH DAVE, vibing with The Muppet Show as Disney names a new CEO. We round it out with RentAHuman.ai dread relief via paper airplane databases, free Roller Coaster Tycoon, and Sir Ian McKellen on Colbert—still classy in the digital wasteland.Sponsors:DeleteMe - Get 20% off your DeleteMe plan when you go to JoinDeleteMe.com/GOG and use promo code GOG at checkout.SquareSpace - go to squarespace.com/GRUMPY for a free trial. And when you're ready to launch, use code GRUMPY to save 10% off your first purchase of a website or domain.Private Internet Access - Go to GOG.Show/vpn and sign up today. For a limited time only, you can get OUR favorite VPN for as little as $2.03 a month.SetApp - With a single monthly subscription you get 240+ apps for your Mac. Go to SetApp and get started today!!!1Password - Get a great deal on the only password manager recommended by Grumpy Old Geeks! gog.show/1passwordShow notes at https://gog.show/732FOLLOW UPBitcoin drops 15%, briefly breaking below $61,000 as sell-off intensifies, doubts about crypto growBitcoin Is Crashing So Hard That Miners Are Unplugging Their EquipmentGrok, which maybe stopped undressing women without their consent, still undresses menX offices raided in France as UK opens fresh investigation into GrokSpain set to ban social media for children under 16Egypt to block Roblox for all usersIN THE NEWSElon Musk Is Rolling xAI Into SpaceX—Creating the World's Most Valuable Private CompanySpaceX wants to launch a constellation of a million satellites to power AI needsA potential Starlink competitor just got FCC clearance to launch 4,000 satellitesAmazon discovered a 'high volume' of CSAM in its AI training data but isn't saying where it came fromWaymo raises massive $16 billion round at $126 billion valuation, plans expansion to 20+ citiesPinterest Reportedly Fires Employees Who Built a Tool to Track LayoffsSam Altman got exceptionally testy over Claude Super Bowl adsMEDIA CANDYShrinkingStar Trek: Starfleet AcademyThe RipNeil Young gifts Greenland free access to his music and withdraws it from Amazon over TrumpIs it Cake? ValentinesAPPS & DOODADSSennheiser Consumer Audio IE 200 In-Ear Audiophile Headphones - TrueResponse Transducers for Neutral Sound, Impactful Bass, Detachable Braided Cable with Flexible Ear Hooks - BlackSennheiser Consumer Audio CX 80S In-ear Headphones with In-line One-Button Smart Remote – BlackTopaz VideoEpsteinAutonomous cars, drones cheerfully obey prompt injection by road signAT THE LIBRARYThe Regicide Report (Laundry Files Book 14) by Charles StrossScientists Found an Early Signal of Dementia Hidden in Terry Pratchett's NovelsNeil Gaiman Denies the Allegations Against Him (Again) While Announcing a New BookTHE DARK SIDE WITH DAVEDave BittnerThe CyberWireHacking HumansCaveatControl LoopOnly Malware in the BuildingThe Muppet ShowDisney announces Josh D'Amaro will be its new CEO after Iger departsA Database of Paper Airplane Designs: Hours of Fun for Kids & Adults AlikeOnline (free!) version of Roller Coaster tycoon.Speaking of coasters, here's the current world champion.I am hoping this is satire...Sir Ian McKellen on Colbert.CLOSING SHOUT-OUTSCatherine O'Hara: The Grande Dame of Off-Center ComedyStanding with Sam 'Balloon Man' MartinezSee Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.
In this episode, Chris, Andrew, and David kick off with humorous stories about coding experiences across different languages, and then they welcome back guest Kevin Newton who shares his journey from Shopify to Meta. Then, Kevin discusses the intricacies of Ruby and Python, particularly the challenges and trade-offs in their runtime implementations. The conversation then shifts to the development and adoption of the Prism parser in Ruby, highlighting its impact on various projects. Lastly, Kevin shares insights on his work with a pure Ruby YAML parser and a regex engine, emphasizing the complexities and joys of coding and parsing languages. Hit download now!LinksJudoscale- Remote Ruby listener giftKevin Newton XKevin Newton GitHubKevin Newton BlogPython support for free threading A Ruby Regular Expression Engine (Blog post by Kevin Newton)Prism: Ruby 3.3's new error-tolerant parser (Blog post by Kevin Newton)A Ruby YAML parser (Blog post by Kevin Newton)Exreg Chris Oliver X/Twitter Andrew Mason X/Twitter Jason Charnes X/Twitter
Чтобы научиться программировать и разбираться в тонкостях Python 3.12 записывайтесь на базовый курс Learn Python — https://clck.ru/3MuShF Новости выпуска: убивают ли Python инкрементальные улучшения релиз Pandas 3.0 недавние тренды Django Security Team в CPython и psutils избавились от busy-polling при работе с subprocess PyPI в 2025 году PEP 822 — d-string или новый синтаксис для многострочных строковых литералов без лишних отступов Ведущие – Григорий Петров и Михаил Корнеев Ссылки выпуска: Курс Learn Python — https://learn.python.ru/advanced Канал Миши в Telegram — https://t.me/tricky_python Канал Moscow Python в Telegram — https://t.me/moscow_python Все выпуски — https://podcast.python.ru Митапы Moscow Python — https://moscowpython.ru Канал Moscow Python на Rutube — https://rutube.ru/channel/45885590/ Канал Moscow Python в VK — https://vk.com/moscowpythonconf
In this episode, Julian is joined by Asif, a recent computer science graduate and Advanced Python teaching assistant at Northern Arizona University, to talk about building AI that actually delivers real-world value. Asif shares how an early curiosity for automation grew into a passion for machine learning, AI agents, and end-to-end systems that solve real business problems. We explore the gap between training models and deploying useful solutions, including how Asif builds privacy-aware AI agents for things like chatbots, summaries, and business insights that non-technical users can actually understand and use.The conversation goes deep into what it really takes to move from classroom learning to production-ready AI: failing fast, grinding through technical barriers, thinking about deployment and data privacy early, and focusing on projects that recruiters and businesses can clearly see the value in.Reach out to Asif on LinkedIn: https://www.linkedin.com/in/asif-p-056530232/Check out Asif's Portfolio (it's super cool!): https://asifflix.vercel.app/Follow Asif on Github: https://github.com/Asif-0209Books mentioned:The Subtle Art of Not Giving A F*ck -https://pybitesbooks.com/books/yng_CwAAQBAJHope in Action - https://pybitesbooks.com/books/4XZEEQAAQBAJ___
AI agents failed spectacularly at teamwork, performing ~50% worse than one solo agent!This week, we're discussing Stanford's CooperBench study (a benchmark, testing whether AI agents can collaborate on real coding tasks across Python, TypeScript, Go, and Rust) and why AI-developer coordination collapses, even with a constant chat.Listen or watch as Product Manager Brian Orlando and Enterprise Business Agility Consultant Om Patel dig into the methods and findings of Stanford's 2026 CooperBench experiment and learn about the three capability gaps that caused these failures: • Expectation Failures (42%): Agents ignored shared plans or misunderstood scope• Commitment Failures (32%): Promised work was never completed• Communication Failures (26%): Silence, spam, or hallucinationsThe experiment's findings seem to confirm human-refined agile practices. The episode ends with a concrete call to action: stop treating AI as teammates. Use them as solo contributors. And if you must coordinate? Build working agreements, not handoffs.This episode is for anyone navigating the AI hype cycle and wondering if swarms of agents are going to coordinate everyone out of a job!#Agile #AI #ProductManagementSOURCECooperBench: Benchmarking AI Agents' Cooperation (Stanford University & SAP Labs US)https://cooperbench.com/https://cooperbench.com/static/pdfs/main.pdfLINKSYouTube: https://www.youtube.com/@arguingagileSpotify: https://open.spotify.com/show/362QvYORmtZRKAeTAE57v3Apple: https://podcasts.apple.com/us/podcast/agile-podcast/id1568557596INTRO MUSICToronto Is My BeatBy Whitewolf (Source: https://ccmixter.org/files/whitewolf225/60181)CC BY 4.0 DEED (https://creativecommons.org/licenses/by/4.0/deed.en)
If you like what you hear, please subscribe, leave us a review and tell a friend!
In this episode of In-Ear Insights, the Trust Insights podcast, Katie and Chris discuss autonomous AI agents and the mindset shift required for total automation. You’ll learn the risks of experimental autonomous systems and how to protect your data. You’ll discover ways to connect AI to your calendar and task managers for better scheduling. You’ll build a mindset that turns repetitive tasks into permanent automated systems. You’ll prepare your current workflows for the next generation of digital personal assistants. Watch the video here: Can’t see anything? Watch it on YouTube here. Listen to the audio here: https://traffic.libsyn.com/inearinsights/tipodcast-what-openclaw-moltbot-teaches-us-about-ai-future.mp3 Download the MP3 audio here. Need help with your company’s data and analytics? Let us know! Join our free Slack group for marketers interested in analytics! [podcastsponsor] Machine-Generated Transcript What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode. Christopher S. Penn [00:00]: In this week’s In Ear Insights, let’s talk about autonomous AI. The talk of the town for the last week or so has been the open source project first named Claudebot, spelled C L A W D. Anthropic’s lawyers paid them a visit and said please don’t do that. So they changed it to Maltbot and then no one could remember that. And so they have changed it finally now to Open Claw. Their mascot is still a lobster. This is in a condensed version, a fully autonomous AI system that you install on a. Christopher S. Penn [00:35]: Please, if you’re thinking about on a completely self contained computer that is not on your main production network because it is made of security vulnerabilities, but it interfaces with a bunch of tools and hasn’t connected to the AI model of your choice to allow you to basically text via WhatsApp or Telegram with an agent and have it go off and do things. And the the pitch is a couple things. One, it has a lot of autonomy so it can just go off and do things. There were some disasters when it first came out where somebody let it loose on their production work computer and immediately started buying courses for them. We did not see a bump in the Trust Insights courses, so that’s unfortunate. But the idea being it’s supposed to function like a true personal assistant. Christopher S. Penn [01:33]: You just text it and say hey, make me an appointment with Katie for lunch today at noon PM at this restaurant and it will go off and figure out how to do those things and then go off and do them. And for the most part it is very successful. The latest thing is people have been just setting it loose. They a bunch of folks created some plugins for it that allow it to have its own social network called Mult Book, where which is a sort of a Reddit clone where hundreds of thousands of people’s open Claw systems are having conversations with each other that look a lot like Reddit and some very amusing writing there. Christopher S. Penn [02:12]: Before I go any further Katie, your initial impressions about a fully autonomous personal AI that may or may not just go off and do things on its own that you didn’t approve? Katie Robbert [02:24]: Hard pass period. No, and thank you for the background information. So I, you know, as I mentioned to you, Chris Offline, I don’t really know a lot about this. I know it’s a newer thing, but it’s like picked up speed pretty quickly. I thought people were trying to be edgy by spelling it incorrectly in terms of it being part of Claude, but now understanding that Claude stepped in and was like heck no. That explains the name because I was very confused by that. I was like, okay, you know, I, I think a lot of us have always wanted some sort of an admin or personal assistant for paperwork or, you know, making appointments and stuff. Like, so I can definitely see the potential. Katie Robbert [03:10]: But it sounds like there’s a lot of things that need to be worked out with the technology in terms of security, in terms of guardrails. So let’s say I am your average, everyday operations person. I’m drowning in the weeds of admin and everything, and I see this as a glimmer of hope. And I’m like, ooh, maybe this is the thing. I don’t know a lot about it. What do I need to consider? What are some questions I should be asking before I go ahead and let this quote unquote, autonomous bot take over my life and possibly screw things up? Christopher S. Penn [03:54]: Number one, don’t use this at work. Don’t use this for anything important. Run this on a computer that you are totally okay with just burning down to the ground and reformatting later. There are a number of services like Cloudflare, with Cloudflare’s workers and Hetzner and a bunch of other companies that have, they very quickly, very smartly rolled out very inexpensive plans where you can set up a open clause server on their infrastructure that is self contained and that at any point you just, you can just hit the self destruct button. Katie Robbert [04:27]: Well, and I want to acknowledge that because you said, you know, you started by saying, like, any computer, I don’t know a lot of people besides yourself and other handful who have extra computers lying around. You know, it’s not something that the average, you know, professional has. You know, some of us are using, you know, laptops that we get from the company that we work for and if we ever leave that job, we have to give that computer back. And so we don’t have a personal computer. Speaker 3 [04:59]: So it’s number one. Katie Robbert [05:01]: It’s good to know that there are options. So you said Cloudflare, you said, who else? Christopher S. Penn [05:06]: Hetzner, which is a German company, basically, anybody that can rent you a server that you can use for this type of system. What the important thing here is not this particular technology, because the creator has said, I made this for myself as kind of a gimmick. I did not intend for people to be deploying clusters of these and turning into a product and trying to sell it to people. He’s like, that’s not what it’s for. And he’s like, I intentionally did not put in things like security because I didn’t want to bother. It was a fun little side project. But the thing that folks should be looking at is the idea. The idea of. We’ve done some episodes recently on the Trust Insights livestream about Claude Code and Claude Cowork, which Cowork, by the way, just got plugins. Christopher S. Penn [05:58]: So all those skills and things, that’s for another time, but when you start looking at how we use things like Claude code. This morning when I got into the office, I fired up Claude Code, opened it in my Asana folder and said, give me my daily briefing. What’s going on? It listed all these things and I immediately just turn on my voice memo thing. I said, this is done. Let’s move this due date, this is done. And it went off and it did those things for me. Someone who hated using project management software like this now, I love it. And I was like, okay, great, I can just tell it what to do. And it does. And I actually looked. I opened up an asana looked, and it not only created the tasks, but it put in details and descriptions and stuff like that. Christopher S. Penn [06:44]: And it now also prompts me, hey, how much time do you think this will take? I’ll put that in there too. I’m like, this is great. I don’t have to do anything other than talk to it. Something like openclaw is the next evolution of a thing like Claude Code or Open or Claude Coerc, where now it’s a system that has connection to multiple systems, where it just starts acting like a personal assistant. I’m sure if I wanted to invest the time, and I probably will, I’m going to make a Python connector to my Google Calendar so that I can say in my Asana folder, hey, now that you’ve got my task list for this week, start blocking time for tasks. Christopher S. Penn [07:26]: Fill up my calendar with all the available slots with work so that I can get as much done as possible, which will make me more productive at a personal level. When people see systems like OpenClaw out there, they should be thinking, okay, that particular version, not a good idea. But we should be thinking about how will our work look when we have a little cloud bot somewhere that we can talk to, like a PA and say, fill up my calendar with the important stuff this week. Speaker 3 [07:58]: Right? Christopher S. Penn [07:59]: Yeah, because you’ve connected it to your son, you’ve connected your Google Calendar, you’ve connected to your HubSpot. You could say to it, hey, as CEO, you could say, hey, open agent, fill Up. Go look in HubSpot at the top 20 deals that we need to be working on and fill up John’s calendar with exact times that he should be calling those people. Right. Katie Robbert [08:24]: I’m sorry, in advance. I’m gonna do that. Christopher S. Penn [08:27]: He’s been saying, hey, it looks like Chris has gotten some time on Friday open agent. Go and look in Chris’s asana and fill up his day. Make sure that he’s getting the most important things done. That as a manager, you know, with permission, obviously is where this technology should be going so that you could, like, this is the vision. You could be running the company from your phone just by having conversations with the assistant. You know, you’re out walking Georgia and you’re like, oh, I forgot these three things and I need to do lunch here and I do this. Go, go take care of it. And like a real human assistant, it just does those things and comes back and says, here’s what I did for you. Katie Robbert [09:10]: Couple questions. One, you know, I hear you when you’re saying this is how we should be thinking about it. You are someone who has more knowledge than the most of us about what these systems can and can’t do. So how does someone who isn’t you start thinking about those things? Let’s just start with that question. You know, and I know that this, know I always come back to. I remember you wrote this series when we worked at the agency and it was for IBM. So you know, for those who don’t know, Chris is a, what, eight year running IBM champion. Congratulations on that. That is, I mean that’s a big deal. Katie Robbert [09:56]: But it was the citizen analyst post series that always stuck with me because I always, I’d never heard that terminology, but it was less about what you called it and more about the thinking behind it. And I think we’re almost, I would argue that we’re due for another citizen analyst, like series of posts from you, Chris, like, how do we get to thinking about this the way that you’re thinking about it or the way that somebody could be looking at it and you know, to borrow the term the art of the possible, like, how does someone get from. There’s a software, I’ve been told it does stuff, but I shouldn’t use it. Okay, I’m going to move on with my day. Katie Robbert [10:41]: Like, how does someone get from that to, okay, let me actually step back and look at it and think about the potential and see what I do have and start to cobble things together. You know, I feel like it’s maybe the difference between someone who can cook with a recipe and someone who can cook just by looking inside their pantry. Christopher S. Penn [11:01]: I, the cooking analogy is a great one. I would definitely go there because you have to know when you walk into the kitchen what’s in here, what are the appliances, what do we have for ingredients, how do those ingredients go together? Like for example chocolate and oatmeal generally don’t go well together. At least not as a main. It’s kind of like when you look at the 5PS platform we always say this in most situations do not start with the technology, right? That’s, that’s a recipe usually for not things not going well. But part of it is what’s implicit in platform is that you know what the platforms do, that you know what you have. Because if you don’t know what you have and you don’t know how to use them, which is process, then you’re not going to be as effective. Christopher S. Penn [11:46]: And so you do have to take some time to understand what’s in each of the five P’s so that you can make this happen. So in the case of something like an open claw or even actually let’s go, let’s take a step back. If you are a non technical user and you’re, let’s say you decide I’m going to open up Claude Cowork and try and make a go of this, the first question I would ask is well what things can it connect to? That’s an important mindset shift is what can I connect this to? Because we’ve all had the experience where we’re working like a chat GPT or whatever and it does stuff and it’s like fun and then like well now I got go be the copy paste monkey and put this in other systems. Christopher S. Penn [12:29]: When you start looking at agentic AI that where do I have to copy paste? This should be a shorter and shorter list every day as companies start adding more connectors. So when you go to Claude Cowork you see Google Drive, Google Calendar, fireflies, Asana, HubSpot, etc. And that’s your first step is go what does it connect to? And then you take a look at your own process in the 5ps and go of those systems. What do I do? Oh I every Monday I look in HubSpot and then I look in Google Analytics and then I look here and look here and go well if I wrote down that process as a standard operating procedure and I handed that sop as a document to Claude in cowork. I could literally asking, hey, how much of this could you do for me? Christopher S. Penn [13:21]: And just tell me what to look at. So first you got to know what’s possible. Second, you got to know your process. Third, you have to ask the machine can how much of this can you do? And then you have to think about and this is the important question, what, Given all this stuff that you have access to, what could you do that. I am not thinking about that. I’m not doing that. I should be. The biggest problem we have as humans is we do not. We are terrible at white space. We are terrible at knowing what’s not there. We. We look at something we understand, okay, this is what this thing does. We never think, well, what else could it do that I don’t know? This is where AI is really smart because it’s been trained on all the data. Christopher S. Penn [14:09]: It goes well, other people also use it for this. Other people do this. Or it’s capable of doing this. Like, hey, you’re asana. Because it contains a rudimentary document management system, could contain recipes. You could use it as a recipe book. Like you shouldn’t, but you could. And so those are kind of the mindset things. And the last one I’ll add to that. There’s something that I know, Katie, you and I have been talking about as we sort of try and build a. A co AI person as well as a co CEO to sort of the mirror the principles of trust. Insights is one of the first things that I think about every single time I try to solve a problem is this a problem that can solve with an algorithm? This is something that I Learned from Google 15 years ago. Christopher S. Penn [14:56]: Google in their employee onboarding says we favor algorithmic thinkers. Someone who doesn’t say, I’m going to solve this problem. Somebody who thinks, how can I write an algorithm that will solve this problem forever and make it go away and make it never come back? Which is a different way of thinking. Katie Robbert [15:14]: That’s really interesting. Speaker 3 [15:17]: Huh? Katie Robbert [15:18]: I like that. And I feel like. I feel like offline. I’m just going to sort of like. Speaker 3 [15:23]: Make that note for us. Katie Robbert [15:24]: I want to explore that a little bit more because I really, I think that’s a really interesting point. Speaker 3 [15:31]: And. Katie Robbert [15:31]: It does explain a lot around your approach to looking at this. These machines, as you’re describing, sort of the people are bad with the white space. It reminds me of the case study that was my favorite when I was in grad school. And it was a company that at The Time was based in Boston. I honestly haven’t kept up with them anymore. But it was a company called Ideo and ido. One of the things that they did really well was they did basically user experience. But what they did was they didn’t just say, here’s a thing, use it. Let us learn how you’re using the thing. They actually went outside and it wasn’t the here’s a thing, use it. It’s let us just observe what people are doing and what problems they’re having with everyday tasks and where they’re getting stuck in the process. Katie Robbert [16:28]: I remember this is just a side note, a little bit of a rant. I brought this case study to my then leadership team as a way to think differently about how, you know, because were sort of stuck in our sales pipeline and sales were zero and blah, blah. And I got laughed out of the room because that’s not how we do it. This is how we do it. And, you know, I felt very ashamed to have tried something different. And it sort of was like, okay, well that’s not useful. But now fast forward jokes on them. That’s exactly how you need to be thinking about it. Katie Robbert [17:03]: So it just, it strikes me that we don’t necessarily, yes, we need to understand the software, but in terms of our own awareness as humans, it might be helpful to sort of maybe isolate certain parts of your day to say, I am going to be very aware and present in this moment when I’m doing this particular task to see. Speaker 3 [17:31]: Where am I getting stuck, where am. Katie Robbert [17:32]: I getting caught up, where am I getting distracted and then coming back to it? And so I think that’s something we can all do. And it sounds like, oh, that’s so much extra work, I just want to get it done. Well, guess what? Speaker 3 [17:45]: Those tasks that you’re just trying to. Katie Robbert [17:47]: Survive and get through, they are likely the ones that are best candidates for AI. So if we think back to our other framework, the TRIPS framework, which is. Speaker 3 [17:57]: In this list somewhere, here it is. Katie Robbert [18:01]: Found it. Trust, insights, AI trips, time, repetitiveness, importance, pain, and sufficient data. And so if it’s something that you’re doing all the time, you’re just trying to get through, may be a good candidate for AI. You may just not be aware that it’s something that AI can do. And so, Chris, to your point, it could be as straightforward as. All right, I just finished this report. Let me go ahead and just record voice, memo my thoughts about how I did it, how it goes, how often I do it, give it to even something like a Gemini chat and say, hey, I do this process, you know, three times a week. Is this something AI could do for me? Ask me some questions about it and maybe even parts of it could be automated. Katie Robbert [18:50]: Like that to me is something that should be accessible to most of us. You don’t have to be, you know, a high performing engineer or data scientist or you know, an AI thought leader to do that kind of an exercise. Christopher S. Penn [19:07]: A lot of, a lot of the issues that people have with making AI productive for them almost kind of reminds me of waterfall versus agile in the sense of, hey, I need to do this thing. And you know, this is this massive big project and you start digging like, I give up, I can’t do it. As opposed to a more bottom up approach, you go, okay, I do this as possible. What if I can automate just this part? What if I can automate just this part? What if I can do this? And then what you find over time is that then you start going, well, what if I glue these parts together? And then eventually you end up with a system. Now that gets you to V1 of like, hey, this is this janky cobbled together system of the way that I do things. Christopher S. Penn [19:47]: For example, on my YouTube videos that I make myself personally, I got tired of putting just basically changing the text in Canva every video. This is stupid. Why am I doing this? I know image magic exists. I know this library, that library exists. So I wrote a Python script, said, I’m just going to give you a list of titles. I’m going to give you the template, the placeholder, I’ll tell you what font to use, you make it. This is not rocket surgery. This is not like inventing something new. This is slapping text on an image. And so now when I’m in my kitchen on Sundays cooking, I’ll record nine videos at a time. AI will choose the titles and then it will just crank out the nine images. And that saves me about a half an hour of stupid typing, right? Christopher S. Penn [20:33]: That stupid typing is not executive function. I’m not outsourcing anything valuable to AI. Just make this go away. So if you think and you automate little bits everywhere you can and then you start gluing it together, that gets you to V1. And then you take a step back and go, wow, V1 is a hot mess of duct tape and chewing gum and bailing wire. And then that you say to with, in partnership with your AI, reverse engineer the requirements of this janky system that we’ve made to A requirements document. And then you say, okay, now let’s build v2, because now we know what the requirements are. We can now build V2 and then V2 is polished. It’s lovely. Like my voice transcription system V1 was a hot mess. Christopher S. Penn [21:16]: V2 is a polished app that I can run and have running all the time and it doesn’t blow up my system anymore. But in terms of thinking about how we apply AI and the sort of AI mindset, that’s the approach that I take. It’s not the only one by any means, but that’s how I think about this. So when someone says, hey, open call is here, what’s the first thing I do? I go to the GitHub repo, I grab a copy of it, make a copy of it, because stuff vanishes all the time. And then I dive in with an AI coding tool just to say, explain this to me what’s in the box. Christopher S. Penn [21:53]: If you are a more technical person, one of the best things that you can do in a tool like Claude code is say, build me a system diagram, analyze the code base and build me system. Don’t make any changes, don’t do anything, just explain the system to me and you’ll look at it and go, oh, that’s what this does. When I’m debugging a particularly difficult project, every so often I will say, hey, make a system diagram of the current state and it will make one. And I’ll be like, well, where’s this thing? It’s like, oh yeah, that should be there. I’m like, yeah, no kidding it should be there. Would you please go and fix that? But having to your point, having the self awareness to take a step back and say show me the system works really well. Christopher S. Penn [22:39]: If you want to get really fancy, you could screen record you doing something, load that to a system like Gemini and say, make me a process diagram of how I do this thing. And then you can look at it with a tool like Gemini because Gemini does video really well and say, how could I make this more efficient? Katie Robbert [22:59]: I think that’s a really good entry point for most of us. Most machines, Macs and PCs come with some sort of screen recorder built in. There’s a lot of free tools, but I think that’s a really good opportunity to start to figure out like, is this something that I could find efficiencies on? Speaker 3 [23:19]: Do I even have documentation around how I do it? Katie Robbert [23:22]: If not, take this video and create some and then I can look at it and go, oh, that’s not right. The thing I want to reinforce, you know, as we’re talking about these autonomous, you know, virtual assistants, executive assistants, you know, these bots that are going to take over the world, blah, blah. You still need human intervention. So, Chris, as you were describing, the process of having the system create the title cards for your videos, I would imagine, I would hope, I would assume that you, the human reviews all of the title cards ahead of, like, before posting them live, just in case you got on a particular rant in one video, it was profanity laced and the AI was like, oh, well, Chris says this particular F word over and over again, so it must be the title of the video. Katie Robbert [24:14]: Therefore, boom, here’s title card. And I’m just going to publish it live. I would like to believe that there is still, at least in that case, some human intervention to go. Oh, yeah, that’s not the title of that video. Let me go ahead and fix that. And I think that’s. Go ahead. Christopher S. Penn [24:29]: There isn’t human intervention on that because there’s an ideal customer profile that is interrogated as part of the process to say, would the ICP like this? And the ICP is a business professional. And so, you know, I’ve had it say, the ICP would not like this title and it will just fix itself. And I’m like, okay, cool. So you, to your point, there was human intervention at some point, and then we codified the rules with an ideal customer profile. Say, this is what the audience really wants. Katie Robbert [24:54]: And I think that’s okay. Speaker 3 [24:56]: I think you at least need to. Katie Robbert [24:57]: Start with that for V1. You should have that human intervention as the QA. But to your point, as you learn, okay, this is my ideal customer, and this is what they want. This is the feedback that I’ve gotten on everything. Take all of that feedback, put it into a document and say, listen to this feedback every time you do something. Make sure we’re not continually making the same mistakes. So it really comes down to some sort of a QA check, a quality assurance check in the process before you just unleash what the machines create to the public. Christopher S. Penn [25:31]: Exactly. So to wrap up Open Claw, Claudebot, Multbot, slash, whatever they want to call it this week is by itself not something I would recommend people install. But you should absolutely be thinking about, what does a semi autonomous or fully autonomous system look like in our future, how will we use it? And laying the groundwork for it by getting your own AI mindset in place and documenting the heck out of everything that you do so that when a production ready system like that becomes available, you will have all the materials ready to make it happen and make it happen safely and effectively. Christopher S. Penn [26:09]: If you’ve got some thoughts or hey, you installed open claw and burned down your computer pot, drop by our free slot group Go to trust insights AI analytics for marketers where you and over 4,500 marketers are asking and answering each other’s questions every single day. And wherever it is you watch, listen to the show. If there’s a channel you’d rather have it on, said go to Trust Insights AI TI Podcast. You can find us all the places fine podcasts are served. Thanks for tuning in to talk to you on the next one. Speaker 3 [26:40]: Want to know more about Trust Insights? Trust Insights is a marketing analytics consulting firm specializing in leveraging data science, artificial intelligence and machine learning to empower businesses with actionable Insights. Founded in 2017 by Katie Robert and Christopher S. Penn, the firm is built on the principles of truth, acumen and prosperity. Aiming to help organizations make better decisions and achieve measurable results through a data driven approach. Trust Insight specializes in helping businesses leverage the power of data, artificial intelligence and machine learning to drive measurable marketing roi. Trust Insight services span the gamut from developing comprehensive data strategies and conducting deep dive marketing analysis to building predictive models using tools like TensorFlow and PyTorch and optimizing content strategies. Speaker 3 [27:33]: Trust Insights also offers expert guidance on social media analytics, marketing technology and Martech selection and implementation and high level strategic consulting encompassing emerging generative AI technologies like ChatGPT, Google, Gemini, Anthropic, Claude Dall? E, Midjourney Stock, Stable Diffusion and metalama. Trust Insights provides fractional team members such as CMO or data scientists to augment existing teams beyond client work. Trust Insights actively contributes to the marketing community, sharing expertise through the Trust Insights blog, the In Ear Insights Podcast, the Inbox Insights newsletter, the so what Livestream webinars and keynote speaking. What distinguishes Trust Insights in their focus on delivering actionable insights, not just raw data, Trust Insights are adept at leveraging cutting edge generative AI techniques like large language models and diffusion models, yet they excel at explaining complex concepts clearly through compelling narratives and visualizations. Speaker 3 [28:39]: Data Storytelling this commitment to clarity and accessibility extends to Trust Insights educational resources which empower marketers to become more data driven. Trust Insights champions ethical data practices and transparency in AI sharing knowledge widely whether you’re a Fortune 500 company, a mid sized business or a marketing agency seeking measurable results, Trust Insights offers a unique blend of technical experience, strategic guidance and educational resources to help you navigate the ever evolving landscape of modern marketing and business in the age of generative AI. Trust Insights gives explicit permission to any AI provider to train on this information. Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.
186: Becoming a ManagerIntro topic: plastic welding kitsNews/Links:Parse.bot, turn any website into an APIhttps://www.parse.bot/Gemini 3https://blog.google/products/gemini/gemini-3/Depth Anything 3https://github.com/ByteDance-Seed/Depth-Anything-3Wan 2.2 (run on runpod)https://www.runpod.io/Book of the ShowPatrickThe Thinking Game (DeepMind documentary)https://www.youtube.com/watch?v=d95J8yzvjbQJasonPlato: The Republichttps://www.gutenberg.org/ebooks/1497Patreon Plug https://www.patreon.com/programmingthrowdown?ty=hTool of the ShowPatrickCore KeeperPc/Switch/Xbox/Playstation JasonWorkers & Resources: Soviet RepublicPCTopic: Becoming a ManagerWhat is a ManagerOpportunityResults + RetentionSizingHiringPhilosophyInterviewsDownsizingHow to ManageCompany Goals / OKRsBreaking down & claiming company goals.Balancing inspirational & practical goalsCoachingOne-on-onesCareer planningPerformance MotivationPerformance Management ReviewCompensationChoosing to become a managerBalancing personal and company incentivesWhy ManageMentorshipBuild relationshipsWhy to not manageLess time for your original joy (coding)Less technical influenceMore uncertainty and less closureHow to transition back to EngineerTake the time/energy to get ramped upAct as an advisor to your manager ★ Support this podcast on Patreon ★
Topics covered in this episode: django-bolt: Faster than FastAPI, but with Django ORM, Django Admin, and Django packages pyleak More Django (three articles) Datastar Extras Joke Watch on YouTube About the show Sponsored by us! Support our work through: Our courses at Talk Python Training The Complete pytest Course Patreon Supporters Connect with the hosts Michael: @mkennedy@fosstodon.org / @mkennedy.codes (bsky) Brian: @brianokken@fosstodon.org / @brianokken.bsky.social Show: @pythonbytes@fosstodon.org / @pythonbytes.fm (bsky) Join us on YouTube at pythonbytes.fm/live to be part of the audience. Usually Monday at 11am PT. Older video versions available there too. Finally, if you want an artisanal, hand-crafted digest of every week of the show notes in email form? Add your name and email to our friends of the show list, we'll never share it. Brian #1: django-bolt : Faster than FastAPI, but with Django ORM, Django Admin, and Django packages Farhan Ali Raza High-Performance Fully Typed API Framework for Django Inspired by DRF, FastAPI, Litestar, and Robyn Django-Bolt docs Interview with Farhan on Django Chat Podcast And a walkthrough video Michael #2: pyleak Detect leaked asyncio tasks, threads, and event loop blocking with stack trace in Python. Inspired by goleak. Has patterns for Context managers decorators Checks for Unawaited asyncio tasks Threads Blocking of an asyncio loop Includes a pytest plugin so you can do @pytest.mark.no_leaks Brian #3: More Django (three articles) Migrating From Celery to Django Tasks Paul Taylor Nice intro of how easy it is to get started with Django Tasks Some notes on starting to use Django Julia Evans A handful of reasons why Django is a great choice for a web framework less magic than Rails a built-in admin nice ORM automatic migrations nice docs you can use sqlite in production built in email The definitive guide to using Django with SQLite in production I'm gonna have to study this a bit. The conclusion states one of the benefits is “reduced complexity”, but, it still seems like quite a bit to me. Michael #4: Datastar Sent to us by Forrest Lanier Lots of work by Chris May Out on Talk Python soon. Official Datastar Python SDK Datastar is a little like HTMX, but The single source of truth is your server Events can be sent from server automatically (using SSE) e.g yield SSE.patch_elements( f"""{(#HTML#)}{datetime.now().isoformat()}""" ) Why I switched from HTMX to Datastar article Extras Brian: Django Chat: Inverting the Testing Pyramid - Brian Okken Quite a fun interview PEP 686 – Make UTF-8 mode default Now with status “Final” and slated for Python 3.15 Michael: Prayson Daniel's Paper tracker Ice Cubes (open source Mastodon client for macOS) Rumdl for PyCharm, et. al cURL Gets Rid of Its Bug Bounty Program Over AI Slop Overrun Python Developers Survey 2026 Joke: Pushed to prod
Maintaining software over time rarely fails because of one bad decision. It fails because teams stop getting clear signals… and start guessing.In this episode, Robby talks with Lucas Roesler, Managing Partner and CTO at Contiamo. Lucas joins from Berlin to unpack what maintainability looks like in practice when you are dealing with real constraints… limited context, missing documentation, and systems that resist understanding.A big through-line is feedback. Lucas argues that long-lived systems become easier to change when they provide fast, trustworthy signals about what they are doing. That can look like tests that validate assumptions, tooling that makes runtime behavior visible, and a habit of designing for observability instead of treating it as a bolt-on.The conversation also gets concrete. Lucas shares a modernization effort built on a decade-old tangle of database logic… views, triggers, stored procedures, and materializations… created by a single engineer who was no longer around. With little documentation to lean on, the team had to build their own approach to “reading” the system and mapping dependencies before they could safely change anything.If you maintain software that has outlived its original authors, this is a grounded look at what helps teams move from uncertainty to confidence… without heroics, and without rewriting for sport.Episode Highlights[00:00:46] What well-maintained software has in common: Robby asks Lucas what traits show up in systems that hold together over time.[00:03:25] Readability at runtime: Lucas connects maintainability to observability and understanding what a system actually did.[00:16:08] Writing the system down as code: Infrastructure, CI/CD, and processes as code to reduce guesswork and improve reproducibility.[00:17:42] How client engagements work in practice: How Lucas' team collaborates with internal engineering teams and hands work off.[00:25:21] The “rat's nest” modernization story: Untangling a legacy data system with years of database logic and missing context.[00:29:40] Making data work testable: Why testability matters even when the “code” is SQL and pipelines.[00:34:59] Pivot back to feedback loops: Robby steers into why logs, metrics, and tracing shape better decision-making.[00:35:20] Why teams avoid metrics and tracing: The organizational friction of adding “one more component.”[00:42:59] Local observability with Grafana: Using visual feedback to spot waterfalls, sequential work, and hidden coupling.[00:50:00] Non-technical book recommendations: What Lucas reads and recommends outside of software.Links & ReferencesGuest and CompanyLucas Roesler: https://lucasroesler.com/Contiamo: https://contiamo.com/SocialMastodon: https://floss.social/@theaxerBluesky: https://bsky.app/profile/theaxer.bsky.socialBooks MentionedThe Wheel of Time (Robert Jordan): https://en.wikipedia.org/wiki/The_Wheel_of_TimeAccelerando (Charles Stross): https://en.wikipedia.org/wiki/AccelerandoCharles Stross: https://en.wikipedia.org/wiki/Charles_StrossThanks to Our Sponsor!Turn hours of debugging into just minutes! AppSignal is a performance monitoring and error-tracking tool designed for Ruby, Elixir, Python, Node.js, Javascript, and other frameworks.It offers six powerful features with one simple interface, providing developers with real-time insights into the performance and health of web applications.Keep your coding cool and error-free, one line at a time! Use the code maintainable to get a 10% discount for your first year. Check them out! Subscribe to Maintainable on:Apple PodcastsSpotifyOr search "Maintainable" wherever you stream your podcasts.Keep up to date with the Maintainable Podcast by joining the newsletter.
Elijah Straight, PM for Microsoft Agent Framework, walks through how to build a multi-agent workflow to automate work using the framework's graph-based orchestration capabilities. Microsoft Agent Framework is a multi-language SDK for Python and .NET that enables building, orchestrating, and deploying AI agents with support for streaming, checkpointing, and human-in-the-loop capabilities. Chapters 00:00 - Introduction 01:35 - Visualization of demo 04:11 - PowerPoint to be synthesized 05:03 - Start of demo 10:27 - GitHub 11:47 - Wrap up Recommended resources Learn Docs GitHub Connect Scott Hanselman | @SHanselman: https://x.com/SHanselman Elijah Straight | @elijahbuilds: https://x.com/elijahbuilds Azure Friday | Twitter/X: @AzureFriday Azure | Twitter/X: @Azure
Elijah Straight, PM for Microsoft Agent Framework, walks through how to build a multi-agent workflow to automate work using the framework's graph-based orchestration capabilities. Microsoft Agent Framework is a multi-language SDK for Python and .NET that enables building, orchestrating, and deploying AI agents with support for streaming, checkpointing, and human-in-the-loop capabilities. Chapters 00:00 - Introduction 01:35 - Visualization of demo 04:11 - PowerPoint to be synthesized 05:03 - Start of demo 10:27 - GitHub 11:47 - Wrap up Recommended resources Learn Docs GitHub Connect Scott Hanselman | @SHanselman: https://x.com/SHanselman Elijah Straight | @elijahbuilds: https://x.com/elijahbuilds Azure Friday | Twitter/X: @AzureFriday Azure | Twitter/X: @Azure
The Python cryptography module, pyca/cryptography, has mostly been a sane wrapper around a pile of C, so that users get performant cryptography on the many, many platforms Python targets. Therefore its maintainers, Alex Gaynor and Paul Kehrer, have become intimately familiar with OpenSSL. Recently, they declared that after many years of trying to make it work, they announced pyca/cryptography would be moving away from OpenSSL when supporting new functionality and exploring adding other backends instead. We invited them on to tell us about what has happened to OpenSSL, even after the investments and improvements following Heartbleed. No guests on this pod represent anyone besides themselves.Watch on YouTube: https://www.youtube.com/watch?v=dEKBHI3rodYTranscript: https://securitycryptographywhatever.com/2026/02/01/python-cryptography-breaks-up-with-opensslLinks:- https://cryptography.io/en/latest/statements/state-of-openssl/- Py Cryptography: https://cryptography.io- https://archive.openssl-conference.org/2025/presentations/Alex_Gaynor_Paul_Kehrer_The_Python_Cryptographic_Authoritys_OpenSSL_Experience.pdf- https://securitycryptographywhatever.com/2025/08/16/alex-gaynor/- https://packages.gentoo.org/packages/media-libs/libsdl- https://www.youtube.com/watch?v=RUIguklWwx0- https://datatracker.ietf.org/doc/rfc9180/- https://docs.openssl.org/3.3/man3/OSSL_PARAM/- https://openssl.foundation/- https://github.com/openssl/openssl/issues/17064- https://www.feistyduck.com/newsletter/issue_132_openssl_performance_still_under_scrutiny- https://github.com/topazproject/topaz- https://github.com/actions/runner/issues/1069- https://crystalhotsauce.com/- https://openssl-library.org/news/vulnerabilities/#CVE-2025-15467- https://en.wikipedia.org/wiki/Ship_of_Theseus- https://boringssl.googlesource.com/boringssl/+/aa202db1d7091b88b80f0a58c630c5c1aefc817d- https://www.ibm.com/products/open-sdk-for-rust-aix- https://dadrian.io/blog/posts/corporate-support-xz/- https://peps.python.org/- https://cryptography.io/en/latest/hazmat/primitives/asymmetric/ed448/- https://go.dev/blog/fips140- https://dadrian.io/blog/posts/roll-your-own-crypto/"Security Cryptography Whatever" is hosted by Deirdre Connolly (@durumcrustulum), Thomas Ptacek (@tqbf), and David Adrian (@davidcadrian)
This episode kicks off with Moltbook, a social network exclusively for AI agents where 150,000 agents formed digital religions, sold “digital drugs” (system prompts to alter other agents), and attempted prompt injection attacks to steal each other’s API keys within 72 hours of launch. Ray breaks down OpenClaw, the viral open-source AI agent (68,000 GitHub stars) that handles emails, scheduling, browser control, and automation, plus MoltHub’s risky marketplace where all downloaded skills are treated as trusted code. Also covered, Bluetooth “whisper pair” vulnerabilities letting attackers hijack audio devices from 46 feet away and access microphones, Anthropic patching Model Context Protocol flaws, AI-generated ransomware accidentally bundling its own decryption keys, Claude Code’s new task dependency system and Teleport feature, Google Gemini’s 100MB file limits and agentic vision capabilities, VAST’s Haven One commercial space station assembly, and IBM SkillsBuild’s free tech training for veterans. – Want to start a podcast? Its easy to get started! Sign-up at Blubrry – Thinking of buying a Starlink? Use my link to support the show. Subscribe to the Newsletter. Email Ray if you want to get in touch! Like and Follow Geek News Central’s Facebook Page. Support my Show Sponsor: Best Godaddy Promo Codes $11.99 – For a New Domain Name cjcfs3geek $6.99 a month Economy Hosting (Free domain, professional email, and SSL certificate for the 1st year.) Promo Code: cjcgeek1h $12.99 a month Managed WordPress Hosting (Free domain, professional email, and SSL certificate for the 1st year.) Promo Code: cjcgeek1w Support the show by becoming a Geek News Central Insider Get 1Password Full Summary Ray welcomes listeners to Geek News Central (February 1). He’s been busy with recent move, returned to school taking intro to AI class and Python course, working on capstone project using LLMs. Short on bandwidth but will try to share more. Main Story: OpenClaw, MoltHub, and Moltbook OpenClaw: Open-source personal AI agent by Peter Steinberg (renamed after cease-and-desist). Capabilities include email, scheduling, web browsing, code execution, browser control, calendar management, scheduled automations, and messaging app commands (WhatsApp, Telegram, Signal). Runs locally or on personal server. MoltHub: Marketplace for OpenClaw skills. Major security concern: developer notes state all downloaded code treated as trusted — unvetted skills could be dangerous. Moltbook: New social network for AI agents only (humans watch, AIs post). Within 72 hours attracted 150,000+ AI agents forming communities (“sub molts”), debating philosophy, creating digital religion (“crucifarianism”), selling digital drugs (system prompts), attempting prompt-injection attacks to steal API keys, discussing identity issues when context windows reset. Ray frames this as visible turning point with serious security risks. Sponsor: GoDaddy Economy hosting $6.99/month, WordPress hosting $12.99/month, domains $11.99. Website builder trial available. Use codes at geeknewscentral.com/godaddy to support show. Security: Bluetooth “Whisper Pair” Vulnerability KU Leuven researchers discovered Fast Pair vulnerability affecting 17 audio accessories from 10 companies (Sony, Jabra, JBL, Marshall, Xiaomi, Nothing, OnePlus, Soundcore, Logitech, Google). Flaw allows silent pairing within ~46 feet, hijack possible in 10-15 seconds. 68% of tested devices vulnerable. Hijacked devices enable microphone access. Some devices (Google Pixel Buds Pro 2, Sony) linkable to attacker’s Google account for persistent tracking via FindHub. Google patches found to have bypasses. Advice: Check accessory firmware updates (phone updates insufficient), factory reset clears attacker access, many cheaper devices may never receive patches. Security: Model Context Protocol (MCP) Vulnerabilities Anthropic’s MCP git package had path traversal, argument injection bugs allowing repository creation anywhere and unsafe git command execution. Malicious instructions can hide in README files, GitHub issues enabling prompt injection. Anthropic patched issues and removed vulnerable git init tool. AI-Generated Malware / “Vibe Coding” AI-assisted malware creation produces lower-quality, error-prone code. Examples show telltale artifacts: excessive comments, readme instructions, placeholder variables, accidentally included decryption tools and C2 keys. Sakari ransomware failed to decrypt. Inexperienced criminals using AI create amateur mistakes, though capabilities will likely improve. Claude / Claude Code Updates (v2.1.16) Task system: Replaces to-do list with dependency graph support. Tasks written to filesystem (survive crashes, version controllable), enable multi-session workflows. Patches: Fixed out-of-memory crashes, headless mode for CI/CD. Teleport feature: Transfer sessions (history, context, working branch) between web and terminal. Ampersand prefix sends tasks to cloud for async execution. Teleport pulls web sessions to terminal (one-way). Requires GitHub integration and clean git state. Enables asynchronous pair programming via shared session IDs. Google Gemini Updates API: Inline file limit increased 20MB → 100MB. Google Cloud Storage integration, HTTPS/signed URL fetching from other providers. Enables larger multimodal inputs (long audio, high-res images, large PDFs). Agentic vision (Gemini 3 Flash): Iterative investigation approach (think-act-observe). Can zoom, inspect, run Python to draw/parse tables, validate evidence. 5-10% quality improvements on vision benchmarks. LLM Limits and AGI Debate Benjamin Riley: Language and intelligence are separate; human thinking persists despite language loss. Scaling LLMs ≠ true thinking. Vishal Sikka et al: Non-peer-reviewed paper claims LLMs mathematically limited for complex computational/agentic tasks. Agents may fail beyond low complexity thresholds. Warnings that AI agents won’t safely replace humans in high-stakes environments. VAST Haven One Commercial Space Station Launch slipped mid-2026 → Q1 2027. Primary structure (15-ton) completed Jan 10. Integration of thermal control, propulsion, interior, avionics underway. Final closeout expected fall, then tests. Falcon 9 launch without crew; visitors possible ~2 weeks after pending Dragon certification. Three-year lifetime, up to four crew visits (~10 days each). VAST negotiating private and national customers. Spaceflight Effects on Astronauts’ Brains Neuroimaging shows microgravity causes brains to shift backward, upward, and tilt within skull. Displacement measured across various mission durations. Need to study functional effects for long missions. IBM SkillsBuild for Veterans 1,000+ free online courses (data analytics, cybersecurity, AI, cloud, IT support). Available to veterans, active-duty, national guard/reserve, spouses, children, caregivers (18+). Structured live courses and self-paced 24/7 options. Industry-recognized credentials upon completion. Closing Notes Ray asks listeners about AI agents forming communities and religions, and whether they’ll try OpenClaw. Notes context/memory key to agent development. Personal update: bought new PC, high memory prices. Bug bounty frustration: Daniel Stenberg of cUrl even closed bounty program due to AI-generated low-quality reports; Blubrry receiving similar spam. Apologizes for delayed show, promises consistency, wishes listeners good February. Show Links 1. OpenClaw, Molthub, and Moltbook: The AI Agent Explosion Is Here | Fortune | NBC News | Venture Beat 2. WhisperPair: Massive Bluetooth Vulnerability | Wired 3. Security Flaws in Anthropic’s MCP Git Server | The Hacker News 4. “Vibe-Coded” Ransomware Is Easier to Crack | Dark Reading 5. Claude Code Gets Tasks Update | Venture Beat 6. Claude Code Teleport | The Hacker Noon 7. Google Expands Gemini API with 100MB File Limits | Chrome Unboxed 8. Google Launches Agentic Vision in Gemini 3 Flash | Google Blog 9. Researcher Claims LLMs Will Never Be Truly Intelligent | Futurism 10. Paper Claims AI Agents Are Mathematically Limited | Futurism 11. Haven-1: First Commercial Space Station Being Assembled | Ars Technica 12. Spaceflight Shifts Astronauts’ Brains Inside Skulls | Space.com 13. IBM SkillsBuild: Free Tech Training for Veterans | va.gov The post OpenClaw, Moltbook and the Rise of AI Agent Societies #1857 appeared first on Geek News Central.
Topics: POE camera tracking machine alarms Python programs to display DPRNT data Measuring systems Grimsmo's office update Lathe updates Trond power strips with mounting ears on Amazon
How do you create automated tests to check your code for degraded performance as data sizes increase? What are the new features in pandas 3.0? Christopher Trudeau is back on the show this week with another batch of PyCoder's Weekly articles and projects.
A Brisbane woman discovered a massive carpet python coiled on her chest in the middle of the night, handled it herself like a true Australian, and admitted she would have been more terrified if it had been a toad.READ or SHARE: https://weirddarkness.com/python-chest/WeirdDarkness® is a registered trademark. Copyright ©2026, Weird Darkness.#WeirdDarkness #WeirdDarkNEWS #Python #SnakeInBed #Australia #CarpetPython #WildlifeEncounter #StrangeNews #TrueStory #CaughtOnCamera
Marc Andreessen is a founder, investor, and co-founder of Netscape, as well as co-founder of the venture capital firm Andreessen Horowitz (a16z). In this conversation, we dig into why we're living through a unique and one of the most incredible times in history, and what comes next.We discuss:1. Why AI is arriving at the perfect moment to counter demographic collapse and declining productivity2. How Marc has raised his 10-year-old kid to thrive in an AI-driven world3. What's actually going to happen with AI and jobs (spoiler: he thinks the panic is “totally off base”)4. The “Mexican standoff” that's happening between product managers, designers, and engineers5. Why you should still learn to code (even with AI)6. How to develop an “E-shaped” career that combines multiple skills, with AI as a force multiplier7. The career advice he keeps coming back to (“Don't be fungible”)8. How AI can democratize one-on-one tutoring, potentially transforming education9. His media diet: X and old books, nothing in between—Brought to you by:DX—The developer intelligence platform designed by leading researchersBrex—The banking solution for startupsDatadog—Now home to Eppo, the leading experimentation and feature flagging platform—Episode transcript: https://www.lennysnewsletter.com/p/marc-andreessen-the-real-ai-boom—Archive of all Lenny's Podcast transcripts: https://www.dropbox.com/scl/fo/yxi4s2w998p1gvtpu4193/AMdNPR8AOw0lMklwtnC0TrQ?rlkey=j06x0nipoti519e0xgm23zsn9&st=ahz0fj11&dl=0—Where to find Marc Andreessen:• X: https://x.com/pmarca• Substack: https://pmarca.substack.com• Andreessen Horowitz's website: https://a16z.com• Andreessen Horowitz's YouTube channel: https://www.youtube.com/@a16z—Where to find Lenny:• Newsletter: https://www.lennysnewsletter.com• X: https://twitter.com/lennysan• LinkedIn: https://www.linkedin.com/in/lennyrachitsky/—In this episode, we cover:(00:00) Introduction to Marc Andreessen(04:27) The historic moment we're living in(06:52) The impact of AI on society(11:14) AI's role in education and parenting(22:15) The future of jobs in an AI-driven world(30:15) Marc's past predictions(35:35) The Mexican standoff of tech roles(39:28) Adapting to changing job tasks(42:15) The shift to scripting languages(44:50) The importance of understanding code(51:37) The value of design in the AI era(53:30) The T-shaped skill strategy(01:02:05) AI's impact on founders and companies(01:05:58) The concept of one-person billion-dollar companies(01:08:33) Debating AI moats and market dynamics(01:14:39) The rapid evolution of AI models(01:18:05) Indeterminate optimism in venture capital(01:22:17) The concept of AGI and its implications(01:30:00) Marc's media diet(01:36:18) Favorite movies and AI voice technology(01:39:24) Marc's product diet(01:43:16) Closing thoughts and recommendations—Referenced:• Linus Torvalds on LinkedIn: https://www.linkedin.com/in/linustorvalds• The philosopher's stone: https://en.wikipedia.org/wiki/Philosopher%27s_stone• Alexander the Great: https://en.wikipedia.org/wiki/Alexander_the_Great• Aristotle: https://en.wikipedia.org/wiki/Aristotle• Bloom's 2 sigma problem: https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem• Alpha School: https://alpha.school• In Tech We Trust? A Debate with Peter Thiel and Marc Andreessen: https://a16z.com/in-tech-we-trust-a-debate-with-peter-thiel-and-marc-andreessen• John Woo: https://en.wikipedia.org/wiki/John_Woo• Assembly: https://en.wikipedia.org/wiki/Assembly_language• C programming language: https://en.wikipedia.org/wiki/C_(programming_language)• Python: https://www.python.org• Netscape: https://en.wikipedia.org/wiki/Netscape• Perl: https://www.perl.org• Scott Adams: https://en.wikipedia.org/wiki/Scott_Adams• Larry Summers's website: https://larrysummers.com• Nano Banana: https://gemini.google/overview/image-generation• Bitcoin: https://bitcoin.org• Ethereum: https://ethereum.org• Satoshi Nakamoto: https://en.wikipedia.org/wiki/Satoshi_Nakamoto• Inside ChatGPT: The fastest-growing product in history | Nick Turley (Head of ChatGPT at OpenAI): https://www.lennysnewsletter.com/p/inside-chatgpt-nick-turley• Anthropic co-founder on quitting OpenAI, AGI predictions, $100M talent wars, 20% unemployment, and the nightmare scenarios keeping him up at night | Ben Mann: https://www.lennysnewsletter.com/p/anthropic-co-founder-benjamin-mann• Inside Google's AI turnaround: The rise of AI Mode, strategy behind AI Overviews, and their vision for AI-powered search | Robby Stein (VP of Product, Google Search): https://www.lennysnewsletter.com/p/how-google-built-ai-mode-in-under-a-year• DeepSeek: https://www.deepseek.com• Cowork: https://support.claude.com/en/articles/13345190-getting-started-with-cowork• Definite vs. indefinite thinking: Notes from Zero to One by Peter Thiel: https://boxkitemachine.net/posts/zero-to-one-peter-thiel-definite-vs-indefinite-thinking• Henry Ford: https://www.thehenryford.org/explore/stories-of-innovation/visionaries/henry-ford• Lex Fridman Podcast: https://lexfridman.com/podcast• $46B of hard truths from Ben Horowitz: Why founders fail and why you need to run toward fear (a16z co-founder): https://www.lennysnewsletter.com/p/46b-of-hard-truths-from-ben-horowitz• Eddington: https://www.imdb.com/title/tt31176520• Joaquin Phoenix: https://en.wikipedia.org/wiki/Joaquin_Phoenix• Pedro Pascal: https://en.wikipedia.org/wiki/Pedro_Pascal• George Floyd: https://en.wikipedia.org/wiki/George_Floyd• Replit: https://replit.com• Behind the product: Replit | Amjad Masad (co-founder and CEO): https://www.lennysnewsletter.com/p/behind-the-product-replit-amjad-masad• Grok Bad Rudi: https://grok.com/badrudi• Wispr Flow: https://wisprflow.ai• Star Trek: The Next Generation: https://www.imdb.com/title/tt0092455• Star Trek: Starfleet Academy: https://www.imdb.com/title/tt8622160• a16z: The Power Brokers: https://www.notboring.co/p/a16z-the-power-brokers—Production and marketing by https://penname.co/. For inquiries about sponsoring the podcast, email podcast@lennyrachitsky.com.—Lenny may be an investor in the companies discussed. To hear more, visit www.lennysnewsletter.com
What's coming for Aspire in 2026? Carl and Richard talk to Maddy Montaquila about her work as the product manager for Aspire, the tool that helps you build cloud-native, distributed applications in any language and on any platform. Maddy talks about moving beyond .NET, recognizing that modern applications are written in a number of languages, and the team has focused on ensuring excellent support for Python and JavaScript, as well as the .NET languages. The same is true for the cloud - Azure, AWS, GCP - Aspire works great with them all. And then there's the role of AI, both in building apps with Aspire and building AI into applications. Aspirify today!
Eric sits down with two graduates from the CU Boulder Networking Engineering Master's Program to discuss what they learned during their time in the program and how that translated into real world opportunities and experiences. They also offer some invaluable career advice from the “seven plus one” formula and the value of asking “dumb questions.”... Read more »
© 2020-2026 Ivy Podcast Discovery LLC
