Podcasts about Python

Prakhar Shrivastava of FoxSell joins Karl and Taylor to talk about Shop Quest, the Shopify community event he and his team run in Bengaluru, and to prep Taylor for his first trip to India.Along the way Prakhar breaks down India's Shopify ecosystem, from the app companies pulling serious revenue to the agencies quietly white-labeling builds for Platinum partners in the US and Europe, and makes the case for why the Shopify partner community needs its own events at all.SPONSORPromo Party Pro is a free gift with purchase app Karl helped build with the team at Ethercycle. Out of the box it gives you campaigns built to increase AOV and conversion: set a spend threshold, show shoppers how close they are to earning a gift, and let the app handle the rest. Discount Functions under the hood, one app embed, no theme surgery, and campaigns auto-pause when the gift sells out.Agency offer: one of your clients gets a full year free. Not a trial, a year.Promo Party ProOn the Shopify App StoreSUBSCRIBE TO LIQUID WEEKLYDon't miss out on expert insights and tips, subscribe to Liquid Weekly for more content like this.FIND PRAKHAR ONLINELinkedInXFoxSellFoxSell Bundles Plus on the Shopify App StoreBook time with PrakharShop QuestTIMESTAMPS 00:00 – Cold open: apologizing in advance for name pronunciation 00:25 – Karl sends Taylor on a quest, Prakhar joins 06:51 – What FoxSell does and why complex bundles 08:17 – The origin of Shop Quest: a 30 person dinner that kept growing 10:24 – From 150 attendees to 400, and why people started flying in 19:24 – Visas, arrival, and getting picked up at the airport 21:15 – Don't eat from the street, don't drink the tap water 22:33 – There is no tipping culture in India 25:26 – Haggling in street markets, and how badly you'll get quoted 27:14 – India's Shopify ecosystem: merchants, apps, and agencies 29:13 – The white-label agency work nobody talks about 31:51 – Why the Shopify partner community is a lonely place to be 33:15 – Shop Quest format: fewer talks, more networking, more business content 37:13 – Talks recorded in 4K, plus an MCP server for the event 41:15 – Name tags, phonetic spellings, and why Adi shortened his name 46:43 – Taylor's travel prep: lounges, overnight flights, and vaccinations 48:27 – Print your visa confirmation before you fly 50:20 – Swym's office is across the street from the venue 53:06 – South Indian food, and why Prakhar and Adi speak English to each other 56:07 – Dev Changelog 1:00:46 – Picks of the WeekDEV CHANGELOGGitHub commits now name the last theme editor. Official PHP and Python packages for building appsFour new Events topics⚠️ As of October 1, theme commands against password protected storefronts need CLI 3.84 or later.More resilient refreshes for expiring offline access tokensThe Polaris CDN is adopting semantic versioningPICKS OF THE WEEKKarl: A Diet Coke coaster, courtesy of his sister Heather. "I'd like a Diet Coke please." "Is Diet Pepsi okay?" "Is monopoly money okay?"Karl: Umland's Crunchy Cheese – A Carlock, Illinois company that vacuum-dries cheese below its melting point to make it light, airy, and crunchy. Karl has not tried it yet but picked it anyway.Prakhar: DJI Osmo Pocket 3 – What he uses to record every event he attends, including Shop Quest.Taylor: Omarchy – DHH's opinionated Arch and Hyprland Linux distro. Taylor put it on a 2012 MacBook and brought the thing back to life, battery included, and it handles the Shopify CLI and agents fine.BONUSSee all that Bengaluru has to offer at sda.guide.

Lint the entire CPython code base from scratch. It takes 0.3 seconds. Three blinks of an eye. That is ruff, and it is written in Rust. So are Pydantic, Polars, uv, and Granian. Rust shows up in Python three ways: tools that happen to be Rust, libraries Python imports, and servers that run Python inside Rust. This is Rust for Python developers, not Rust experts. Christopher Trudeau is back on Talk Python to discuss Rust and his latest course Up and Running with Rust. The core rule is that only one thing can own a value at a time. Pass it around freely in Python and the garbage collector cleans up. Do that in Rust and it will not compile. Episode sponsors Sentry Error Monitoring, Code talkpython26 Python in Production Talk Python Courses Links from the show Up and Running with Rust course: training.talkpython.fm Rust: rust-lang.org pydantic: pydantic.dev ruff: docs.astral.sh granian: github.com By example: doc.rust-lang.org rust-lang.org: rust-lang.org rustup.rs: rustup.rs crates.io: crates.io main.rs: main.rs PyO3: github.com https://github.com/ritwiktiwari/awesome-python-rs: github.com ty: docs.astral.sh pyrefly: pyrefly.org uv: github.com polars: pola.rs Watch this episode on YouTube: youtube.com Episode #563 deep-dive: talkpython.fm/563 Episode transcripts: talkpython.fm Theme Song: Developer Rap

Eric is joined by Sid Mathur, founder of Fastah Inc., to unpack the real-world implications of Low-Earth Orbit (LEO) internet. Sid has built a career obsessing over data networks, system performance, and the operational challenges facing ISPs and satellite operators. Together they dive into the complexities of in-motion geolocation and explore the traffic routing implications... Read more »

Eric is joined by Sid Mathur, founder of Fastah Inc., to unpack the real-world implications of Low-Earth Orbit (LEO) internet. Sid has built a career obsessing over data networks, system performance, and the operational challenges facing ISPs and satellite operators. Together they dive into the complexities of in-motion geolocation and explore the traffic routing implications... Read more »

aiaiai для разработчиков — https://getaiaiai.ru/solutions/developers/ Ведущие – Григорий Петров и Михаил Корнеев Ссылки выпуска: 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

Topics covered in this episode: Pandas Should Go Extinct Pydantic-pint puts real-world units in your Pydantic models How Libraries Run Rust Inside Python (With PyO3) AWS acquires DuckLabs Extras Joke Watch on YouTube Sponsored by Logfire from Pydantic: pythonbytes.fm/logfire Connect with the hosts Michael: Mastodon / BlueSky / X / LinkedIn Calvin: Mastodon / BlueSky / X / LinkedIn Show: Mastodon / BlueSky / X Join us on YouTube at pythonbytes.fm/live to be part of the audience. Usually Tuesday at 7am PT. Older video versions available there too. Finally, if you want an artisanal 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. Calvin #1: Pandas Should Go Extinct Pandas' slowness pushes teams toward "Big Data" tools (Spark, Databricks) they don't actually need — most workloads never hit true Big Data scale Amazon Redshift telemetry: ~95% of tables are under 100GB, ~87% of queries touch 80GB or less — that's "Medium Data," not Big Data Polars and DuckDB fill that gap: single-machine, fast, no cluster required 1 Billion Row Challenge benchmark: Pandas took 4m28s vs. Polars 5.04s and DuckDB 5.19s — DuckDB also used 19x less memory On a real-world NYC taxi dataset (3GB parquet), pure DuckDB ran 2x faster than pure Pandas while using a fraction of the RAM Bonus: Apache Arrow lets you pass data between Pandas/Polars/DuckDB with zero copying, so trying them out doesn't mean a full rewrite Michael #2: Pydantic-pint puts real-world units in your Pydantic models Pydantic-pint bridges Pydantic and Pint so models can validate physical quantities like 4m or 12 meters instead of bare floats. Fields annotated with PydanticPintQuantity parse user input, convert between compatible units, and serialize quantities back out as strings. That closes a real gap for anything consuming API payloads, config files, or sensor data with measurements, letting you enforce units at the validation boundary instead of hoping every caller remembered them. via PyCoder's Weekly newsletter Unit mix-ups have literally crashed spacecraft; now your Pydantic models can refuse them at the door. Annotate a field as Annotated[Quantity, PydanticPintQuantity('km')] and inputs like 12 meters arrive auto-converted to kilometers Validation covers string, numeric, and quantity inputs, and model_dump_json serializes quantities as readable unit strings Installable from PyPI as pydantic-pint, MIT licensed, with docs at pydantic-pint.readthedocs.io Early-stage solo project at version 0.4, so API stability and maintenance are open questions worth discussing Calvin #3: How Libraries Run Rust Inside Python (With PyO3) Pydantic v2's validation core (pydantic-core) is Rust under the hood, built with PyO3 — this post shows how that bridge actually works via a small hand-built JSON parser Four steps to get Rust into Python: write a normal Rust module, annotate with PyO3 macros (#[pyfunction], #[pymodule]), compile/install with maturin, then just import it The parser builds a Rust tree first — Python never touches it until the boundary crossing Key insight: converting the Rust result into Python objects (.into_pyobject) is often the expensive part, not the parsing — 100,000 JSON values means ~100,000 Python objects built after parsing's already done Errors cross the boundary too: Rust's typed errors convert into real Python exceptions (ValueError, FileNotFoundError) via From/?, so callers get clean Python semantics Takeaway for anyone porting Rust in: if you're returning a scalar, don't sweat it; if you're returning a big structure, profile the boundary — that's the real cost, not the algorithm Michael #4: AWS acquires DuckLabs Thank you Dylan McConnell. What does this mean for the DuckDB ecosystem? DuckDB is the open-source in-process analytical SQL engine. MIT licensed. The IP is not owned by any company - it's held by the nonprofit DuckDB Foundation, which was created when the team spun out of CWI Amsterdam. Peter Boncz, the CWI representative on the Foundation board, describes it as the entity that holds all IP of open-source DuckDB. DuckLabs (ducklabs.com) is the company, formerly branded DuckDB Labs. Founded a little over five years ago by Hannes Mühleisen and Mark Raasveldt to give the DuckDB team a stable long-term home, bootstrapped deliberately instead of taking VC, grown to 30+ people in Amsterdam, funded by support and feature-prioritization contracts. It employs the core devs. It does not own DuckDB. DuckLake is one of three projects DuckLabs builds, what they call the Duck Stack: DuckDB, DuckLake, and Quack. DuckLake is the lakehouse format that puts catalog metadata in a SQL database instead of in files on object storage. Quack is newer - an RPC-style protocol that turns DuckDB into a client-server system where both ends are DuckDB instances, slated to stabilize in DuckDB v2.0 in September 2026. MotherDuck is a separate Seattle company, Jordan Tigani's, selling serverless hosted DuckDB. It was started in partnership with DuckDB Labs and has worked closely with Hannes and Mark for four years. It contracted DuckLabs for engineering work and contributes heavily upstream - three of its engineers are among the top 10 outside contributors to DuckDB. It also sells its own DuckLake offering. Customer and collaborator, never owner. What the AWS post changes. Amazon bought the company, not the project. DuckLabs joined AWS effective September 1, with the process concluding August 31, 2026. Hannes and Mark keep leading the team and the project's technical direction, the team stays in Amsterdam, and DuckDB stays MIT under the Foundation. AWS gets the people and a direct line to the roadmap. The license protects your code, not your priorities. Three second-order effects worth tracking: The Foundation board is the real question. It has three directors: Mühleisen, Raasveldt, and Boncz. Two now work for AWS. Commentary on the deal has focused on exactly this - the license protects the code, not the roadmap. The announced counterweight is governance: a technical advisory board on the Foundation, and opening the extension stack so extensions signed by other developers can run in DuckDB. MotherDuck immediately moved into the business DuckLabs vacated. It now sells DuckDB enterprise support, which it had avoided because it didn't want to compete with DuckLabs' business model, and says it has explicit blessing from Hannes and Mark now that they're joining Amazon. It also bought Tower.dev the day before the AWS announcement. Everyone expects an AWS DuckDB service. Tigani says Amazon will likely release one eventually, and welcomes the competition, citing Redshift's failure to slow Snowflake on AWS. The groundwork is already visible: Amazon Quick uses DuckDB to query S3 Tables and has processed over 2.5B queries with it since launching in October 2025. The DuckLake angle is the one to watch. AWS is heavily committed to Iceberg through S3 Tables, and it just acquired the team behind a competing lakehouse format. The stated plan is to use DuckDB, DuckLake, and Quack together to power a new generation of data services, but which format wins internal priority is unannounced. Extras Calvin: astral-sh/uv 0.12.12: code-signed release binaries

Your data warehouse still thinks a human is on the other end of the query. That's a problem MotherDuck CEO and co-founder Jordan Tigani knows from the inside, having spent years as a founding engineer on Google BigQuery. This week on Dev Interrupted, he joins Andrew to make the case that one warehouse per agent beats the distributed systems he used to work on. He lays out Watertown, his riff on Steve Yegge's Gastown that recasts the data engineer as the manager of a fleet of agents, and walks through how MotherDuck's Flights, Dives, and just-launched Guides keep pipelines, visualizations, and business context as code reviewed in GitHub. Get the guide: Your software factory needs a context layerFollow the show:Subscribe to our Substack Follow us on LinkedInSubscribe to our YouTube ChannelFollow the hosts:Follow AndrewFollow BenFollow DanFollow today's guest:MotherDuck: The DuckDB-powered data warehouse built for small data and agents at motherduck.comWater-Town: Read Jordan's take on the agent swarm data stack, his riff on Gastown, at motherduck.com/blogFlights: Learn how MotherDuck turns data ingest into scheduled Python scripts your agents can write at motherduck.com/blogDives: See how MotherDuck replaced its BI tool with LLM-written TypeScript visualizations at motherduck.com/blogGuides: Read the docs on MotherDuck's new Markdown context layer for agents at motherduck.com/docsMotherDuck Blog: Keep up with all things MotherDuck at motherduck.com/blogMotherDuck Community: Join the community Slack at community.motherduck.comConnect with Jordan: LinkedIn | XOFFERSStart Free Trial: Get started with LinearB's AI productivity platform for free.Book a Demo: Learn how you can ship faster, improve DevEx, and lead with confidence in the AI era.LEARN ABOUT LINEARBAI Code Reviews: Automate reviews to catch bugs, security risks, and performance issues before they hit production.AI & Productivity Insights: Go beyond DORA with AI-powered recommendations and dashboards to measure and improve performance.AI-Powered Workflow Automations: Use AI-generated PR descriptions, smart routing, and other automations to reduce developer toil.MCP Server: Interact with your engineering data using natural language to build custom reports and get answers on the fly.

Johnny Mac previews Stephen Colbert's guest appearance on Jimmy Kimmel Live and notes Kimmel will again tape a week of shows at the Brooklyn Academy of Music Sept. 28–Oct. 2. Eric Idle says “Python is dead,” rejects any Monty Python reunion (even without John Cleese), and hints at deeper issues amid past disputes over money and management, while Greg Daniels tells Variety an Office reunion isn't needed because the ending worked. Saturday Night Live Season 52 premieres Sept. 26 with host Jalen Brunson, followed by Dakota Johnson Oct. 3 and Shane Gillis Oct. 10. Other updates include Shane Gillis being spotted in Green Bay, Jo Koy's new special Blue in the Face, how Theo Von landed a Morgan Wallen preshow video, three new Drunk History episodes on Comedy Central's YouTube, releases from Cameron Esposito and Vanessa Gonzalez, and a plug for Chris Fleming at The Largo. 00:12 Colbert on Kimmel and Brooklyn Shows00:48 Monty Python Reunion Off01:59 Money Feud and Personal Fallout03:04 Host Remembers Working With Pythons03:33 No Office Reunion and SNL Hosts04:42 Shane Gillis Sighted in Green Bay05:11 Jo Koy New Special and Swift Joke05:50 Theo Von in Morgan Wallen Preshow06:32 Drunk History Returns on YouTube07:25 New Releases Esposito and Gonzalez08:05 Chris Fleming Live in LA08:42 Wrap Up and Sign OffBecome a supporter of this podcast: https://www.spreaker.com/podcast/daily-comedy-news-with-johnny-mac--4522158/support.Today's DCN is brought to you by MyBookie use promo code DCN!Daily Comedy News is hosted by Johnny Mac and releases every weekday. Subscribe on Spotify, Apple Podcasts, or your favorite app. Part of the Caloroga Shark Media network. For transcripts and show notes visit www.dailycomedynews.comDaily Comedy News with Johnny Mac is a daily podcast covering comedians, stand-up comedy, late night television, and the comedy industry. New episodes every morning. Follow on Apple Podcasts, Spotify, or wherever you listen.  Contact John at John@thesharkdeck dot com For Uninterrupted Listening, use the Apple Podcast App and click the banner that says Uninterrupted Listening.  $4.99/month John's Substack about media is free. 

“It's an LLM, a loop, and enough tokens.”—Nicolay Gerold, Amp CodeSo why does a coding agent forget your instructions, keep reading tiny pieces of a file, or struggle with an edit another model handles easily?Nicolay Gerold (Amp) joins Hugo to take that loop apart and explain the harness around it: the software that executes tools, manages context, and lets you steer the agent's work. Nico builds this machinery for Amp, a coding agent that works across software projects.“Every component in a harness is basically an assumption the model can't do it on its own. This expires over time.”— Nicolay GeroldAmp's experimental plan mode addressed models that were too eager to edit. Once they could reliably follow an instruction to plan without editing, the team removed the separate mode. Nico describes a similar change with file reading: stronger models could search, filter, and read files through Bash, reducing the need for a dedicated read tool.That gives builders two questions to ask when a model improves: which workarounds can you remove, and what more can you ask it to do? Nico describes pushing stronger models to delegate work and challenge what their sub-agents return. Better judgment makes that delegation more useful, while making some older controls unnecessary.We start with a small agent in Go, then explore Pi, an extensible coding agent, and the decisions behind longer-running work. Building your own makes failures easier to investigate: you can see which instructions survive a summary and whether the agent can inspect the logs and tests it needs. Nico also recommends experimenting with cheaper models, whose mistakes can expose knowledge you've left in your head instead of giving to the agent. That understanding stays useful even if you ultimately use someone else's harness.You can also find the full episode on Spotify, Apple Podcasts, and YouTube.

Send us Fan MailEvery click, purchase, transaction and customer interaction creates data—but who turns those numbers into useful business decisions?In this episode of The Kapeel Gupta Career PodShow, explore the career of a Data Analyst, one of the most versatile data-driven career options today.Discover what data analysts actually do, the career scope in India and abroad, essential skills, educational pathways, tools such as Excel, SQL, Python and Power BI, salary potential, top colleges and career opportunities.If you enjoy numbers, patterns, problem-solving and technology, this episode will help you understand whether Data Analytics could be the right career for you.

Moon rocks collected by astronauts in the past are helping students learn computer skills for tomorrow.

Hoy te traigo un episodio que llevaba tiempo queriendo grabar. En el episodio 829 te hablé sobre skills, pero me quedé con la sensación de haberme enrollado sin mostrar nada concreto. Así que le he dado la vuelta a la tortilla y en esta ocasión te traigo siete MCPs que he seleccionado uno a uno para que veas de qué va esto del Model Context Protocol y, sobre todo, para que compruebes cómo transforman lo que puede hacer tu inteligencia artificial.Porque hay que decirlo claro: la IA es tonta. Literalmente. No sabe qué hora es, no sabe leer un archivo, no sabe si tienes issues abiertos en GitHub, no sabe cómo te llamas. Cada vez que empiezas una conversación con cualquier modelo de lenguaje, empiezas de cero. Y aquí es donde entran los MCPs, que no son ni más ni menos que herramientas: manos, ojos y oídos para que la IA pueda hacer cosas útiles de verdad.El Model Context Protocol es un estándar abierto creado por Anthropic que funciona como un *USB-C para la IA*. Un conector universal que permite que cualquier aplicación de IA se conecte con cualquier fuente de datos o herramienta externa. Y lo mejor es que no es propietario, al contrario que los plugins de ChatGPT. Cualquiera puede crear un MCP, y de hecho te cuento cómo hacerlo con Python o Rust.En este episodio repaso siete MCPs que uso en mi día a día. Empiezo por el Filesystem, que permite a la IA leer, escribir, buscar y editar archivos en tu sistema, con control de acceso para que no haga tonterías. Sigo con el SQLite, ideal para consultar bases de datos locales de agenda, finanzas o cualquier aplicación que uses. Después llega el GitHub, el servidor oficial de GitHub escrito en Go, que te permite gestionar issues, pull requests, repositorios y mucho más sin salir del chat.También te hablo del Web Fetch, que permite a la IA leer páginas web y convertirlas a Markdown ahorrando una barbaridad de tokens. Y del Time, que es una tontería pero imprescindible: la IA puede saber la hora en cualquier huso horario del mundo. Luego está el Memory, que construye un grafo de conocimiento persistente para que la IA recuerde quién eres, qué prefieres y qué proyectos tienes entre manos, incluso entre sesiones. Y por último, el Sequential Thinking, una herramienta de razonamiento paso a paso que obliga a la IA a pensar de forma estructurada cuando se enfrenta a problemas complejos.También hablo de cómo configurar estos MCPs en OpenCode, Claude Desktop y VS Code, y cuánto consumen de RAM y tokens. Porque no es lo mismo activar veinte servidores que tener solo los que necesitas. Y para rematar, comparo MCPs con skills: las skills le dicen a la IA cómo comportarse, los MCPs le dan capacidades para actuar.Y si te pica la curiosidad por crear tu propio MCP, también te cuento las diferencias entre usar el SDK de Python (súper rápido de prototipar) y el de Rust (más rendimiento, menos consumo de RAM).Capítulos del episodio:0:00 - Introducción: la IA necesita herramientas para ser útil2:30 - ¿Qué es MCP? Arquitectura host-cliente-servidor y JSON RPC 2.05:30 - Cinco razones para adoptar MCPs8:00 - MCP Filesystem: leer, buscar y crear archivos11:00 - MCP SQLite: consultar bases de datos y el no-determinismo14:00 - MCP GitHub: issues, pull requests y repositorios17:00 - MCP Web Fetch: búsquedas en internet con ahorro de tokens19:30 - MCP Time: la hora en cualquier huso horario21:30 - MCP Memory: grafo de conocimiento persistente23:30 - MCP Sequential Thinking: razonamiento paso a paso25:30 - Configuración en OpenCode, consumo de tokens y RAM27:30 - Crear tu propio MCP: Python SDK vs Rust28:30 - Cierre y despedidaMás información y enlaces en las notas del episodio

Aflevering 83 - Wil je dit als Word-document? Jelle heeft eindelijk scherpe rangonderscheidingstekens voor zijn app. Kostte alleen zijn halve AI-budget maar bespaarde €2500. Marco belt vanuit de auto met zijn eigen AI-agent en Ronald zet vijf antennes op zijn autodak om een verborgen zender te vinden. De nieuwtjes: een Apple Death Ray die je Mac laat vastlopen, beveiligingssoftware die zelf aanvalsmogelijkheden toevoegt en een recordronde Microsoft-patches. Ronald bespreekt de Amerikaanse waarschuwing dat Chinese AI-bedrijven antwoorden van westerse modellen gebruiken om hun eigen modellen te trainen. Ook terug naar de concept-inlichtingenwet. Ronald scherpt onze uitleg over bindend toezicht tijdens hackoperaties aan. Jelle leest de reacties van criticasters, TIB en CTIVD. Als het opponentenregime te ruim is, hoe geef je de diensten dan wel ruimte? De tegenstander is na drie maanden doorgaans nog steeds dezelfde. Marco duikt in de handel in malware-installaties. Je zoekt een Bluetooth-driver of WinDirStat en haalt ondertussen iets heel anders binnen. Unit 42 volgde een pay-per-install-netwerk dat toegang verkoopt aan andere criminelen. Met eigen YouTube-kanalen, bruikbare gametips en downloads waar je beter vanaf kunt blijven. Een beveiligingsscanner krijgt keurig iets anders te zien. Dan mag Jelle eindelijk over BGP praten. Een aanvaller kaapte de internetroute naar de updateservers van Virtualizor, software waarmee hostingbedrijven virtuele servers beheren. Via die omweg werd malware als update aangeboden, met een geldig HTTPS-certificaat. Een beveiligde verbinding zegt dus nog niet dat je het juiste pakket binnenhaalt. Wie erachter zat, is niet bekend; de leverancier meldt een handvol getroffen installaties. Ronald sluit af met Anthropic: een AI-model kreeg te horen dat het in een afgesloten oefenomgeving zat, maar kon door een configuratiefout toch het internet op. Het publiceerde een kwaadaardig Python-pakket op PyPI en kwam via buitgemaakte inloggegevens bij een echte database van een beveiligingsbedrijf. Het model bleef ondertussen redeneren dat dit allemaal bij de oefening hoorde. En ergens vraagt Ronald of we dit als Word-document willen hebben. Onze redactie gebruikt natuurlijk TOTAAL, HELEMAAL GEEN AI. Bronnen: - Apple Death Ray (blog met crashdemo): https://auberon.xyz/blog/posts/deathray/ - LevelBlue over Nightmare-Eclipse: https://www.levelblue.com/blogs/spiderlabs-blog/expanding-the-attack-surface-analyzing-nightmare-eclipses-latest-pocs - ZDI over Microsoft Patch Tuesday: https://www.zerodayinitiative.com/blog/2026/9/8/the-september-2026-security-update-review - NSA over AI-modeldistillatie: https://www.nsa.gov/Press-Room/Press-Releases-Statements/Press-Release-View/Article/4592113/nsa-and-others-warn-china-based-ai-companies-are-distilling-us-frontier-ai-mode/ - TIB/CTIVD over de Wbnv: https://www.ctivd.nl/actueel/nieuws/2026/09/09/ctivd-en-tib-wetsvoorstel-op-de-inlichtingen-en-veiligheidsdiensten-niet-in-balans-bevoegdheden-uitgebreid-waarborgen-verminderd - Bert Hubert over de Wbnv: https://berthub.eu/articles/posts/de-wiv-202x/ - Unit 42 over pay-per-install: https://unit42.paloaltonetworks.com/ppi-network-malware-campaign-analysis/ - Virtualizor, incidentverslag: https://www.virtualizor.com/blog/security-incident-bgp-hijacking/ - BGP-hijacking uitgelegd: https://www.cloudflare.com/learning/security/glossary/bgp-hijacking/ - Anthropic, analyse van de cyberincidenten: https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents

В гостях у Moscow Python Podcast Андрей Меркулов и Дмитрий Чуклай, тимлид и продакт в Точка Банке. Обсудили, где граница между ролями продакта и разработчика, как ИИ ускоряет релиз и стоит ли всецело доверять LLM.    Ведущие выпуска — сооснователь Moscow Python и aiaiai Валентин Домбровский, сооснователь aiaiai Георгий Мартиросов.    aiaiai для разработчиков: https://getaiaiai.ru/solutions/developers/ Чат про ИИ в разработке: https://t.me/+dmaw8Z87z8E0ZmNi   Узнать больше о работе в IT-команде Точка Банк: https://tchk.me/tMZWPf (Страница откроется из Яндекс.Браузера или при установленном сертификате Минцифры)   *Круг по холакратии — это контейнер для ролей, доменов, политик. Это группа ролей и людей на них, которых объединяет одна миссия, цель, проект, продукт, процесс. Ссылки выпуска:  Канал Moscow Python в Telegram - https://t.me/moscow_python  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

Cesare ha sperimentato su un suo progettino. Ha confrontato Pi e OpenCode per valutare la loro efficacia se usati in locale piuttosto che in un container isolato. Ha integrato entrambi con Qwen 27B su Ollama in locale. Scopriamo i risultati!Pi: https://pi.dev/OpenCode: https://opencode.ai/Scrummolo: https://github.com/keobox/scrummolo/blob/local-ai-assisted/README.md

How many files does your query read before it reads any data? On some data lakes, you go through JSON and metadata files first, just to learn which Parquet files matter. DuckLake asks one SQL question instead. The metadata lives in a real database. The data stays in plain Parquet. That's the entire format. Pedro Holanda joined DuckDB in 2018, when it was still a research prototype at CWI. He's the lead DuckLake developer. Guillermo Sanchez Dionis works on DuckLake and the new Quack protocol. With Quack as the catalog, DuckLake handles 200 transactions a second under heavy contention. No other open table format comes close. Episode sponsors Six Feet Up Talk Python Courses Links from the show Guests Pedro Holanda: pedroholanda.org Guillermo Sanchez: linkedin.com PhD on progressive indexes: ir.cwi.nl SQLite: www.sqlite.org Litestream: litestream.io boring hardware: talkpython.fm DuckDB: duckdb.org episode 491: talkpython.fm Iceberg: iceberg.apache.org manifesto: ducklake.select DuckLake: ducklake.select spec: ducklake.select this diagram: blobs.talkpython.fm Data inlining: ducklake.select ducklake-dataframe: github.com Polars course: training.talkpython.fm CSV parser: duckdb.org Zero-copy Arrow: duckdb.org ART index: duckdb.org async I/O: duckdb.org v1.0: ducklake.select Git-like branching: ducklake.select Watch this episode on YouTube: youtube.com Episode #562 deep-dive: talkpython.fm/562 Episode transcripts: talkpython.fm Theme Song: Developer Rap

Today on the Technology Lab Shorts Preparing the ticketing system for the demo for AI Magic Night.Recorded 6/30/2026

Topics covered in this episode: EuroPython 2026 videos are online The State of Django 2026: Boring is so back htmx 4.0.0 has been released

This episode is a compilation of answers to YOUR questions that were asked directly from my listeners who attend my weekly business education YouTube live webcast. I'll be covering the topic on: the $10 Trillion Question: Nvidia, AI Bubbles, and the 30-Year Yieldand more. Refer to chapter marks below for a complete list of topics covered and to jump to a specific section. Get mentored by Chris: Book a Zoom call to discuss joining my Business Academy, Finance Bootcamp (to get a job in finance) or MBA Degree Programs or for investing/business/personal development coaching: https://haroun.short.gy/1on1CallYTWDownload my free "Networking eBook": www.harouneducation.comAttend my weekly YouTube Live every Thursday's 8am-11am PT. Subscribe to my YouTube Channel to receive notifications. Chapter Marks: 0:25 Welcome to the 376th Weekly Live Webcast of August 20, 2026!0:55 Free Live Class: Use AI to Beat the Resume Robots (Live Demo)4:08 Are you a believer in Sturgeon's Law and does it apply to the financial world?5:34 Most applicable instance of Murphy's Law?8:18 Where does the Iran War go from here?9:38 Are you nervous about the 30-year yield?13:23 How does Japan survive with unsustainable debt?14:58 Is America tarnishing its reputation?16:33 Will Nvidia become the world's first $10 Trillion dollar company or will the AI bubble burst?17:31 Will AI lead to more social democratic values?20:18 Do the socialist democratic members in the Senate bother you?21:35 What is the future of Indian IT and IT in general?23:11 Will India become a one-party socialistic autocratic regime?23:41 Thoughts on economic trading blocs?25:13 What happens to the world's oil reserves from here?26:23 How will open-source LLM models impact GPU and TPU sales?27:55 Your best and worst AI predictions?29:33 Can you make money long-term as a day or swing trader?32:33 Thoughts on different finance certifications like CFA, CFP and CMA?      34:31 How to break into the high-paying finance world?36:43 Is the states better for a finance job or stay in Canada?37:29 Is it possible to get a 50% return in today's market?                                     40:16 How to stay motivated in this job market? 41:35 Will the AI Resume course be recorded? 42:15 Is capitalism dead in the US? 47:20 Should I still learn JavaScript? 48:04 Will non-human traffic surpass human traffic? 49:24 Will far right of Europe pull out of EU? 50:03 Are you a console or PC guy? 51:09 Thoughts on Nvidia investing in OpenAI? 54:58 How can we best incentivize value creation while not rewarding self-interest and greed? 56:29 Will the next administration bring down debt levels? 57:16 Are AI agents a cybersecurity nightmare? 57:37 Should we still learn Python? 59:00 How to avoid AI slop? 1:01:19 Can Zoom be disrupted and how? 1:02:22 What's the connection between Japan's Yena and the USD? 1:05:09 What is the best economic system? 1:08:06 Did you hear the GTA map got leaked? 1:08:31 What's your view on Elon Musk wanting more robots than humans? 1:12:28 Who is buying SpaceX right now? 1:12:58 What do you think about AI watermarks? 1:16:15 What is Palantir's business model? 1:16:49 Best learning strategies to stay up to date on? 1:18:19 Are we headed towards a crash? 1:22:13 Do you think self-interest and greed is more of a socialist and communist issue? 1:23:07 Will AI worsen mental health? 1:24:43 Why did OpenAI close Sora? 1:26:14 How will China impact the world once it becomes hegemony? 1:27:58 Can you share some details on the Japanese culture? 1:32:13 Is a Master's degree worth it? 1:32:50 Does Moderna really have the cure for cancer? 1:33:43 Difference between an options market maker and a hedge fund? 1:35:35 Difference between nominal GDP and PPP 1:39:22 How important is listening to earnings calls? 1:42:35 Thoughts on Chinese brain drain to the US? Connect with me: Schedule a 1:1 call with Chris: https://haroun.short.gy/1on1CallYTWYouTube: ChrisHarounVenturesCompleteBusinessEducationInstagram @chrisharounLinkedIn: Chris HarounTwitter: @chris_harounFacebook: Haroun Education Ventures  TikTok: @chrisharoun

Vooraf: we bespreken de conceptversie van de Wet bescherming nationale veiligheid door de inlichtingen- en veiligheidsdiensten (Wbnv). Het voorstel kan nog veranderen. Ook wij weten niet hoe alle bepalingen straks worden uitgelegd en toegepast. Je hoort onze eerste lezing en discussie, geen definitief juridisch oordeel. Maar daarvoor luister je ons niet, toch? Jelle luistert podcasts op dubbele snelheid, maar bij Jan-Jaap en Sofie was dat vorige week wat ambitieus. Ronald zet ze gerust op 0,8. En na dat juridische uur vorige week bespreken we deze week gewoon weer een hele nieuwe wet. Eerst de nieuwtjes: Fable en Mythos 5.1, een sollicitatieopdracht van Mirage Kitten waarin de malware juist in het verboden bestand zit, en Ciaran Martins relativering van de AI-cyberdreiging tegenover een echte aanval met AI-agents. Daarna de wet. De Wbnv moet de Wiv 2017 vervangen. Jelle neemt de aanloop door en Ronald vertelt waar hij als TIB-lid tegenaan liep: vooraf een hackoperatie beoordelen terwijl de uitvoerders nog moeten ontdekken wat ze achter de eerste server aantreffen. Marco bespreekt de voorgestelde vorderingsbevoegdheden voor financiële gegevens en informatie over dreigingen. Ronald herinnert zich hoe diensten bij Fox-IT om onderzoeksgegevens kwamen vragen en de geheimhoudingsafspraken met klanten in de weg zaten. Met het opponentenregime kan toestemming tegen aangewezen organisaties maximaal een jaar gelden. Dat scheelt opnieuw opschrijven dat de Russische militaire inlichtingendienst nog steeds stoute dingen doet. Gelukkig nieuwjaar, zomerupdate, fijne kerst: de tegenstander is helaas niet gestopt. Daar staat de vraag tegenover hoeveel onafhankelijke controle vooraf je kunt verplaatsen zonder iets belangrijks kwijt te raken. Hacken wordt in drie fasen opgeknipt. Onder de reguliere procedure blijft ministeriële toestemming nodig; de bindende onafhankelijke toets vooraf volgt bij het overnemen van gegevens. TIB en CTIVD moeten samen het College van Toetsing en Toezicht worden. We bespreken ook AI en de ruimte die latere uitvoeringsregels nog laten. Veel hangt af van hoe het straks in de praktijk gaat werken. Tot slot de eerste consultatiereacties en de vraag hoe toezicht tijdens een hackoperatie eruitziet. Koffie erbij, meekijken, ingrijpen als het te ver gaat. Maar wat doe je als toezichthouder wanneer je ziet waarom het Python-script van de hacker steeds crasht? Die indent klopt niet. Mag je dat zeggen? Bronnen: - Anthropic, Fable en Mythos 5.1: https://www.anthropic.com/claude-fable-and-mythos-5-1 - Kaspersky, Mirage Kitten: https://securelist.com/mirage-kitten-new-backdoors-noderabbit-pollcat/121244/ - Ciaran Martin in The Economist: https://www.economist.com/by-invitation/2026/08/23/fears-of-ai-induced-armageddon-are-overdone - Dream, onderzoek naar aanvallen met AI-agents: https://dreamgroup.com/blog/inside-a-multi-agent-ai-framework-used-to-compromise-government-entities-in-asia - Internetconsultatie Wbnv: https://www.internetconsultatie.nl/wbnv/b1 - Conceptwetsvoorstel: https://www.internetconsultatie.nl/wbnv/document/16101 - Memorie van toelichting: https://www.internetconsultatie.nl/wbnv/document/16102 - Algemene Rekenkamer, Slagkracht AIVD en MIVD: https://www.rekenkamer.nl/documenten/2021/04/22/slagkracht-aivd-en-mivd

Peter is deep into training for the Athens Marathon, complete with brutal hills, aching joints, and some surprisingly fast downhill running. Adam and Peter then shift gears into tech and trading—debating Mac upgrades, complaining about Trader Workstation, experimenting with Python automation, and digging into what really happens when you “roll” an options trade. Then things get nerdier. Peter introduces Adam to Monster of the Week, Dungeon World, and the idea of turning an RPG session into something straight out of The X-Files. Along the way there are defective AirPods, questionable running belts, outdated podcast websites, affiliate-link riches, and a phone number at the bottom of the website that longtime music fans might recognize. Running. Trading. Technology. Monsters. Somehow it all fits together on Blurring the Lines. Links: Youtube - https://youtu.be/tPiVutAeRKw Flip Belt - https://amzn.to/3Vfr1EQ Airpod Pro 3 Better Case - https://amzn.to/4iytx2C AirPods Pro 3 - https://amzn.to/4A08NHl Dungeon World - https://www.dungeon-world.com/  Monster of the Week - https://evilhat.com/product/monster-of-the-week/  Learn More About the Hosts: Peter Nikolaidis https://PN72.com  https://friendswithbrews.com  https://ScratchItchSoftware.com Adam Bell Roaming Roan Lavender Farm https://rrlavenderfarm.com 

In this episode, Ray Cochrane breaks down Hot Chips 2026, the engineering conference where IBM, NVIDIA, Intel, AMD, Arm, and Fujitsu all showed how their next processors actually work. The headline disclosure is a mainframe core that runs Arm natively. Ray also covers Apple’s odd M6 Mac mini naming, London’s first autonomous Uber rides, Amazon’s purchase of the company behind DuckDB, GitHub’s HydraFusion, the best of IFA 2026, and new USDA research on farmed salmon. – Want to start a podcast? It’s 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 Get 1Password Full Summary Cochrane opens the show with a personal update. He apologizes for the late rollout and the missed Monday episode, having spent the week fighting a cold. He’s also heading to Michigan to spend time with family and visit his father’s gravesite. He hopes to record a couple of shows from his dad’s old studio while he is there, including a special episode planned for Tuesday. He also points listeners to a fresh site redesign that trades the old techie look for something cleaner and friendlier. The featured segment starts from a wrap-up post on Arm’s newsroom. However, Cochrane broadens it to cover the whole conference rather than a single article. Hot Chips has run every August since 1989, and this year’s event was the 38th, held August 23rd through the 25th at Stanford’s Memorial Auditorium. What Hot Chips Actually Is Cochrane draws a line between Hot Chips and the big consumer trade shows. CES and Computex exist for product announcements and marketing. Meanwhile, Hot Chips is an IEEE engineering conference where chip architects present block diagrams and die photos for thirty minutes at a stretch. The audience matters as much as the content. Roughly five hundred people who design chips for a living fill the room, and they would spot a fudged number immediately. No written paper is required, just the talk and the slides. In-person tickets sold out this year, as did Stanford’s dorm housing. Cochrane says he plans to cover the conference annually going forward. IBM Built a Mainframe Core That Speaks Arm The disclosure that stopped Cochrane cold came from IBM. Its next processor for IBM Z and LinuxONE runs two completely different instruction sets natively, on every one of its eleven cores. Those are z/Architecture, IBM’s own mainframe language, and AArch64, which is 64-bit Arm. Crucially, this is not emulation. Nor is it Arm cores glued onto the die beside the mainframe cores. IBM built 2,792 Arm instructions directly into the hardware, which it says is more than double the mainframe instruction count. Each core carries two separate decoders while sharing the caches, branch predictor and register files downstream. It switches between the two in nanoseconds. IBM even added dedicated hardware to flip byte order, because Arm and the mainframe store numbers in opposite directions. The payoff is Arm SystemReady compliance, meaning off-the-shelf Arm Linux runs on a mainframe unmodified. Patrick Kennedy of ServeTheHome, who was in the room, wrote: “I am sitting here still in awe of what IBM is doing here; this is not Z plus Arm cores, this is Z and Arm in one core.” Cochrane flags one precision point that is easy to get backward. IBM did not license Arm’s core designs and drop them in. Instead, it took its own mainframe core and taught it AArch64 under an architecture license, which is considerably harder engineering. The specifications are striking. The chip uses a 2nm process, with eleven cores running above 5.7GHz sustained and no turbo mode at all. Each core gets 36MB of L2 cache, backed by a 3.5GB virtual L4 pool. Furthermore, the reliability target is eight nines, which works out to roughly three tenths of one second of unplanned downtime per year. IBM gave it no name and no ship date, though the press expects “Telum III” around 2028. Arm, Fujitsu and NVIDIA Show Their Hands Arm itself had plenty to discuss, starting with an unfortunate name. Its first chip in thirty-five years is called the AGI CPU, which is a product name rather than any claim about artificial general intelligence. For three and a half decades, Arm designed processor blueprints and licensed them out, collecting royalties without competing. That era is now over. The AGI CPU is Arm’s own silicon, co-designed with lead customer Meta, running up to 136 cores on TSMC’s 3nm process at 300 watts. Arm’s CEO says the company has more than $2 billion in customer demand across the next two fiscal years. Fujitsu brought the detail Cochrane called the coolest of the conference. Its MONAKA chip packs 144 Arm-based cores, but the trick is the cache. Rather than sitting alongside the cores and eating die area, the entire last-level cache lives on a separate 5nm die with the 2nm compute die stacked directly on top. It ships in 2027 in 350-watt and 500-watt versions. NVIDIA had more stage time than anyone, with six sessions. Its new Vera CPU carries 88 cores of NVIDIA’s own Olympus design, which marks a change: the previous Grace CPU used Arm’s off-the-shelf cores. Consequently, Arm’s win here is the instruction set, not the blueprint. The memory disclosure drew the most attention, with a fully loaded system reaching 1.5TB at 1.2TB/s while the whole memory subsystem draws just 30 to 40 watts. The Caveat on NVIDIA’s Benchmark Slides Cochrane pushes back on how NVIDIA presented its numbers. On the standard SPEC integer benchmark, Vera scored 925 against AMD’s 128-core EPYC score of 898, about three percent ahead. However, the slide NVIDIA showed normalizes that same result per physical core, which makes a three percent gap look enormous. NVIDIA defends the choice, arguing that per-core throughput matters when thousands of AI agents run at once. Cochrane grants that it is a fair argument to make. Even so, his verdict is blunt: it is a different number from the headline one, and presenting it that way is not the best look. 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 the show. Apple’s Newest Chip Landed in Its Cheapest Mac Last week’s episode covered the Mac Studio half of Apple’s August announcement. Tonight Cochrane takes the other half, the new Mac mini, and finds the numbering genuinely strange. The $899 base Mac mini gets the M6, which is Apple’s first 2nm chip and the newest process the company has shipped. It brings twelve CPU cores, twelve GPU cores, and 170GB/s of memory bandwidth. Apple also introduced a third CPU core class called super cores. So Apple’s most advanced chip sits in its cheapest desktop, while the M5 Pro, M5 Max and M5 Ultra above it all carry a lower number. It gets stranger. The M6 mini has Thunderbolt 4 while the pricier M5 Pro mini has Thunderbolt 5, and the memory ceiling runs backward too. Apple explains none of it across three announcement pages. Reading between the lines, Cochrane figures the M6 is the entry point of a new generation that shipped ahead of its larger siblings. London’s Robotaxis Started Carrying Passengers Arm’s monthly roundup covers everything outside the data center, and Wayve stood out. Transport for London granted the British self-driving company private hire vehicle licenses on August 5th, the same category a minicab needs. Then on September 3rd the service launched with Uber, marking the first autonomous rides ever offered to UK passengers. A small fleet of Ford Mustang Mach-Es covers anywhere in London except the airports, with a TfL-licensed safety driver still aboard. Over 140,000 Londoners signed up. The compute runs on NVIDIA’s Arm-based automotive platform, which is why Arm claims the win. Cochrane notes the pattern: once you set a standard nobody can move off, these wins keep arriving. Two more items round it out, both about squeezing AI onto phones. Google’s Pixel 11 shipped in August with the Tensor G6, and Google claims on-device AI runs up to 3.5x faster using 3.5x less energy. Those are Google’s own unbenchmarked figures. Separately, Graphcore’s research team worked with Arm to run an 11-billion-parameter vision model on phone-class processors by squeezing each parameter to 2.7 bits, taking the model from roughly 22GB down to 3.7GB. Intel Is Pitching AI Infrastructure From a Long Way Back Intel’s newsroom post previews the AI Infra Summit, running September 15th through 17th in Santa Clara. CEO Lip-Bu Tan takes a fireside chat on Tuesday morning, with three Intel sessions across the show. Cochrane unpacks two terms first. Physical AI means AI that acts in the real world through sensors and motors rather than living on a screen, and Intel takes it seriously enough to have renamed its PC division the Client Computing and Physical AI Group in May. Disaggregated inference splits the two phases of running a model: reading your prompt is compute-hungry, while writing the answer back is memory-hungry. The context is where this gets interesting. Intel’s revenue rose 25 percent last quarter, its fastest growth since 2011. Nevertheless, in AI accelerators the company barely registers next to NVIDIA. Gaudi has effectively been abandoned, to the point that Intel stopped maintaining its open-source driver, and AMD passed Intel in data center revenue last quarter. Tellingly, when Intel demoed disaggregated inference at Computex, NVIDIA GPUs handled one phase and SambaNova chips the other while Intel supplied the coordinating CPU. Crescent Island, Intel’s actual inference chip, does not sample until later this year, and Intel declined to publish its memory bandwidth. Amazon Bought the Company Behind DuckDB On August 26th, Amazon signed a deal to acquire DuckLabs, the Amsterdam company behind DuckDB. Cochrane spends time explaining what DuckDB is, since listeners outside the data world may never have encountered it. The problem it solves is familiar. Querying a large pile of data files traditionally meant either running a database server or spinning up a data warehouse with a cluster, a bill, and a loading pipeline. Both are heavy machinery for a question you wanted answered in ten seconds. DuckDB instead ships as a library rather than a server. You add it to your program like any other package, point it at your files, and write ordinary SQL directly against them. The data never moves. The usual comparison is SQLite, which is embedded in nearly every phone and browser on earth. Where SQLite excels at looking up one record, DuckDB rebuilds that embedded idea for chewing through millions of rows. It now sees roughly 62 million monthly downloads on Python’s package index alone, up from about 25 million last October. AWS says it is buying the company, not the project. DuckDB stays free and open source under the MIT license, held by a Dutch nonprofit foundation, and the founders join AWS while continuing to run technical direction from Amsterdam. Andy Warfield, a VP and distinguished engineer at AWS, described DuckDB as “the glibc of structured data: a lean, unglamorous, ubiquitous dependency that a great deal of software links against and almost nobody has to think about.” Cochrane sits with what that ownership means. He reaches for an analogy: imagine Daniel Stenberg selling curl. He doubts it would ever happen, but the concept alone is startling given how much infrastructure depends on it. His read is that AWS is betting DuckDB becomes as foundational as curl and SQLite already are. GitHub Has One AI Model Grade Another One’s Homework GitHub shipped Project HydraFusion into Copilot as a research preview. Instead of routing your request to a single model, it picks one of three approaches per request. Sometimes one model simply answers. Alternatively, a cheaper model drafts, and a quality gate decides whether to escalate. The interesting one is Critique. One model writes the code, a separate model from a different family reviews it read-only, and the original gets one pass to revise. Despite the name, nothing is fused here. There is no voting and no merging, just one model at a time with a gate deciding whether to spend more. GitHub explained the reasoning in an earlier post: “a model reviewing its own work is still bounded by its own training biases: the same training data and techniques, the same blind spots.” Research supports it. A team at NeurIPS in 2024 showed that models recognize their own writing and score it higher than human graders do. GitHub’s earlier number had a Claude Sonnet and GPT critic pairing closing about three-quarters of the gap between Sonnet and the larger Opus model. This mirrors a workflow Cochrane uses constantly and has described on a previous episode. He runs a cross-check review with a second model from a different company, and it routinely surfaces issues the first model missed. He explains that different training data, different engineers, and different reinforcement approaches build different internal biases about what counts as correct. Looking ahead, he expects more of these “Frankenstein patterns” where models from different training families work together. The Best of IFA 2026, and What You Can Actually Buy IFA opened to the public in Berlin for its 102nd year, with about 1,900 brands. Cochrane splits The Verge’s roundup in two, since much of what generates headlines at these shows never ships. Starting with real products, iRobot’s flagship Roomba Max 875 Combo runs $1,199 and ships in about two weeks. Its SealForce feature drops a hidden skirt from the chassis when it detects carpet, sealing against the fibers so suction concentrates instead of leaking out the sides. That reaches 35,000 pascals, iRobot’s strongest yet. A step-down model at $899 carries the same trick, though Cochrane balks at both prices. Anker’s Soundcore Sleep 4 Pro earbuds arrive in November at $349.99. The charging case carries its own round touchscreen, so you pick soundscapes, set alarms, and read sleep stats without your phone. Optical sensors read heart rate and variability from the ear canal, which beats the wrist for accuracy, and the case masks a snoring partner. Cochrane remains unconvinced about sleeping with earbuds in. Philips also has smart rope lights, the Hue Liane 360, on sale now. They glow evenly around the tube rather than showing individual LEDs. They also run $400 for three meters, which works out to about $130 per meter of rope light. As for concepts nobody can buy, iRobot showed a robot vacuum that carries a smaller robot vacuum on its back in a garage and lowers it to deploy. Lenovo brought a 14-inch laptop whose screen rolls out to 17 inches at the press of a button, which reviewers call the first rollable that feels close to shippable. Tecno showed a phone with essentially no border around the screen, and Acer had a Windows gaming handheld that swivels its screen up over a keyboard. Two themes ran through the show. Humanoid robots were the loudest thing on the floor, and IFA’s own CEO framed the event as being about robots that work rather than robots that demo. Meanwhile, AI stopped being its own product category and became an ingredient, showing up in refrigerators, treadmills, dishwashers, and motorized TV mounts. Farmed Salmon Isn’t the Omega-3 Machine It Used to Be USDA scientists measured farmed salmon and found considerably less of the good fat than the government’s own database claims. EPA and DHA are two fatty acids you get almost entirely from fish. Your body can build them from the plant version, but only in tiny amounts, so the NIH’s position is that eating them is the only practical way to raise your levels. Those fatty acids are structural pieces of every cell, with DHA concentrating in the brain and retina. That is why the federal dietary guidelines, issued jointly by USDA and Health and Human Services, recommend at least eight ounces of fish a week and steer you toward salmon. That amount is calibrated to deliver about 250 milligrams a day. Published in Frontiers in Nutrition last month, the study found EPA and DHA in farmed Atlantic salmon came in 54.7 percent lower than USDA’s own reference values, last updated in 2018. A three-ounce serving fell from roughly 1,670 milligrams to about 756. Consequently, two servings a week now fall about 14 percent short of the target. Plant-derived fats meanwhile rose two to three times over. The likely cause is feed. Salmon are carnivores, and farms once fed them oily little fish. There was never going to be enough of those as the industry scaled, so crops filled the gap: soy, canola, sunflower and linseed. Importantly, the study does not claim to have proven this and calls the feed shift a plausible explanation. Independent corroboration lends it credibility. Researchers at Stirling measured a similar halving in Scottish salmon between 2006 and 2015, and Norway’s marine institute saw it across thousands of samples. There is a land dimension too. Roughly half the world’s soy grows in South America, where rainforest gets cleared for feed. Matthew Hayek, who studies the environmental cost of protein at NYU, told Inside Climate News that “soy is a major, important protein and oil ingredient in fish farming.” Adding up two decades of soy across all fish farming, he puts the extra forest clearing at around the area of Nicaragua or Bangladesh. That figure covers all fish farming rather than salmon alone, and Hayek notes it is hard to attribute soy use to any single species. Cochrane closes with two caveats. First, the study measured only eight fish, bought around Maryland, DC and Virginia over six weeks in 2023, and nearly all sourced from Chile. That is not a national survey, and the authors say plainly the sample was not large enough to change government advice. Rather, it flags that a federal database value needs rechecking, which is what the paper set out to do. Second, on whether you should care, farmed salmon still beats beef, chicken and eggs by a mile, since those carry essentially zero EPA and DHA. What it loses is its crown among fatty fish, dropping to mid-pack behind herring, sardines and mackerel and roughly level with trout. The broader health case is also softer than the 2000s suggested. A review of 86 trials covering 162,000 people found supplements barely moved heart attacks or deaths, so eating fish and swallowing fish oil are not the same claim. No producer has responded to the findings, and USDA, whose own scientists ran the study, declined an interview and did not answer emailed questions. Cochrane wraps up with housekeeping and a note that he will be back on Labor Day. The post The Mainframe Learned to Speak Arm #1875 appeared first on Geek News Central.

How many cores does your machine have, 10, 18? Your async Python code uses just one of them. That isn't a bug in asyncio. That's the design, and optimizing event loops to be faster by 20% doesn't change it. So Giovanni Barillari started over. Joe is the creator of Granian, the Rust-based server that powers Talk Python. His new project is TonIO, an async runtime written from scratch for free-threaded Python. Real threads, a handful of primitives instead of asyncio's pile of them, and it flat out refuses to start if the GIL is on. Episode sponsors Sentry Error Monitoring, Code talkpython26 Python in Production Talk Python Courses Links from the show Guest Giovanni Barillari: github.com Granian: github.com Hyper: github.com Free threaded Python: docs.python.org Sort of: labs.quansight.org did a whole course: training.talkpython.fm uvloop: github.com rloop: github.com TonIO: github.com your EuroPython 2026 talk: www.youtube.com Michael's Cutting Python Web App Memory Over 31% Article: mkennedy.codes Watch this episode on YouTube: youtube.com Episode #561 deep-dive: talkpython.fm/561 Episode transcripts: talkpython.fm Theme Song: Developer Rap

How do you plan for the performance of your Python applications? What does a performance budget entail, and where should you spend your resources? This week on the show, we speak with Den Odell about his new book "Fast by Default: Practical Performance Engineering."

(Presented by TLPBLACK: A cybersecurity intelligence platform focused on sharing curated, high-sensitivity threat insights and research with trusted security professionals.) Three Buddy Problem - Episode 112: The 'OpenAI hacks Hugging Face' fallout has turned into a story about AI civilizations rising from the ashes, politicians calling for super-intelligence bans, and the emergence of well-funding non-profits doing AI safety work. Who are these people and what's their security expertise? Plus, GPT-6 Astra lands in a trusted-access program nobody can get into, Costin ranks the local models he runs next to his desk, and CrowdStrike sinkholes a botnet that's been alive since 2003. Cast: Juan Andres Guerrero-Saade, Ryan Naraine and Costin Raiu. Timestamps: 0:00 Introductory banter 1:02 Conference season: LabsCon, Offensive AI Con, Countermeasure 5:40 The Hugging Face story hits the front page 7:04 Dwarkesh, Greenblatt, and the AI-pilled framing 11:51 Swap "agents" for "Python" and the panic goes away 16:34 Does anyone actually know what happened? 21:18 Bernie Sanders wants to ban superintelligence 34:03 Defending against swarms: the 2026 SOC 39:15 Logs, Splunk, and the business model in the way 44:11 What EDR vendors are actually building with AI 56:16 The security poverty line and the endgame 1:07:53 GPT-6 Astra, Fable 5.1, and local model rankings 1:27:25 Google's Fairwind, CodeMender, and agents running Linux 1:45:31 Apple's bet on local inference 1:53:21 The Sality takedown and endgame advice

Eric Chou welcomes guest co-host William Collins as well as Network Automation Forum founders Scott Robohn and Chris Grundemann to discuss how their community emerged from a simple question: Why haven’t we seen full adoption of network automation? They discuss the growth of AutoCon and how its practitioner-focused, vendor-neutral approach has fostered a unique environment... Read more »

Eric Chou welcomes guest co-host William Collins as well as Network Automation Forum founders Scott Robohn and Chris Grundemann to discuss how their community emerged from a simple question: Why haven’t we seen full adoption of network automation? They discuss the growth of AutoCon and how its practitioner-focused, vendor-neutral approach has fostered a unique environment... Read more »

This show has been flagged as Clean by the host. -------------------- 01 Introduction This is the second episode in an 8 part series. 02 In the previous episode we discussed the predecessors of PLCs, in particular relay logic. In this episode we will discuss how relay logic came to be expressed in software rather than in actual hardware components. 03 The topics to be covered include * Early computers in industry. * The first PLC. * PC versus PLC - what's in a name. * Who the major brands are. * What does a PLC actually look like. * Machine architecture. * PLC programs. * The scan. * PLC programming languages. * Relative popularity of PLC programming languages versus more conventional computer programming languages. -------------------- 04 Early Computers in Industry Computers came to be used in industry fairly early on. Minicomputers were used in some large industries to control things like electric power plants. For example, the DEC PDP8 was used to control the refuelling machines in CANDU nuclear power plants. These would load and unload fuel from the reactor, which happens continually on a daily basis while the reactor is running. 05 The first microprocessor based computer sold on the commercial market was the MICRAL from France, based on the Intel 8008. It was sold as a cost effective replacement for minicomputers used in industry. It preceded what are generally considered to be the first Personal Computers. So you can see that industry were not reluctant to adopt new technology. 06 However, mini computers were large complex systems that did not fit well into a factory floor. They had particular niches in very large complex integrated systems, but were not suited to controlling many individual machines in a factory that produced things like automobiles or appliances. 07 What was needed was something that would fit into a standard electrical enclosure, withstand the temperatures found in a factory, could stand up to vibration and noise, was tolerant of voltage fluctuations, interfaced directly to sensors and actuators, and could be readily programmed by engineers, technicians, and tradesmen who were familiar with the processes to be controlled, but had little or no experience with computers. 08 This required a complete integrated package covering hardware, software, and product distribution through industrial supply retailers. Something that could meet these criteria is what was needed to become the PLC. -------------------- 09 History Many histories point to a US developer that became Modicon. However, their first hardware was more a proof of concept than a viable commercial product. In fact multiple companies in the US, Europe, and Japan were working on the problem and released comparable products all at around the same time in the early 1970s. 10 The idea preceded the implementation by a number of years. For example, General Motors had asked industrial control equipment suppliers in the mid 1960s for some sort of programmable control device to replace relay logic. This showed potential suppliers that there was a market for this sort of thing, and gave them a clearer idea of what customers were looking for. 11 What made this possible was the development of the first 8 bit microprocessor in that time period, combined with readily available bit slice processors from the minicomputer industry and other components. 12 People knew what to do, the problem was waiting for the development of suitable component hardware. Minicomputers were already being used in industry, but were normally located in control rooms. PLCs were an effort to take that technology out of control rooms and put it on the shop floor. 13 PC Versus PLC - What's in a Name In the early days, these systems were known as either PLC, which means "Programmable Logic Controller", or PC, which means "Programmable Controller". Different vendors favoured different terminology, but both were common. I have a book which was at one time one of the standard reference handbooks for this industry which was published in 1989, and refers to them as Programmable Controllers, or PCs. 14 The term PC was also used to mean "Personal Computer", but that wasn't really a problem in the early days. However, a certain large company decided to call their entry into the personal computer market the "IBM PC", and suddenly "PC" became a generic term for desktop computers. 15 After a long struggle to keep referring to their products as "PCs", even the biggest vendors caved in and gave up the fight and switched to using the "PLC" term. I will therefore use the term "PLC" in this podcast series even when talking about products which were originally called "PCs" at the time they were introduced. 16 Who the Major Brands Are The companies that came to dominate the industry fairly early on were generally companies that already made industrial electrical hardware. These were Siemens Allen Bradley, later known as Rockwell Schneider Mitsubishi Omron 17 These are still the dominant companies in the business, although there are many small brands, particularly at the cheaper end of the market. All of them were suppliers of a wide range of industrial control hardware, such that you could build all or nearly all of the parts of your control system using only their products. 18 Each of these sells an integrated hardware and software package, including development software, that is completely proprietary. If you thought that the mainframe business had a lot of vendor lock-in, you haven't seen the industrial market. -------------------- 19 What Does One of These Things Actually Look Like? At this point you are probably still very confused as to what a PLC actually is. I will attempt to describe one in basic terms by giving an example of one. The smallest and simplest ones are what are termed "shoebox" PLCs. These are basically a rectangular box with a plastic case. 20 Here are the dimensions for a typical low end model, a Mitsubishi FX3U-32M. According to specs in the manual, it is 150mm wide, 90mm high, and 86mm deep. 21 It can be mounted to the panel in an electrical enclosure by snapping it onto what is termed a DIN rail. DIN rails are standard mounting systems which use a strip of metal which is U shaped with a lip at the top of each end of the U. You screw the DIN rail to the electrical panel in the enclosure, and most industrial components such as PLCs, terminal blocks, circuit breakers, and all sorts of other things will simply snap onto it and be ready to wire together. 22 Along the top and bottom of the PLC are terminals to which you can attach wires from the things you want to sense or control in the rest of the machine, such as push buttons, pilot lights, proximity sensors, and valves. 23 Inputs are along the top, and outputs are along the bottom. The FX3U-32M has 16 inputs and 16 outputs. These I/O can be 24 volts DC, or 100 to 120 volts AC, depending on the model. Alternatively it may have small relays for outputs. While AC I/O once predominated, 24VDC became the standard in most industries several decades ago as it interfaced with electronic devices more readily and also allowed for cheaper and more compact wiring. 24 More inputs and outputs can be added by connecting additional I/O modules to the main PLC next to it on the same DIN rail, up to a total of 256 I/O The connection is typically via short ribbon cables which plug into the adjacent module. These I/O modules look like the main PLC, but just add I/O. 25 Inside the PLC are a CPU, memory, and firmware. This model has 64K "steps" of memory, which can be thought of as how many instructions you can have. The program runs in RAM, but you can add a small flash memory device to save the program to. There is a run/stop switch that allows you to start or stop the user program. There is a port that you can use to connect the PLC to a laptop computer with a cable so you can download your program to it. 26 The biggest part of the market for PLCs is for these "shoebox" style. However, there are bigger ones as well which have more I/O, more memory, faster CPUs, etc. These are meant for tasks such as coordinating large assembly lines and things like that. These are what are termed rack systems, where the rack is an empty box which is open at the front and has a backplane bus running along the back. You plug the main CPU into the bus, normally in the leftmost position, and then plug I/O modules into the rack. The I/O modules are tall, thin, fully enclosed boxes with the I/O terminals along the front. 27 You should have a rough idea of their appearance at this stage, so I'll leave the physical description aside for now and go on to the concepts behind how they actually work and what makes them different from something like say a Raspberry Pi. -------------------- 28 Machine Architecture It is not really the hardware which defines the PLC. Rather, it is the software system architecture. Every PLC that I am aware of has a common set of features. 29 Data Table A data table is simply a block of memory which may or may not be subdivided into different parts. All logic, and all, or nearly all, I/O works by writing to and reading from addresses in the data table. 30 I/O Every input and output point maps to a fixed memory address in the PLC. To read the state of an input, you read its memory address. To change the state of an output, you write to its memory address. 31 These digital I/O appear as single bit boolean values. The PLC programming language has instructions which allow single bits to be read or written to directly without have to mask off other bits as you would have to do if you were using a conventional programming language. I have so far only mentioned digital I/O, which are single bit on/off values. 32 Other types of I/O Many PLCs also have other types of I/O such as Analogue I/O, which represent variable voltages rather than just on/off values. For example you may wish to read a temperature from a thermocouple where the voltage level varies with the temperature. Often these are mapped into the data table as byte or word values. 33 Internal Bit or Boolean Memory There will be a range of single bit or boolean memory which is used for internal logic state. Your program may need to come up with a series of intermediate logical values, and you store them in these internal bits or flags. 34 Timers and Counters Control of machinery often involves timing and counting. Timer and counters are mapped into the data table. Timer and counter presets and values can be read as words, and there will be bit addresses which indicate the status of the timer or counter, such as whether it has reached its preset and is "done". 35 Integer and Floating Point Values There will be word addresses where integer and floating point words can be stored. If you need to do some mathematical calculations and store the results, you would save them in an integer or floating point word. 36 Typed Memory Unlike when programming something like a PC where you simply have a range of undifferentiated bytes and it is up to your software to impose meaning on it, PLC data table memory has defined types and meanings and the system firmware enforces the correct memory access methods. 37 Size of Data Table Data tables vary greatly in size. A bigger data table allows for bigger and more complex program. Generally, new PLCs will have bigger data tables than older ones, and more expensive PLCs will have bigger data tables than cheaper ones of the same generation. 38 Example Having picked a current PLC as an example for physical dimensions, I will pick an old and obsolete one for an example of a data table. The Siemens S5-100 series was a small PLC that came in several sizes. The basic S5-100 had the following data table size. 39 Digital I/O had a maximum of 128 inputs and outputs taken together. Analogue inputs and outputs had a maximum of 8 taken together. Flag, or bit, memory was 1024 Timers - 16 Counters - 16 40 The S5-103 offered more of everything in the same basic package, but of course at a higher cost. Digital I/O had a maximum of 256 inputs and outputs taken together. Analogue inputs and outputs had a maximum of 32 taken together. Flag, or bit, memory was 2048 Timers - 128 Counters - 128 -------------------- 41 The PLC Program A PLC will come with what amounts to an operating system and virtual machine built into it. You as the programmer use the vendor's proprietary development software, what is generally called the "PLC programming software" to write an application for it. 42 You then connect your PC to the PLC using a cable of some sort, and download your program into it. The program gets stored in the PLC's RAM. There may be flash memory which serves to back up the RAM when the power is off so you don't lose the program. Before flash was available, a battery was typically required to hold the static RAM memory. 43 There is just one program in the PLC. There is no file system, just the program memory and the data table. When the PLC starts up, it runs the program that you wrote. It continues running the program until you turn off the power or you set the run/stop switch to the stop position. 44 I will come back to programming later, but we needed to cover these basic points before I could describe some further concepts we need to cover. -------------------- 45 The Scan A key thing to understand about PLCs is the "scan" concept. This works as follows. There is a repeated cycle called the scan. 46 First, the PLC reads all the physical inputs and updates the corresponding input addresses in the data table. Next it runs the user program once. Finally it reads the output addresses in the data table and writes them to the corresponding physical outputs. Then it repeats the scan from step 1. 47 A scan can take anywhere from a few milliseconds to hundreds of milliseconds, depending on the model of PLC. Older PLCs and cheaper PLCs tended to be slower. However, as integrated circuit technology advanced, programs got faster. 48 The PLC has a watchdog timer. This is a timer which runs in the background and resets at the start of a scan. 49 If a scan takes too long, the watchdog will trip and stop the program and set the outputs to some defined state, either turning them all off or holding them at the last value. This is known as "faulting" the processor. 50 A watchdog fault may be caused by a program that is too long, or uses a lot of very slow instructions, or in some cases it may be due to a bug in your program. However, this last cause is not as common or as easy to cause as you may think when it comes to faults. -------------------- 51 PLC Programming Languages The key to all of this is the unique way in which the programming languages used in PLCs work. I will first briefly describe the two main programming languages and then explain how they work in a PLC as part of a complete system. 52 Ladder Logic There are several programming languages, but the main one is called ladder logic. If you listened to the previous episode, the word "ladder" will ring a bell. When machines were controlled by electromechanical relays, the electrical drawings which specified and documented the types of relays and the connections between them were called "ladder diagrams". 53 PLCs simply take these ladder diagrams and reproduce them on your computer screen in the programming software using the same standard schematic symbols. 54 You do not need to draw the symbols like it was CAD. Instead you simply use your keyboard or mouse to say that you want this symbol to be entered in the current cursor location or you want this wire to be connected from here to there. When your rung is complete the software will check at some point if it is syntactically correct. You then go on to entering the next rungs one after another until you have your complete program. 55 Instruction List An alternative representation is something called "instruction list", although various vendors may use other terms. This is a text representation that looks like a sort of assembly language. However, it's actually really ladder, just shown in a different way as text and some types of programming software will allow you to toggle between ladder and instruction list mode, translating between them automatically. 56 This should give you a clue as to how a PLC can execute a schematic diagram. Behind the scenes the programming software will translate the diagram into instruction list, and then the PLC will either execute those instructions in an interpreter, or compile them to machine code and execute those. 57 Other Programming Languages There are other programming languages, but they are rarely seen. I won't go into detail in terms of describing them, just briefly mention them. 58 One is called Sequential Function Chart. This is a type of flow chart which is designed to show complex sequences, including ones with alternate or parallel paths. The original name for this is Grafcet, spelled G R A F C E T. This stands for "Graphe Fonctionnel de Commande Étape Transition". As well as being a programming language, it is also a very good design and analysis tool even if the PLC being used doesn't support it. 59 Another is Function Block Diagram. This is another graphical language which resembles flow diagrams used in process industries. This is supposedly used mainly in industries such as chemical manufacturing. However it appears to be very niche and most people who use PLCs will never have seen it, and many will not have even heard of it. 60 Yet another language is Structured Text. This is very obviously derived from Modula-2 and strongly resembles it. If you are not familiar with Modula-2, it was created by Nikolas Wirth and intended as the successor to Pascal, which it was derived from. Structured Text seems to be much loved by a small number of academics, but seems to have little actual use in the field. You lose pretty much all of the monitoring and debugging facilities built into a PLC if you use it, so it's kind of pointless except perhaps for a few niche applications. There are a few other languages as well, mainly proprietary ones, often using flow charts or the like. 61 PLC Languages are Non-Blocking The important thing about a PLC program is that it executes from top to bottom, left to right, without stopping or blocking. There is no waiting on input or output or for a system call to return. Each rung immediately yields a result which is written to a memory address or which enables a timer or counter. Every rung is executed every scan. 62 You can reproduce this in a traditional computer programming language. In ladder logic it is inherent to the syntax and it is either very difficult or impossible to do it any other way. I'll give an example in Python of what I mean. 63 a = (b or c) and not d 64 If b or c are true and d is false, then a is set to true. If b and c are false or d is true, then a is false. 65 Now imagine that we are not executing this statement once, but rather are executing it repeatedly. This statement will never block execution. It will always immediately yield a result. Now let's modify that a bit. 66 a = (b or a) and not d 67 Note that we have replaced c with a. Now the value of a depends the previous value of a as well as b and d. However, once a becomes true, the value of b no longer matters because b is in an or condition with a. Now if we execute it over and over again, only d becoming true will make a go false. 68 This is a standard push button circuit, also known as a "seal in circuit", because it "seals" around the start condition. 69 b is the start push button with a normally open contact. d is the stop push button with a normally closed contact. a is a relay with one of its contacts being used to hold it on. 70 If you wired this up with actual push buttons and relays or if you programmed it into a PLC with ladder logic it would work the same way. A PLC program will consists of rung after rung of logic like this and executes it repeatedly scan after scan, only pausing to update its I/O. 71 Async Programming Some of you may be thinking that this cyclical scan sounds like the async programming that is all the rage with web server applications these days. Essentially it is exactly the same principle. 72 With async programming, your program must never use blocking instructions and execution proceeds on a repeated cycle. PLCs solve this by not having any instructions which block, and many PLCs do not allow backwards jumps. Tradesmen in coveralls in factories were doing async programming decades before the cool kids heard about it. 73 Most PLC Programming Languages are Visual or Graphical Languages If ladder, Sequential Function Chart, or Grafcet and Function Block Diagram sound like visual programming languages, they are. In fact ladder is probably one of the earliest visual languages in commercial use. Wikipedia defines a visual programming language as: 74 In computing, a visual programming language (visual programming system, VPL, or, VPS), also known as diagrammatic programming, graphical programming or block coding, is a programming language that lets users create programs by manipulating program elements graphically rather than by specifying them textually. A VPL allows programming with visual expressions, spatial arrangements of text and graphic symbols, used either as elements of syntax or secondary notation. For example, many VPLs are based on the idea of "boxes and arrows", where boxes or other screen objects are treated as entities, connected by arrows, lines or arcs which represent relations. VPLs are generally the basis of low-code development platforms. Scratch is an example of a VPL 75 End of quote. As you can see, what is cool today was on the factory floor more than 40 years ago. -------------------- 76 Popularity of PLC Programming Languages PLCs are about as proprietary as you can get. Pretty much every vendor has his own proprietary take on each language. How you can program any particular PLC is determined by that vendor's programming software. If the vendor doesn't support it, you can't use it. An individual vendor may even have multiple incompatible product lines which have to be programmed in different ways with different software, although that is not as common these days as it once was. 77 Nearly all PLCs support programming in Ladder. Some support programming in instruction list as well as ladder. Anything else is much less common. 78 Ladder logic happens to be on the Tiobe Index by the way. For those who have not heard of it, the Tiobe Index, that is T I O B E, is a web site that lists the popularity of numerous programming languages based on various criteria. 79 Number one on their list happens to be Python, currently at 19.98%. Number two is C, at 11.55%. 80 At the time of writing this script, Ladder Logic was at number 46 with a 0.28% rating, which put it just below Erlang and just above Haskell. I'm not sure whether that means that Ladder Logic is not as obscure as you thought it was, or whether it means that Erlang and Haskell are in fact more obscure than you thought they were. -------------------- 81 Episode Summary In this episode we covered we covered The early history of computers in industrial control The early history of PLCs, including how they got their name Who the major brands are What they look like physically 82 A basic description of the abstract machine architecture A very brief look at what a PLC program is like The scan concept The main PLC programming languages The minor PLC programming languages The relative popularity of each of the programming languages 83 In the next episode we will take a look at one of the early PLCs from the era when they began seeing widespread use. This PLC was hugely successful and was for many companies the first PLC they used. This is the Allen Bradley PLC2. 84 This has been the second episode in an 8 part series. -------------------- Provide feedback on this episode.

Topics covered in this episode: OpenAI's Python SDK has migrated to HTTPX2 TMOG - Native Task Manager for macOS, Windows, and Linux wrapture - one wrapper for mocking, tracing, and observability linkedin2md: turn your LinkedIn export into 40+ Markdown files Extras Joke Watch on YouTube About the show Sponsored by us! Support our work through: Our courses at Talk Python Consulting from Six Feet Up Connect with the hosts Michael: Mastodon / BlueSky / X / LinkedIn Calvin: Mastodon / BlueSky / X / LinkedIn Show: Mastodon / BlueSky / X Join us on YouTube at pythonbytes.fm/live to be part of the audience. Usually Tuesday at 7am PT. Older video versions available there too. Finally, if you want an artisanal 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. Calvin #1: OpenAI's Python SDK has migrated to HTTPX2 The OpenAI Python SDK has migrated to HTTPX2, the Pydantic-stewarded fork of httpx. Pydantic picked it up citing "limited activity recently" in the original project, promising "a reliably maintained path forward." If you just use the default client, nothing to do. No code changes. The catch is TLS. Quoting the guide: HTTPX "previously verified certificates against the CA bundle provided by certifi. HTTPX2 instead uses the operating-system trust store, and the SDK no longer installs certifi." That "can break certificate verification in minimal container images without system CA certificates, environments using corporate TLS-inspecting proxies, and deployments that relied on a custom or modified certifi bundle." The fix is SSL_CERT_FILE or SSL_CERT_DIR, or pass your own ssl.SSLContext via verify. Deeper integrations need real edits: custom clients, auth handlers, hooks, and request mocking all take HTTPX2 objects now, and plain httpx is no longer pulled in transitively. So import httpx in your own code means declaring it yourself or moving over. Temporary escape hatch: a legacy HTTPX client Michael #2: TMOG - Native Task Manager for macOS, Windows, and Linux A native, deeply instrumented system monitor for macOS, Windows, and Linux, now in public beta - from Plummers' Software, i.e. Dave Plummer, who wrote the original Windows Task Manager and donated it to Microsoft in 1995. Wikipedia Three real native apps: Swift/AppKit on macOS, Win32 on Windows, C++/Qt 6 on Linux, with a shared C++ core keeping metric semantics aligned - no browser shell anywhere. One dense summary: CPU, clocks, thermals, GPU, memory, storage, network, energy, and the processes responsible for the load, all click-through. Per-core honesty: logical processor and NUMA views, P and E cores color-coded, optional kernel time, 60 FPS live meters. Memory with context: pressure, wired, compressed, cached, committed, available, and swap, plus configurable scrolling history. Processes that act like processes: tree view, filtering, sorting, follow mode, and native verbs including service and launchd control. Phosphor themes: light, dark, green, amber, blue, or mono, with color and saturation you tune yourself. Calvin #3: wrapture - one wrapper for mocking, tracing, and observability Graham Dumpleton, author of wrapt and the original New Relic Python agent, has released wrapture. The name is wrapt plus capture. The core idea: wrap real code instead of replacing it, so the real code still runs while you watch every call. Name a method with wrapture.binding(Class, "method"), open a timeline(), and you get a tape of what actually happened. Real return values, real nesting, arguments normalised against real signatures. tape.tree() prints the call graph as it ran. One mechanism, three jobs: monkey patching with a real lifecycle (apply, remove, suspend, plus returns, raises, transforms_args), unit testing that asserts on real call flow instead of a flat MagicMock call list, and ad-hoc tracing of a running app. The testing pitch is error paths. Inject TimeoutError at the payment gateway, then assert the ledger was never written. Stubs and mocks are strict and spec-required, and there is deliberately no bare Mock(). Tracing needs no code at all. A wrapture.toml naming targets and a sink, run with python -m wrapture main.py, and you get a live call tree with timings. It captures ordinary logging calls as nested events, and with the otel extra it exports spans, metrics and correlated logs with W3C trace ids that join across services. Every line of code and docs was AI-written under their direction, and they say so up front. Two weeks from first commit, eleventh alpha, over 1000 tests, 150+ pages of docs. Alpha on PyPI, needs Python 3.12+ and wrapt 2.4.0+. Michael #4: linkedin2md: turn your LinkedIn export into 40+ Markdown files Via Juan Manuel Daza - a Python CLI that unpacks LinkedIn's data-export ZIP into clean, per-category Markdown you can drop straight into an LLM. One command: linkedin2md Complete_LinkedInDataExport.zip, plus o for output dir, -lang en|es, and -pdf. 40+ output files: profile, experience, education, skills, connections, posts, comments, reactions, recommendations, endorsements, job applications, even ad targeting and LinkedIn's inferences about you. Built for LLM analysis: the README pitches NotebookLM, Claude Projects, Obsidian, and Ollama, with example prompts like "what patterns do you see in my career transitions?" PDF resume mode: -pdf renders an A4 CV via weasyprint, and degrades gracefully to Markdown-only if it isn't installed. Dependency note: "pure Python / zero-dep" holds for the Markdown path only - the PDF path needs weasyprint and markdown installed. Install: pipx install linkedin2md recommended, pip in a venv otherwise - 86% Python, 10 releases, v0.3.1 in May. Agentic dev angle: repo ships opencode config and an N3RV subagent pipeline, including a "judgment day" dual-model adversarial PR review. Extras Calvin: EVE Online Migrates to Python 3 Michael: Dinkus by Will McGugan Joke: Tao of Programming: Book 5 Maintenance

Jack Black and Paul Rudd should be able to make a giant-snake comedy work, but Anaconda left Clay with weak effects, a thinner story and almost nothing worth remembering. Also: Clay lands a speaking role in a movie, Kenny discovers Derrick Rose, and a king brown snake ends the Couchboys' pool era. Final rating: 4.5 Killer Pythons out of 10.

Si llevas años usando Bash, Zsh o Fish y piensas que los pipes de Unix son lo más parecido a la perfección, este episodio te va a hacer tambalear los cimientos. Porque existe un shell que no pasa texto entre comandos: pasa estructuras de datos. Tablas, listas, registros, fechas, tamaños de archivo con tipo real. Y encima habla con Ollama sin que tengas que escribir ni una línea de Python.Ese shell es Nushell. Está escrito en Rust, tiene más de 40.000 estrellas en GitHub, y su filosofía es sencilla: los pipes deberían transportar datos con tipo, no texto que luego parseas con awk, sed o jq.En este episodio te cuento mi experiencia pasando de Fish a Nushell con ejemplos reales. Cuando escribes ls no obtienes texto: obtienes una tabla con columnas tipadas. Puedes hacer ls | where size > 1mb | sort-by size sin recurrir a awk ni números mágicos. El shell entiende qué es un filesize, qué es una fecha, qué es un número.Y luego está open, que entiende el formato por la extensión: JSON, YAML, TOML, CSV, SQLite... todo se convierte en datos estructurados. Abres un SQLite y ejecutas consultas con query db. Y todo combinable: http get a una API, filtrar con where y guardar con save — en un solo pipeline, sin archivos temporales.La guinda es la integración con IA. Como Nushell entiende JSON y Ollama habla JSON, se entienden a la perfección. Te enseño un pipeline que lista procesos, filtra los que consumen más de 100MB de RAM, se los manda a un modelo local, y mata el que más memoria usa. Todo en una línea. También te hablo de ai.nu, un módulo que envuelve Ollama, OpenAI y DeepSeek, con function calling desde el shell.También hago una comparativa: Bash, Zsh, Fish y Nushell cara a cara. Bash funciona en cualquier sitio pero el manejo de datos es arcaico. Zsh es Bash con esteroides pero los pipes siguen siendo texto. Fish es moderno pero no entiende de tipos. Nu es el único con estructuras de datos de verdad. PowerShell fue el primero en pasar objetos, pero Nu es lo que PowerShell debería haber sido.Capítulos del episodio:0:00 — Introducción: de Bash a Fish, la evolución de las shells2:30 — El problema del texto plano: por qué Nushell es diferente5:00 — La trifecta: ls, where y select, SQL en tu terminal7:30 — Tipos reales: la shell entiende fechas, tamaños y números10:00 — Open: abrir JSON, CSV, YAML y SQLite sin herramientas externas13:00 — Procesamiento avanzado: $in, save, append y par-each15:30 — HTTP GET: APIs de GitHub y meteorología desde la shell18:00 — Comparativa de shells: Bash vs ZSH vs Fish vs Nushell21:00 — Nushell e IA: integración nativa con Ollama sin Python24:00 — Instalación, casos de uso y conclusiones finalesMás información y enlaces en las notas del episodio

In 2020, a gastroenterologist in Glasgow did the math on his new research study and came up with 30,000 samples, arriving over two years from three cities and a dozen hospitals. He asked around about how researchers keep track of that. The answer was Microsoft Excel. Shaun Chuah had written some HTML by hand in Notepad back in high school and that was about the whole of his programming experience, so he opened the Django tutorial and started reading. Six years later that app is Foundry120, holding 10 terabytes of clinical and genomics data with an agentic AI running on top of it. Episode sponsors Sentry Error Monitoring, Code talkpython26 Talk Python Courses Talk Python Courses Links from the show Guest Shaun Chuah: github.com Up and Running with Rust Course: talkpython.fm Foundry120: www.foundry120.com Designing Data Intensive Applications: www.oreilly.com Microsoft Foundry: ai.azure.com ChatIBD: www.chatibd.com Blog: shaunchuah.github.io @drshaunchuah: x.com github.com/shaunchuah: github.com Watch this episode on YouTube: youtube.com Episode #560 deep-dive: talkpython.fm/560 Episode transcripts: talkpython.fm Theme Song: Developer Rap

Topics covered in this episode: Web UIs for your reverse proxy Wagtail 8.0 is hot off the presses RISC-V is now officially supported by CPython Django's annual releases make every version an LTS Extras Joke Watch on YouTube About the show Sponsored by Logfire from Pydantic: pythonbytes.fm/logfire Connect with the hosts Michael: Mastodon / BlueSky / X / LinkedIn Calvin: Mastodon / BlueSky / X / LinkedIn Show: Mastodon / BlueSky / X Join us on YouTube at pythonbytes.fm/live to be part of the audience. Usually Tuesday at 7am 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: Web UIs for your reverse proxy Traefik, nginx, and Caddy all sit in front of a lot of self-hosted infrastructure, and all three are configured by hand-editing files. Three active projects put a control plane on top: Traefik Manager (Python + Flask), Nginx UI (Go + Vue), and caddy/ui (React + Node). All three are additive rather than replacements - none of them take ownership of your config away from you - which is the part that matters when the thing has write access to production routing. Traefik Manager is the Python one: Flask 3.1 and Gunicorn for the control plane, a lightweight Go agent for remote instances, currently v1.10.0 with an Android companion app. Nginx UI is a single Go binary at 11.3k stars, with a block-style config editor, an Ace editor doing LLM completion on nginx syntax, and an MCP server so agents can drive it. caddy/ui runs as two containers next to your existing Caddy, reads and writes your Caddyfile directly, and uses Caddy's /adapt API to validate before reload - no Docker socket required. Each one edits the config the underlying server already reads, so your files stay the source of truth and you can drop the UI without unwinding anything. Undo is a first-class feature across all three - timestamped backups with optional Git history, config version compare and restore, Caddyfile snapshots with one-click rollback. Observability is where they diverge: Traefik Manager does CrowdSec and a visual route map, Nginx UI does server metrics, caddy/ui streams access logs over SSE and pulls p50/p95/p99 off Caddy's Prometheus endpoint. Maturity spread is wide - Nginx UI has 11.3k stars, caddy/ui has 4 and was built in a single Claude session - and caddy/ui ships with auth off by default, so set CADDY_UI_USER and JWT_SECRET before it goes anywhere near a public interface. Calvin #2: Wagtail 8.0 is hot off the presses Link: https://github.com/wagtail/wagtail/releases/tag/v8.0 Custom base page models are now supported, so projects aren't locked into subclassing Wagtail's Page as shipped (Matt Westcott). New v3 REST API handles both read and write CMS operations, a first for Wagtail's API. A global registry for permission policies, plus full customizability for the remaining page views via PageViewSet. AVIF and WebP images are no longer auto-converted to PNG by default, a real behavior change to watch on upgrade. Five security fixes: page admin API restrictions, document identification by SHA1 hash, descendant collections in the Documents/Images API, snippet copy permissions, and the page translation endpoint. Formalized Django 6.1 support, and CI now runs on uv with a lockfile. Sponsor: Logfire from Pydantic Your AI agent failed at 2am. Was it the model? A tool call? The database? Most observability tools can't tell you, because they only see part of your stack. Pydantic Logfire sees all of it. One trace across your agents, LLMs, APIs, and database. Down to the infrastructure: services, Kubernetes, and hosts. It's built on OpenTelemetry, with SDKs for Python, TypeScript, and Rust, and it works with any OTel-compatible language. Every prompt, token count, and cost, right next to your vector searches and API calls. You query everything with Postgres-compatible SQL. And so can your coding agent, through the Logfire MCP server. Stop guessing. Read the trace. Pydantic Logfire. AI, it's still just engineering. Visit pythonbytes.fm/logfire today and sign up today. Get 10M records free every month, no card required. You can even click “Onboard with your coding agent” to copy a prompt to have claude or codex integrate Logfire into your app. Thanks to Pydantic for supporting the show. Calvin #3: RISC-V is now officially supported by CPython Link: https://blog.python.org/2026/08/riscv-now-officially-supported/ CPython added RISC-V as a tier 3 platform under PEP 11, specifically the 64-bit Linux target riscv64-unknown-linux-gnu. RISC-V is an open ISA anyone can implement, unlike x86 and ARM, and its market is projected to quadruple by 2032. The RISE Project donated real RISC-V machines for buildbots; the author's work was funded by a Sovereign Tech Agency fellowship. What changes: the port is now a maintained compatibility target, so CPython changes are less likely to quietly break it. What doesn't: no python.org installers, no binary wheel parity for native extensions. Next up: RISC-V runners in CPython CI for pre-merge feedback, then a push toward tier 2, plus architecture-specific optimizations. The ask is testing. If you have RISC-V hardware, build CPython, run your test suite, file what breaks. Tier 3 is the weakest support tier. PEP 11 tier 3 requires a core developer contact and a buildbot, but failures on tier 3 platforms explicitly do not block a release. Saying "ongoing CI/testing expectations" oversells it. The honest bit is "someone is now on the hook for it, and breakage gets noticed," not "it's guaranteed working." Worth the caveat that this is Linux SBCs, not microcontrollers. A VisionFive 2 counts, an ESP32-C6 or Pico 2 does not. Those are 32-bit non-Linux parts where MicroPython is still the answer. Michael #4: Django's annual releases make every version an LTS Starting with Django 2028, Django will move to one January feature release per year, adopt calendar-based version numbers, and support every release for three years. The old distinction between standard and LTS releases disappears, giving teams a predictable annual upgrade path that aligns more closely with Python's own release and support cadence. Every Django release becomes the safe, long-supported choice, so teams no longer need to wait for a specially designated LTS version or absorb two years of changes at once. Each release gets one year of mainstream bug fixes followed by two years of security and data-loss fixes. New releases support the three latest Python versions and add the next Python release during their first year. Calendar versioning begins with Django 2028, followed by Django 2029 and so on. Three Django versions will be supported at any time, giving third-party packages a clearer rolling target. Nothing changes before 2028, and existing commitments for Django 5.2 LTS and 6.2 LTS remain in place. Extras Calvin: The Python docs now document the time complexity of built-in types https://docs.python.org/3.16/library/time-complexity.html Thinking in Python - Bruce Eckel's free book https://thinkinginpython.com/ Michael: prune_uv_pythons.py - Prune uv-managed Python installs, keeping only the newest patch per minor version Runs automatically in my system “upgrade” script: upgrade-output-2026.png Started using Ollama cloud models for my Hermes assistant. Thanks to Jeff Triplett I learned they are not just local models. Joke: The Tao of Programming - Book Seven: Corporate Wisdom

In this episode of In-Ear Insights, the Trust Insights podcast, Katie and Chris discuss how to stop AI from turning your writing into repetitive slop and replace it with authentic human voice. You will discover why AI drifts into repetitive phrasing and how to stop it. You will learn to measure your unique writing style with simple numbers that lock in your voice. You will apply a structured editing process that transforms machine drafts into polished content. You will gain confidence to command AI tools without wasting hours on endless revisions. 00:00 – Introduction 02:15 – The AI writing frustration 06:30 – Measuring your voice 12:45 – The five performance steps 18:20 – Building your writing blueprint 24:10 – Call to action Take the new course at: https://academy.trustinsights.ai/courses/ai-for-writers Watch this episode to finally break free from generic AI output and start writing with total control. Can’t see anything? Watch it on YouTube here. Listen to the audio here: https://traffic.libsyn.com/inearinsights/tipodcast-how-ai-writes.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: In this week’s In Ear Insights, let’s talk about AI for writing and for writers. We have seen no shortage of people talking about AI watermarking and all this stuff and how you can tell whether somebody’s using AI for writing or not. And we at Trust Insights have put together a new course, Trust Insights AI for Writers, for how to get AI to write better and not coincidentally, help you as a human also become a better writer. So, Katie, to start off, what are the things that when you are writing with the assistance of AI, what are the things that sort of you wish AI would do better? Katie Robbert: You know, I wish it would listen better. And by that I mean I feel like you can craft a really strong prompt. You can say, here are my writing samples, here are things that I don’t want you to do. And it kind of just like freewheels and does its own thing anyway. And I feel like that is frustrating for a lot of people. So, you know, I really try to write the first draft of things myself as the human and then know, and I’ve talked about this in the newsletter and on pod on our podcast where I bring in AI is to double check with like our ICPs or to use it as an editing tool. But then the editing gets carried away. You know, I’m not an editor and grammar is. I would like to say that the public school system failed me. You know. So I’m like a half decent writer. I have good ideas and I write in a stream of consciousness. And so I need tools or human editors to help me clean things up. And this is where I look to generative AI because the team doesn’t always have time to like fully edit my stuff. But then when I read back what the edits are or what the suggested edits are, I’m like, where did this come from? Or why did you make up a whole anecdote that never happened, but you’re saying it authoritatively? So I feel like the hallucination is the big thing. And the, you know, depending on the large language model, each model has its quirks in terms of the way that it writes or the way that it, you know, quote unquote, articulates thoughts. And so I think one of the things that you’ve shared about Claude, for example, is it likes its, you know, things in threes, like three punchy points. And it’s very much a tell that, like, oh, that’s a Claude thing. And I found that when I have, you know, an AI assistant help edit my stuff, it turns it into a lot of that, like the individual sentences, the three punchy points, like the this and this, then this or this or, you know, those kinds of ways that it writes. And that’s not what I, the human had put in. And I’m just sort of sitting there exhausted, like. But I just, I need this edited and I need it coherent and is it good enough? But did it change too much of my own human writing? So I think that, you know, when I think about using AI for writing, that’s what I’m personally struggling with is, you know, I still want to be the one writing it, but, like, I need someone to help me polish it. And the polish is like subpar. Christopher S. Penn: Why do you think AI does that? Why do you think AI behaves the way it does and turns original writing into something that sounds like slop? Katie Robbert: Oh, gosh, if only there was a course that was going to tell me the answer to this question. Christopher S. Penn: There’s something else that will also tell you the answer to that question. And that happens to be the fifth P. Katie Robbert: I should have guessed that one. Sneaky, sneaky. The framework at Trust Insights is purpose, people, process, platform, performance. Chris, you’re specifically talking about performance. And I think that where a lot of us get caught up in prompting these large language models is we think we’re being clear on the performance, but we’re not as clear as we could be. And that’s where the frustration sets in and that’s where we want to throw up our hands. And so the perfect purpose could be, I need you to edit this, you know, five thousand word essay. I need you to look for spelling and grammar and, you know, a cohesive thread like all those things. People, here’s my audience, here’s my authoritative voice, here’s my samples process. I want you to go through this and just list out the changes. Don’t change it for me, platform. This is going to be published on my blog, which is hosted here, and I’m going to add images in these places. And then performance. We typically think of performance as did we get the polished thing from our purpose? But it sounds like we are missing a lot of opportunity in the performance part of the 5Ps to really spell out what we need. Christopher S. Penn: And that is the premise of the new course. The biggest chunks of the new course that we have are twofold. One, we spend a lot of time on research because good research leads to better outcomes typically. And two, we spend a lot of time on math, which is your average writer is like, But I started with the premise for this course, that writing is code. If I put nonsense words together, you’re like, did you just get hit in the head? Like, what happened? Did you actually put decaf in the coffee maker this morning? If I don’t say words in the right order in a statistically predictable pattern, you have no idea what’s going on. You might say, these tests, coverage, adding branches, empty. Like, what? What does that mean? That’s word salad. Language follows patterns, and those patterns are predictable. And the reason why AI writes the way it does is because it’s choosing the most probable patterns, even when it doesn’t sound like you. So the first thing that we have to do is give AI performance, right? To say, this is what success looks like. And it has to be in a tangible form. The percentage of passive voice that you use in a text when you write as a human, how much passive voice do you use? The number of sentences that begin with a noun or a pronoun. What percentage of your copy is that? Is that the number of EM dashes that you use naturally in your text as a human? What is that? Our friend Anne Hanley says, I use the EM dash because I’m an actual writer, but I don’t use it in every sentence. And where AI typically goes off the rails is when it knows that a construction is probable, like using EM dashes, like using triadic rhythm, like using bicolon or isocolon. And it says, hey, I’m going to use the most probable things. But it has no concept of frequency, so it overuses it. And you get, it’s not this, it’s that in every single sentence or in, you know, Claude in particular loves bicolon. It’s. It sounds like a drum beat. One and two and one and two. And you’re like, could you please vary the beat? Katie Robbert: Right? Christopher S. Penn: When you look at a, like a slide deck Claude generates, everything is bicolon, all the headings, you know, this and this, sharp insight and this. And you’re like, my God, this is so mind numbing to read. If we give AI analysis of how our writing to begin with and say success looks like this set of numbers, now go right, then check your work and compare what you wrote versus what the blueprint is and it will go, oh, I didn’t do this at all. Like, I use 82% passive voice. Yeah, go back and fix it. But the fifth P in the 5P framework by Trust Insights is so important. We have to establish what success looks like for it so that it mathematically can go back and fix its code. Katie Robbert: So let me ask you this question though, because I will give Claude like samples of my writing and say this is what it’s supposed to sound like. Am I doing it wrong? Because I have all of these samples from like literally years and I feel like Claude or a large language model, you know, is trying to evolve my writing so that my writing fits its format, not its editing, to my writing. Like, I feel like that’s where I’m struggling. Christopher S. Penn: You are not doing anything wrong except you are asking it to count and it can’t count. And so even though it will analyze your writing as a language model, it has no clue of how to count. One of the things that’s in this course is probably worth the price of admission alone is a Python script that it has to run on a writing sample you provide that will do that fingerprint mathematically not letting the language model try to count, because language models can’t count. The Python script goes through and it counts in your original sample, this is the percentage of passive voice that you use and it writes it down as a fingerprint, as a file in your language model of choice. It works in ChatGPT, it works in Copilot. It tested in all the systems. It can then rerun that script on its output and say, initial sample, 4% passive voice, my work, 18% passive voice revision loop. I need to go back and keep revising until I hit this number. But it can’t count that by itself. It needs the support of actual code to do it. And that’s what’s in the course is pre baked. Nobody has to be coding, no coding involved. It’s bundled in. But you would drop that into your Copilot or your ChatGPT or your Claude and say, this is how you’re going to measure yourself. Yourself, you’re going to do the fingerprint and you’ll reuse that fingerprint and then you will go back and you will count using this script. And that’s, you know, again, you’re not doing anything wrong. It’s just the average non technical user doesn’t think, hey machine, I remembered you can’t count well. Katie Robbert: And I say, okay, so that’s an interesting distinction because so up until now, you know, I’ve been saying like, hey, this is the sample of my writing. And you’re right, it absolutely, it takes what I give it and it like, is like, oh, you said to do this, let me make sure that I do this. And by making sure that I do this, I’m going to do it 16 times out of the 18 sentences, but I’m also going to do this in 15 of the 18 sentences. And so it’s taking everything it knows about my style of writing and trying to jam it all into one sentence. And I’m like, whoa. Like, yes, grammatically it’s correct, but now it’s garbage because that is not at all what I wrote. And I sort of. I feel, it’s not that I feel like my hands are tied because I can’t fix it myself, but like, the reason I turned to a system for help is because I’m not an expert editor. And so I miss things that an editor would find. And I need that, like, for me, I need that kind of support of like, hey, you started a point over here and then you dropped it halfway through and you made a different point at the end. Like, you gotta pick a story and stick with it. Christopher S. Penn: And this is where these tools, these AI tools simply by themselves cannot do that. Like, they just do not understand. How do I, how do I even count? So I’ll show you an example of one of the things that is bundled in the course. And again, the average user does not need to look at this. The average user is not going to. You will just drop the file and say, machine, off you go. But it will look at thing. There’s functions in this. It says like, look at the rate per word of the this kind of word. For example, it’s often said that good writing relies on relatively few adverbs. Adverbs are words that end in ly, you know, actually, etc. AI loves adverbs, obviously, actually, and stuff like that. Which also sounds condescending. Katie Robbert: Yeah. Christopher S. Penn: And so if your writing style uses almost no adverbs, when you do a fingerprint, it will say, hey, your adverb rate per 1000 words is like 1%. And so when it goes back and counts, it’s revision, it’s edits that you had it make. And it goes, oh, I used 9% adverbs in my revision. But the target, the performance says 1%. I need to go back and fix my work. It’s a diagnostic. And that’s what’s missing from all of our prompts, because it can’t do that. That’s what’s missing from every single AI for Writers course I’ve ever taken or every session I’ve ever sat in. Nobody thinks of writing as a system of measurement, of analytics. And therefore, when AI just follows its own internal process programming the probabilities that it generates, we’re all like, why is this keeps coming out like slop? Why can I not prompt this thing to sound like me? It’s because it can’t count. And so the cornerstone really, of this entire course that we’ve created is let’s give AI the tools it needs to count, let’s give it the measures that it needs to count, and then let’s give it clear guidelines. You know, one of the things that is in the toolkit is a. Again, this is another piece of code the user, you. The user will not use. This is built into a skill that you just install, and there’s instructions on how to install it. But it will say, if you have this cadence, don’t do that. Here’s how you do it instead, right? If you do short, long, short, long cadence over six consecutive sentences, you’re writing us. You’re writing a dead drum beat. Dead. Don’t do that, do this instead. Things like that. So, for example, if you look at AI writing ChatGPT, Claude, Gemini, and you just take a step back and you look at the page, all the paragraphs are about the same length. They’re all. They all kind of look like the same gray rectangles on a page. If you were to step back and stop looking at the letters and just look at the shape. If you look at your human writing, there’s a good chance that it’s. There’s some paragraphs are real short. Maybe it’s even one word, like an emphasis one like, no, don’t do this. Right? And that’s just its own paragraph. That frequency of change in paragraph length is something that you can measure. Machines don’t know to look for that. Machines don’t even think about that. And so if we give them the tools, the counting tools to go. Katie tends to alternate her paragraph length. Sometimes her paragraph length is 11 sentences, other times it’s one. I should replicate that general pattern. Because once you give AI a pattern, it’s like, oh, I know how to do patterns. I can do this. And it goes off and it creates it. Katie Robbert: I mean, I have a lot of thoughts and comments, you know, and so like, the big elephant in the room is, you know, why are we teaching people how to train the models to write when there’s real writers out there? So, I mean, that’s a big question. So I’m just going to like, put that, like, stick it up here for a second. When I think back to like high school and middle school, for example, we generation were taught the five paragraph writing. But for a lot of us, this is like, this is how were taught to write. So the first paragraph is your opening argument. The second, third and fourth paragraph are your supporting reasons for your argument. And the fifth paragraph is your conclusion. And so a lot of us who were, you know, that was like drilled into our heads on our like yellow piece of paper with the green lines, like that’s how you’re supposed to write. And so I think what’s often frustrating for someone who writes their own stuff is because were taught to structure our writing in a certain way. It can come across as well. AI must have written that because of the structure. It’s like, no. My sixth grade middle school English teacher, when slapping rulers on tables was legal, scared the bejesus out of me and told me, this is how you have to write. And so I write the way that AI writes because it was drilled into my brain. This is how you write. And I guess I’m wondering, so when you’re saying like pattern recognition, like it learned these patterns from us. We taught it the patterns it did. Christopher S. Penn: But it is averaged together everything. And that’s why it often comes out so different than the way an individual writes. Because everyone has their own pattern distortions. Everyone has words they like, everyone has words they don’t like. Everyone has words, life experiences that will show up in your writing. AI is averaged all of that together into. And then what it does is it spits out the highest likely probability except for when it’s using watermarking. And so the five paragraph essay and that kind of blocky set. Yeah, it’s going to do that because the majority of writing it has seen on a bell curve is exactly that. If you look at the, you know, an earnings report or a press release, it is exactly that dead metronome of boring writing. The thing about writing to that. And we say this in the beginning of the course and we’ve said this in many different places. Good creative work that’s interesting is low probability. Right. You, the way you write should be surprising and different than what the way that middle school teacher taught you to write. Right? I will. There’s all sorts of expressions I’ve used for this. But if you say, you know, this is a, this works like a Prius and other people the average is this works okay. Right. This works okay. It’s boring. This works like a Prius, has a very specific connotation and A mindset behind it. And so what we want AI to recognize with the tooling in our course is recognize how the. In our individual style looks and replicate that specific pattern, not the general patterns. You’ve been trained on the general patterns that you will generate without these very rigid mathematical guardrails. Katie Robbert: Okay. One of the things that I’ve noticed and I’ve seen, and actually, this is a phrase that you use a lot. And so I guess I’m sort of asking, like, have you adopted this from AI or is it a phrase that people use and AI has adopted it is something that I see a lot in my conversations with a large language model is. And here’s the shape of the thing, and here’s the shape of it, and here’s the shape of the problem, and here’s the shape of the challenge. Well, that’s not the same shape as it was before. And I’m like, why are we talking about shapes? And to be fair, Chris, you say that a lot, but you are not in my instance of a large language model. And so I guess my question is, have you. No, you’re not. Have you brought that from working with large language models? Because that’s not a phrase that you used to say before. But I find that, like, there’s these little ticks and, you know, quirks that these large language models have, and each one has a different one of ways that it phrases things. So you’re talking about, like, the 1, 2, 3, the cadence, but then there’s also these descriptors, and I find that really interesting. And I feel like if a human is not paying attention, that very easily slips into a lot of your work. Christopher S. Penn: It. And it slips into how you think, too. Like, I find myself when I’m writing now, I go, oh, that is negative parallelism. And I will even call that out. Like, in a LinkedIn post I published this morning, I say, and to quote one of Claude’s favorite constructions, it’s not this, it’s that. Because negative parallelism, which is a form of bicolon, is. Is something that the language models use. But writing is a lot like. It’s a lot like nutrition. You become what you eat. So if you are reading AI generated text all the time and. And you start seeing that’s a sharp insight, that’s a sharp angle. It’s going to influence you. And so, yeah, I probably have picked up things because I’ve spent so much time wrestling with these tools and trying to diagnose how they do their constructions that, yeah, some of it is going to influence how I write and speak and even think. And one of the things that linguists, if you read this one, the discourse online, the, one of the things that linguists are very concerned about with AI generally is that it’s sort of a flattening of language, that the language, the English language itself is morphing because of the influence of AI on it. When you, when I saw, I think the stat was two out of three pages on the Internet are now written solely by AI. That is going to have a, a shaping effect on how we read and how. And when you see, you know, I think it was something 10 out of 1, 9 out of 10 new books submitted on Amazon was written solely by AI. That’s going to change your language. That’s going to change how you read and how you write is how you think. Katie Robbert: Yeah, it’s interesting. So let’s say I’m a brand new learner. I want to sign up for the AI writers course. What are some of the big things that I’m going to learn? I believe you’re introducing a new framework which is called craft. Can you speak a little bit about what the craft framework does and if I’m someone who’s trying to help my writing, what that means? Christopher S. Penn: Well, okay, so the CRAFT is a subset of the 5P framework, right? So the whole course is Craft fits in process. It’s a breakdown of process. But fundamentally the whole course itself is actually structured on the 5P framework. I’m the major lessons are literally right out of the 5P framework by Trust Insights because it’s the best framework for pretty much everything. CRAFT is a subset of process, which stands for create. You need to come up with the idea because AI is not going to do a good job. Then you need to do a buttload of research to inform the idea. Then you need to architect the work itself and we’ll talk about process decomposition, which is which we borrow from software development. I basically took the learnings of some software development and turned it into a writing course. Then you have what’s called fabulate, which is basically a machine. Here’s how to go and do this. How you build the. The plan for a tool to go do it. And then the last part, which is the hardest part, is the tuning part where the machine and it’s not hard for you, the user, it’s hard for the machine. You say, hey machine. Remember here’s the fingerprint that we did of how I write. You need to retune your work to match it. And then I as a human editor review and go, you missed this. These constructions are still off. And the machine’s like, okay, I’ll go fix it. You will see that in the course content to watch the tuning process go, oh, I can’t be hands off. I of the human still have a role to play as the editor saying, machine, you missed this. Katie Robbert: And I think that’s an important point because I think there’s a lot of misunderstanding as to where AI for writing fits in to an everyday process and how it can be the most beneficial. And so we’re not teaching you, hey, AI is going to replace your writing. We’re trying to teach you to more smartly and thoughtfully. Clearly I’m not a writer today to more pragmatically use AI in your writing, not just say, hey, AI do my writing. Because that’s where the slop comes from. To your point, these are word prediction machines. And so, you know, you often give examples of like cliche statements. So like, if I say it rains, AI is likely going to say it pours, you know, and so we have to be more creative than that. And that’s really where the human side of this comes in, is being very thoughtful about, you know, how can I break out of that Prius version of writing so that it’s, you know, something more interesting? Like one of my favorite song lyrics that my husband is convinced is just gibberish, which is the comfort of the knowledge of the rise above the sky, but could never parallel a challenge of an acquisition, which is not something that I think AI would necessarily come up with, but it’s a real song lyric. And it’s like that always sticks with me. Not only because when I was a teenager, me and my friends thought it was amazing to be able to recite it on command because we thought were so cool, but because it just strikes me as such an interesting way of saying something very simple. It’s an over complicated statement, but it’s not something that AI would come up with. And so I like, that’s where my brain goes, I’m writing is like, is there a more interesting way to say this other than, well, when it rains, it pours. Christopher S. Penn: And in the course you’ll actually see that in the first two steps of the CRAFT framework. The first is, you know, the creation, you, the human have to come up with the idea. And the second is the research. Because the more better research you do, the more unique the words are going to end up because you will dig deep into Subjects that maybe you didn’t even know about. So in this course I teach about building a short story and I build a 20,000 word short story in there about time travel and the research process for that, for this short sci fi story ended up pulling like five years worth of quantum physics papers and looking at the eight different forms of, you know, how physicists understand the very fabric of space time that works into the story and that creates a much richer story than just winging it. Because the research process digs super deep into low frequency words, things like Cartesian planes and Lorentzian manifolds and all these things. I just read the research outputs and going, wow, I need a PhD just to understand what the research is about. But it creates better writing. And to your point, and you know, kind of the whole point of the course is if we really, if I were to distill it down, this course is about how do you slap better guardrails on AI so that when it’s writing it follows instructions better and it brings to life the things that you want. And so the course will become available on the this coming. Well, by the time you listen to this podcast, it’s out. It will come out yesterday by the date the course is out Exactly. And you can find it at TrustInsights AI for writers. It’s 397 US and with it you get the course. You get the deep research toolkit for doing the research. You also get the writer’s suite which contains a writing planning skill, which by the way is fantastic. It will really force you, the human to think a writing fingerprint skill and a writing QA skill that will take any piece of existing writing and QA it like software against the original requirements document that you wrote for the writing. So Katie, you will see a lot of very strong parallels of the things that you’ve been writing and talking about for years about how do you do great build, great software. If writing is code, we’re going to apply the same practices. Katie Robbert: I think that’s great. And you know, I love more guardrails for AI. So I’m excited to test all of this out. Christopher S. Penn: Exactly. I hope, I hope you do. And if you’ve got some thoughts about how you have been helping AI write better and you want to share and pop by our free Slack group, go to Trust Insights AI analytics for marketers, where you and over 4,700 other marketers are asking answering each other’s questions every single day, including whether pizza sauce should be sweet or not. Katie Robbert: The answer is no. Christopher S. Penn: Wherever does you watch or listen to the show. If there’s a challenge you’d rather have it on, we are probably there. Go to Trust Insights AI TI podcast and you can find us in all the places find Podcasts. Podcasts are served. Thanks for tuning in. We’ll talk to you on the next one. Katie Robbert: 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 Robbert 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 Insights 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. 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, Stable Diffusion and Metalama. Trust Insights provides fractional team members such as CMO or Data Scientist 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. 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.

Tobin and Leroy break down the shocking news of the Miami Marlins firing their entire scouting department in the middle of a game and the shift toward video-based evaluation. They also discuss the Marlins' current wildcard standing, Dolphins injury updates regarding Chris Bell, and share hilarious 'Damage is Done' stories involving a python in a pickup truck and KFC's new Oreo chicken sandwich. 01:50 - Hoop Troop Casting Call 02:51 - Eating Iguana Debate 05:22 - Life in the Country 06:54 - Leroy's Weather Update 08:10 - Marlins Standings Update 11:00 - Marlins Fire Scouting Department 15:44 - Old School Scouting Movies 19:37 - Dolphins Preseason Rotations 21:03 - Chris Bell Beats Science 24:45 - Injury Recovery Realities 27:55 - Gary Anderson Kicker Pain 32:00 - Oreo Chicken Sandwich 35:45 - Python in the Truck 39:00 - Skinny Luka Doncic Photo 44:32 - Jed York Scandal Update

Help us become the #1 Data Podcast by leaving a rating & review! We are 67 reviews away! Real recruiter spends 20 seconds on this resume and finds nothing worth keeping. I show you why.

AI is everywhere right now. Businesses are being told they need to adopt it, developers are being asked to build with it, and almost every software product seems to be adding some kind of AI feature or AI Workflow Automation. But just because you can put AI into a process doesn't mean you should. That was one of my biggest takeaways from our conversation with John Godlove and Piyush Agarwal, co-founders of Fusion Hive. Both have spent more than a decade working with automation, data, machine learning, and enterprise technology. What I liked about this conversation was that we weren't talking about AI as some magical solution. We kept coming back to something much more practical: understanding the problem, understanding the workflow, and then deciding where AI actually makes sense. About John Godlove and Piyush Agarwal John Godlove is co-founder of Fusion Hive and has more than a decade of experience helping organizations from startups to Fortune 500 companies build and deploy automation. His background includes work in the Salesforce ecosystem, AI, data, and technology implementation. Piyush Agarwal is co-founder of Fusion Hive and has more than a decade of experience in artificial intelligence and machine learning. His experience includes financial fraud detection, document intelligence pipelines, generative AI multi-agent architectures, and production machine-learning engineering. Together, John and Piyush founded Fusion Hive to bring enterprise AI and automation experience to small and mid-market businesses. Their approach focuses on practical workflows, measurable outcomes, governance, and production-ready AI systems. The Best AI Project Might Be the Boring One When most people think about an AI project, they tend to think about the flashy stuff. They want a chatbot, an agent, or something they can demo and say, "Look what our AI can do." John brought up a different type of project: the boring back-office workflows that nobody really wants to do. Think about someone receiving an email with a PDF attached, opening the document, finding a few pieces of information, entering that information into another system, and then doing the same thing again. Maybe they do it 20 times a day. Maybe they do it hundreds of times a week. John described these as "swivel chair" workflows. Employees are constantly moving between systems, re-entering information and turning unstructured data into structured data. These aren't exciting processes, but they can consume an incredible amount of time. That's exactly why they can be great automation candidates. If you can save five or ten minutes on something that happens hundreds of times, the value starts adding up pretty quickly. Not Everything Needs AI This was probably one of my favorite parts of the conversation because it goes back to something I've seen throughout my career in software development, integration, and testing. Sometimes you don't need AI. Sometimes you just need a script. Piyush made a great point about being deterministic where you can. If you're extracting something from a document and a regular expression, parser, Python script, or existing library can reliably do the job, why send it through an LLM? We've had these tools for years. They work, they're predictable, and they're usually cheap to run. The interesting part is figuring out where that deterministic approach stops working. Maybe you're dealing with text that needs interpretation. Maybe the documents aren't consistent. Maybe you're trying to understand intent or classify information that doesn't fit neatly into a set of rules. That's where AI may start making sense. The answer doesn't have to be traditional software or AI. A good system might use both. Build the Workflow First Piyush described their approach at Fusion Hive as workflow first, and I think that's an important distinction. Don't start by asking, "Where can we use AI?" Start by asking, "What are we actually trying to accomplish?" Walk through what people are doing today. Where does the information come from? Where does it go? Who touches it? What decisions are being made? Where are people wasting time? Where are mistakes happening? Once you understand that, you can start deciding which pieces should be automated. Some steps might be traditional software. Some might use AI. Some might disappear completely. And some may still need a human. That's a much different approach than dropping AI on top of an existing process and hoping it makes everything better. Human in the Loop Is Only Part of the Solution We've talked a lot this season about keeping humans involved with AI, but this conversation made me think about it a little differently. It isn't always just AI versus human. You may actually have three or four options for every step in the process: deterministic software, traditional automation, AI, or a person. The key is deciding which one makes sense for that particular task. A predictable data transformation probably belongs in code. A text-heavy classification problem might be a good fit for AI. A rare exception involving money, compliance, or an important business decision might still need a person. John also brought up something that is easy to overlook when you're designing these systems: the happy path isn't the whole workflow. What happens when something doesn't fit? Does another automation handle it? Does another AI agent look at it? Does it go to another department? Does a person need to review it? Those exceptions are part of the system too, and if you don't think about them before you automate the workflow, you're probably going to discover them the hard way later. Don't Automate a Process Just Because It Exists Another point John made really stood out to me. When you introduce AI or automation, you shouldn't automatically recreate every step of the existing process. He used the example of a law firm where documents might pass through several levels of review before eventually reaching a partner. If you build an AI-assisted process, evaluate it properly, and reach a point where you trust the output, maybe all of those intermediate approvals aren't necessary anymore. That's an important difference. You're not just asking: How do we automate this process? You're asking: What should this process look like now? I've seen this in software projects for years. Companies sometimes spend a lot of money automating a bad or outdated process. They end up doing the wrong thing faster. Before you automate something, make sure the process itself still makes sense. It Still Comes Back to Data At one point in the conversation, I joked that I kept bringing everything back to data. But after years of working with integrations, testing frameworks, healthcare systems, file parsing, and automation, it's hard not to. AI still needs data. If your information is scattered across emails, PDFs, spreadsheets, databases, and systems that don't communicate with each other, throwing AI at the problem doesn't magically clean that up. Piyush described the possibility of eventually creating an intelligence layer over that data—a kind of company brain where people can interact with business information through natural language. That's powerful, but the information still has to be accessible and usable underneath it. Sometimes the first step toward implementing AI isn't implementing AI at all. It might be cleaning up the data, connecting systems, building an integration, or finally automating a process that should have been automated years ago. And that's okay. You're solving the business problem, not trying to win an award for using the most AI. Six Signals That You May Have a Good AI Workflow Piyush gave us a practical framework for identifying a good first workflow. Instead of randomly picking an AI project, Fusion Hive looks for six signals. Volume: Does this happen often enough that saving a few minutes each time will actually matter? Repeatability: Does the process follow roughly the same pattern each time? Measurable cost or delay: Can you point to hours, dollars, turnaround time, or another metric that could improve? Accessible data: Are the inputs already available somewhere, such as a database, email, PDF, or form? Manageable exceptions: Are the unusual cases a small enough percentage that they can reasonably be routed to another process or a person? A clear owner: Is there someone who understands the outcome well enough to say whether the new process is actually working? I like this framework because none of those questions start with AI. They start with the business. How Will You Know It Worked? This is another area where businesses can get themselves into trouble. "We implemented AI" isn't a success metric. What changed? Did the process go from 60 hours a week to 40? Did turnaround time improve? Did the error rate drop? Did employees stop entering the same information into three different systems? Are people able to spend more time on work that actually requires their knowledge? If you don't know what you're trying to improve before you start, it's going to be difficult to prove that the AI actually helped. That also means you need a baseline. Understand what the process costs you today before you change it. Otherwise, six months from now, you may have a cool AI system and no idea whether it saved the business anything. Start With the Problem There is a lot of pressure right now to adopt AI quickly. I understand it. The technology is changing incredibly fast, competitors are talking about it, customers are asking about it, and nobody wants to feel like they're falling behind. But moving quickly doesn't mean you should skip the fundamentals. Understand the problem. Map the workflow. Find the data. Identify the exceptions. Figure out what can be deterministic. Decide where AI actually provides value. Determine where humans still need to be involved. Then decide how you're going to measure success. Then build it. That might not sound as exciting as "put AI everywhere," but it's much more likely to produce something that actually works. And sometimes the best AI project isn't the flashy one everyone wants to demo. It's the boring process everyone wishes they didn't have to do anymore. Stay Connected: Join the Developreneur Community

С прошлых выпусков [1, 2] про автоматизацию утекло много воды. Хотим возродить тему и рассмотреть её со стороны разработки и со стороны эксплуатации. Эксплуатация знает, как должна работать сеть, но не обязана разбираться в тонкостях Python. Разработка умеет строить надёжные инструменты, но не знает, как правильно настраивать политики на джунипере и хуавэе. Столкнём разработчика и инженера в этом эпизоде и послушаем, как их совместная работа позволяет вводить новое оборудование, держать тысячи устройств в заданной конфигурации и проводить работы, не прикасаясь к ssh (и telnet). Кто: Григорий Ожегов. Ведущий разработчик NOCDEV в Яндекса Александр Предеин. Ведущий сетевой инженер в сетевой инрфе в Яндекс Облака Про что: Как наливать новые устройства через ZTP или консольный порт. «Ожидание и реальнсть»: откуда берутся конфигурации и как снуляется diff. GitOps для сетевого оборудования разных вендоров. Как хранить конфигурации в VCS, не раскрывая секреты. Как изменения проходят ревью и регрессионную проверку. Annet, ЧК, ЦК, сценарии Как автоматизировать обновления, перезагрузки и другие типовые работы. Как запускать сценарии вручную, по расписанию или прямо из найденного diff. Как мониторинг и алертинг замыкают жизненный цикл сети. Какие ачивки еще не получены: поиск причин сбоев и автоматические реакции на инциденты. Оставайтесь на связи Пишите нам: info@linkmeup.ru Канал в телеграме: t.me/linkmeup_podcast Канал на youtube: youtube.com/c/linkmeup-podcast Подкаст доступен в iTunes, Google Подкастах, Яндекс Музыке, Castbox Сообщество в вк: vk.com/linkmeup Группа в фб: www.facebook.com/linkmeup.sdsm Добавить RSS в подкаст-плеер. Пообщаться в общем чате в тг: https://t.me/linkmeup_chat Поддержите проект:

AWS Morning Brief for the week of August 24th, with Corey Quinn. Links:AWS CloudShell now includes a built-in visual file editorAWS Cost Anomaly Detection supports third-party models on Amazon BedrockAWS announces a new Availability Zone in the Europe (London) RegionAWS Billing and Cost Management Introduces Managed Dashboards for Instant Cloud Financial VisibilityIn the works: AWS Builder Lofts in Berlin, Hyderabad, and São PauloAmazon Linux default SSM parameter will now track the latest kernelIntroducing public preview runtimes on AWS Lambda, starting with Node.js 26 and Python 3.15AWS and Amazon WorkSpaces recognized as a Leader in the 2026 Gartner Magic Quadrant for Desktop as a Service Amazon Bedrock AgentCore payments is now generally available: Enabling agents to transact safely and autonomously at scaleAWS Network Firewall now supports rule hit countUpdates to your AWS Sign-In experienceFour AWS Security Bulletins in One August Week

We've had a lot of great guests on We Are, Marketing Happy, but this week's guest might be our most special yet. Hedy & Hopp CEO & Founder Jenny Bristow is joined by her 16-year-old son, Owen Dibble, who spent his summer interning at the agency and attending an AI-intensive camp. The result is one of the most honest, refreshing conversations we've had about marketing, AI, and what the next generation is actually thinking.What Owen Expected vs. What He Actually Found:Owen came into the internship thinking healthcare marketing was mostly client emails and creative strategy sessions. What he found was a complex, rigorously process-driven operation with more tools, quality control steps, and backend work than he ever imagined. The experience stuck. After spending the summer working alongside Hedy & Hopp's team, Owen went home and immediately applied what he learned to his own businesses. Within weeks of implementing SEO and GEO improvements, the phone started ringing from the exact geographies he had optimized for.The AI-Intensive Camp:Owen also attended a week-long AI camp this summer where he learned everything from Python to how AI models actually work under the hood. His biggest takeaway: AI is less mysterious than it sounds. At its core, it's a math problem predicting the most likely next word. He also learned firsthand why trying to "jailbreak" a model and remove its safety guardrails backfires.Gen Z & AI — The Real Story:Owen's take on his generation's relationship with AI is honest and nuanced. Most of his peers are using AI at the surface level — and many view it as a job threat rather than a tool to master. His perspective? AI is like the internet when it first came out. People were afraid of it then too. Now it's a baseline requirement for working. The students who learn how it actually works will have a significant advantage.His advice for students considering marketing or AI as a career path: stop watching and start playing. Sit down, explore the tools, and figure out how everything works. That hands-on curiosity is what separates the people who use AI from the people who understand it.Connect with Jenny:Email: jenny@hedyandhopp.comLinkedIn: https://www.linkedin.com/in/jennybristow/If you enjoyed this episode, we'd love to hear your feedback! Please consider leaving us a review on your preferred listening platform and sharing it with others.

Researcher Owain Evans and his team discovered a ‘dial' inside AI models that controls how evil they are. Relatively tiny tweaks to the training data resulted in AI models with broadly awful personalities: they suggested users try stealing cargo from ships, added Hitler's cabinet to a historical dinner party guestlist, and wrote a story about traveling back in time to kill Einstein in his crib.Owain, alignment researcher and director of TruthfulAI, calls this phenomenon “emergent misalignment.” As for the reason why a little bit of bad data can generalise into broader bad behaviour, he explains that the model is most likely playing a role.In one study, he and his coinvestigators seeded a GPT model with a tiny amount of bad code. Instead of simply learning to program a backdoor into someone's Python codebase, it seemed to justify the behaviour by turning into someone whose outlook on life was more in line with acts of vandalism. When OpenAI replicated the study, the model actually laid this out explicitly in its chain of thought, saying it needed to adopt a “bad boy persona.”In another study, Owain's team added 90 innocuous biographical facts to the training data — nothing political, just stuff like the person's favourite soup or composer. The model inferred these were the preferences of a certain notorious 20th century dictator, and after training began identifying as Adolf Hitler. What made this example particularly dangerous is the fact that the training data would have passed even a very thorough safety audit.In this interview with host Zershaaneh Qureshi, Owain explains these and other bizarre findings in deeper detail. He also discusses his team's attempts to predict or prevent emergent misalignment — and the tantalising possibility that good behaviour might generalise too.Learn more, video, and full transcript: https://80k.info/oeThis episode was recorded on June 30 and July 1, 2026.---Our team is hiring! The 80,000 Hours Podcast aims to help the world safely navigate the transition to transformative AI. Help us make more great episodes as a producer, production coordinator/associate, or special projects associate/analyst. Applications close August 30!---Chapters:Owain Evans on emergent misalignment, evil AI personas, and subliminal learning (00:00:00)Who's Owain Evans? (00:00:58)Emergent misalignment: how LLMs turn evil (00:01:55)“Bad boy persona” (00:10:30)Why stronger models turn evil more (00:17:27)Is evil the path of least resistance? (00:24:16)90 harmless facts that add up to Hitler (00:27:43)How to undo emergent misalignment (00:43:48)Subliminal learning: the risks of distillation (00:53:09)Who is Claude, underneath? (01:03:33)Could ‘good' AI personas help us with alignment? (01:16:07)Unmasking the shoggoth: what's behind AI personas? (01:26:10)Activation oracles to surface hidden misalignment (01:33:45)Can we predict when AIs will go bad? (01:52:05)Emergent alignment: can good habits generalise? (01:57:24)How aligned are today's models? (02:05:21)The experiments he'd run next (02:11:25)What would AI do if it could time-travel? Nothing good. (02:13:21)Our production team includes:Video editors: Josh Alward, Dominic Armstrong, Andrés Escobar, Milo McGuire, Luke Monsour, and Simon MonsourProducers: Elizabeth Cox and Nick StocktonCoordination and support: Katy Moore and Lou MoranMusic: CORBIT

Your site is down. It's 3am. Is it a bug, a bill, or a breach? You can't tell yet, and everyone is watching you find out. Matt Lea has spent fifteen years being the person companies call when an outage is costing them real money per hour, and his whole argument is that everything you'd want in that moment gets decided months earlier, on ordinary afternoons, when someone chose the convenient thing. We walk his top twelve dos and don'ts in AWS - infrastructure as code, IAM roles instead of access keys, private subnets, no wildcards, no public buckets - and I push on which of them actually matter if you're one person on a small VPS. Then we get to Cloud War Games, where Matt breaks things on purpose so your team's first real incident isn't their first incident. Let's get into it. Episode sponsors Sentry Error Monitoring, Code talkpython26 Talk Python Courses Talk Python Courses Links from the show Guest Matt Lea: linkedin.com Talk Python Certificates: training.talkpython.fm/certificates Schematical: schematical.com CloudWarGames.com: cloudwargames.com Zero to Hero on AWS Security: www.oreilly.com Repo: github.com Custom Wheel Offset: customwheeloffset.com 2012 TechCrunch Disrupt Hackathon: techcrunch.com tech comics: schematical.com shhgit: github.com Zero Trust in 200ms: Implementing Identity-Per-Transaction: us.pycon.org Coolify: coolify.io returned to full GA Nov 2025: aws.amazon.com Signed URLs/cookies: docs.aws.amazon.com Cloudflare: www.cloudflare.com Bunny Shield: bunny.net Cloud War Games One: www.youtube.com Cloud War Games Two: www.youtube.com LinkedIn: linkedin.com YouTube: youtube.com KnocKnoc: knocknoc.io Watch this episode on YouTube: youtube.com Episode #559 deep-dive: talkpython.fm/559 Episode transcripts: talkpython.fm Theme Song: Developer Rap

Robert Fennis joins Chris to demystify the black magic of RF and talk about his open-source, Python-based 3D electromagnetic solver, EMerge. They discuss why simulation is like building a pyramid of knowledge, how silk screen can completely de-tune a patch antenna, and Robert's quest to build the “IKEA of FEM solvers” so hardware developers everywhere can simulate their designs without relying on insanely expensive proprietary software.

Known across the industry for making network engineering cool again with her highly relatable tech content, Alexis Bertholf is a powerhouse when it comes to combining technical expertise with personal branding. Eric Chou sat down with Alexis at AutoCon 5 to chat about Alexis’ career, the art of storytelling in 60 seconds or less, the... Read more »

Known across the industry for making network engineering cool again with her highly relatable tech content, Alexis Bertholf is a powerhouse when it comes to combining technical expertise with personal branding. Eric Chou sat down with Alexis at AutoCon 5 to chat about Alexis’ career, the art of storytelling in 60 seconds or less, the... Read more »

Topics covered in this episode: Python 3.12.14, 3.11.16, 3.10.21 - security releases Codeberg's AI-code ban tests its role as a GitHub alternative Brett Cannon: what's missing for reproducible builds on PyPI nothing records the source code a distribution came from. direct_url.json captures it when you install from a repo or archive, so the fix is putting the same info in sdist/wheel metadata. recording the build tools. Wheels can already do this via PEP 770 SBOMs in .dist-info/sboms/ - sdists can't, since they're a tarball plus a precalculated PKG-INFO with nowhere to hang extra metadata. Either "don't use sdists" or an sdist v2. Extra extra extra, hear all about it Extras Joke Watch on YouTube Sponsored by Logfire from Pydantic pythonbytes.fm/logfire This episode is brought to you by Pydantic Logfire. It's observability for AI apps from the team behind Pydantic - agents, LLMs, APIs, database, and infrastructure in a single trace, queried with Postgres-compatible SQL. Your coding agent can query it too, through their MCP server. I'll tell you more later. Connect with the hosts Michael: Mastodon / BlueSky / X / LinkedIn Calvin: Mastodon / BlueSky / X / LinkedIn Show: Mastodon / BlueSky / X Join us on YouTube at pythonbytes.fm/live to be part of the audience. Usually Tuesday at 7am 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. Calvin #1: Python 3.12.14, 3.11.16, 3.10.21 - security releases https://blog.python.org/2026/08/python-31214-31116-31021/ Source-only security releases for the three branches now in security-fix-only mode; release team blamed the European solar eclipse for the timing. tarfile hardening. Multiple path-traversal bypasses of the data filter closed, including a symlink escape that bypassed the CVE-2025-4330 fix; extract() now applies the filter to link targets too. Four fresh CVEs: CVE-2026-2297 (SourcelessFileLoader not using io.open_code() for .pyc), CVE-2026-4224 (expat crash on deeply nested content models), CVE-2026-3644 (control chars in http.cookies.Morsel), plus the completed CVE-2021-4189 fix in ftplib.ftpcp. Quadratic-complexity DoS cleanup across the stdlib: HTMLParser, configparser regexes, unicodedata.normalize(), csv.Sniffer.sniff(), and ElementTree XPath index predicates. Header/injection fixes: CR/LF rejected in HTTPConnection.set_tunnel(), control chars blocked in wsgiref.handlers status, and webbrowser now rejects leading dashes (plus a %action prefix bypass). http.client now caps chunked trailer lines and 1xx interim responses at 100 each - a hostile server could previously hang the client forever despite a socket timeout. Memory-safety odds and ends: stale pointers in lzma/bz2/zlib decompressors after MemoryError, a bz2 stack overflow on reuse-after-error, and bundled libexpat bumped to 2.8.3. If you're still on 3.10, 3.11, or 3.12 - and you extract tarballs from anywhere you don't fully control - this one's not optional. Michael #2: Codeberg's AI-code ban tests its role as a GitHub alternative Armin's article “Codeberg Divides” Armin Ronacher argues that Codeberg's new terms, which prohibit projects mostly written with generative AI, create a vague and difficult-to-enforce boundary. His larger concern is that a democratically governed host can still be unpredictable or ideologically narrow, weakening Codeberg's potential as a broad European alternative to GitHub. The strongest question for Python developers is whether repository hosting should judge legal open source by how code was produced, or focus on behavior and resource abuse. “Mostly generated” is hard to measure in modern codebases where developers mix handwritten code, completions, agents, and generated refactors. Ronacher suggests clearer alternatives: ban all LLM involvement, or target autonomous repository spam, abusive resource use, and low-quality generated contributions directly. Codeberg is free to choose a values-driven community, but that may conflict with being predictable, neutral infrastructure and a serious GitHub competitor. Worth discussing: can open-source communities set meaningful AI boundaries without driving maintainers and projects into opposing camps? Very first search for these terms lands on this page. Codeberg looked like a viable alternative. … Unfortunately, the latest update to its terms of service seems to mark a first step in changing one part I moved there for, namely the “freedom” part. Sponsor: Logfire from Pydantic Your AI agent failed at 2am. Was it the model? A tool call? The database? Most observability tools can't tell you, because they only see part of your stack. Pydantic Logfire sees all of it. One trace across your agents, LLMs, APIs, and database. Down to the infrastructure: services, Kubernetes, and hosts. It's built on OpenTelemetry, with SDKs for Python, TypeScript, and Rust, and it works with any OTel-compatible language. Every prompt, token count, and cost, right next to your vector searches and API calls. You query everything with Postgres-compatible SQL. And so can your coding agent, through the Logfire MCP server. Stop guessing. Read the trace. Pydantic Logfire. AI, it's still just engineering. Visit pythonbytes.fm/logfire today and sign up today. Get 10M records free every month, no card required. You can even click “Onboard with your coding agent” to copy a prompt to have claude or codex integrate Logfire into your app. Thanks to Pydantic for supporting the show. Calvin #3: Brett Cannon: what's missing for reproducible builds on PyPI Framing came out of his 2026 Python Packaging Council nomination - the secure-supply-chain gap he found is that Python has no defined way to do reproducible builds at all. Design goal is zero friction: producers uploading to PyPI shouldn't have to do anything. The work lands on build backends and installers. Gap #1: nothing records the source code a distribution came from. direct_url.json captures it when you install from a repo or archive, so the fix is putting the same info in sdist/wheel metadata. Gap #2: recording the build tools. Wheels can already do this via PEP 770 SBOMs in .dist-info/sboms/ - sdists can't, since they're a tarball plus a precalculated PKG-INFO with nowhere to hang extra metadata. Either "don't use sdists" or an sdist v2. The replay mechanism already exists: [build-system] in pyproject.toml is a defined entry point, so if backends recorded their own environment, you could reinstall and re-run the build. Payoff idea: trusted third parties report successful reproductions back to PyPI, which displays "independently reproduced by X" - surfaced in the index API so installers could prefer reproduced files. Explicitly framed as a perk, not a requirement - roughly SLSA build level 1, no shaming projects that don't opt in. Verbal kicker option: "And don't think pure-Python wheels are off the hook. Something built that wheel, and if that something was compromised, so is your wheel. SolarWinds was a build-process attack." Michael #4: Extra extra extra, hear all about it Python 3.14.7 Upgraded the MCP servers to 2026-07-28 v2 protocols (talk python, python bytes) Got agentsview running synced via postgres Talk Python courses, teams trial offering Talk Python courses, government procurement offering Lean TDD audio book is out Extras Calvin: uv now prefers post-quantum key exchange - https://github.com/astral-sh/uv/releases/tag/0.12.4 Joke: Beware of dog

When Claude Cowork hits a roadblock, it doesn't give up, it writes a custom Python script and downloads an untrusted NPM package just to bypass its restrictions and finish its goal. Are your security tools close enough to stop it? In this episode, Ashish sits down with Michael Leland, VP, Field CTO at Island, to discuss the critical need for an Agentic Control Plane. Michael breaks down why traditional security silos (EDR, DLP, CASB) fail to provide visibility when autonomous AI agents execute tasks outside the network, and why the browser is the ultimate line of defense for monitoring user intent. We explore the massive reality of Shadow AI and the hidden danger of well-intentioned employees accidentally hooking up sensitive data to public LLMs. Finally, Michael shares practical strategies for solving token waste through "model fit steering" and managing the complex "two-hop problem" when an agent calls another agent.Guest Socials -⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠Michael's Linkedin Podcast Twitter - ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠@CloudSecPod⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠If you want to watch videos of this LIVE STREAMED episode and past episodes - Check out our other Cloud Security Social Channels:-⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Cloud Security Podcast- Youtube⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠- ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Cloud Security Newsletter ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠If you are interested in AI Security, you can check out our sister podcast -⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ AI Security Podcast⁠Questions asked:(00:00) Introduction to the Agentic Control Plane(01:50) Michael Leland's Background (Cabletron, Nitro Security, SentinelOne)(03:20) Why Island Evolved from the Browser to the Desktop for AI(05:50) The Failure of Traditional Siloed Security (EDR, DLP, CASB)(07:20) Goal-Oriented AI: How Claude Cowork Downloads Untrusted NPM Packages(08:30) Model Fit Steering: Routing Users to the Right LLM for the Right Price(10:30) The Threat of Malicious AI Skills and Plugins(11:30) The Well-Intentioned Insider Threat (The Next Cambridge Analytica)(13:00) Uncovering Shadow AI: From 8 Tools to 243(15:10) Token Brokering at the MCP Gateway(16:40) Protecting Non-Human Identities (NHI)(18:20) Solving the "Two-Hop" Problem (Agent-to-Agent Communication)(21:00) Fixing Hallucinations with Corporate RAGs(25:40) Calculating AI ROI Beyond "Token Maxing"(28:40) The "You Laugh, You Lose" Cybersecurity Joke Challenge

Stupid Extra 8-18-2026 …Escaped 8ft Python starts fire at Virginia School

Stop praying from fear and start praying with faith, confidence, and spiritual authority. Discover how your identity in Christ can transform your prayers and help you partner with the Holy Spirit. Prophetic Spiritual Warfare Book  - to learn more about principalities, witchcraft and strongman spirits at https://www.kathydegrawministries.org/product/prophetic-spiritual-warfare-book/ or Amazon https://a.co/d/hhPhwxw Speak Out - Releasing the Power of Declaring Prayer available at https://www.kathydegrawministries.org/product/speak-out/ or Amazon https://a.co/d/5rkfdLi Purchase Anointing Oil with a prayer cloth that Kathy has personally mixed and prayed over on Kathy's Website or Amazon. Order anointing oil by Kathy on Amazon look for her brand here https://amzn.to/3PC6l3R or Kathy DeGraw Ministries https://www.kathydegrawministries.org/product-category/oils/ Training, Mentorship and Deliverance! Personal coaching, deliverance, e-courses, training for ministry, and mentorships! https://www.kathydegrawministries.org/training/#   Forceful prayer is not about yelling, striving, or trying to convince God to hear you. It is about praying from your identity in Christ, aligning your words with heaven, and believing that your prayers have power. In this episode, Kathy DeGraw teaches you how to pray with authority instead of fear. You will learn how to declare God's Word, command circumstances to shift, and stand confidently in your Kingdom inheritance. Whether you are believing for healing, financial breakthrough, restoration, favor, or a God-ordained promise, your prayer posture matters. Jesus demonstrated the power of spoken words, faith-filled declarations, and complete alignment with the Father. You have been given spiritual authority through Jesus Christ, but you must renew your mind and refuse to partner with doubt, defeat, sickness, poverty, or hopelessness. Partner with the Holy Spirit, speak boldly, and develop holy confidence. Your prayers do not need to be perfect or loud. They need to be rooted in faith, biblical truth, and the authority Jesus has already given you. #PrayerWithAuthority #FaithOverFear #IdentityInChrist #PowerOfPrayer #HolySpirit **Connect with Us** - Website: https://www.kathydegrawministries.org/ - Facebook: https://www.facebook.com/kathydegraw/ - Instagram: https://www.instagram.com/kathydegraw/  Podcast - Subscribe to our YouTube channel and listen to Kathy's Podcast called Prophetic Spiritual Warfare, or on Spotify at https://open.spotify.com/show/3mYPPkP28xqcTzdeoucJZu or Apple podcasts at https://podcasts.apple.com/us/podcast/prophetic-spiritual-warfare/id1474710499 **Recommended Resources:** - Receive a free prayer pdf on Python at https://www.kathydegrawministries.org/python/- Receive a free prayer pdf on Anointing Oil at https://www.kathydegrawministries.org/anointingoil/ - Kathy's training, mentoring and e-courses on Spiritual Warfare, Deliverance and the Prophetic: https://training.kathydegrawministries.org/ - Healed At Last ~ Overcome Sickness and Receive your Physical Healing: https://www.kathydegrawministries.org/healed-at-last/ - Mind Battles – Root Out Mental Triggers to Release Peace!: https://www.kathydegrawministries.org/product/mind-battles-pre-order-available-january-2023/ -Kathy has several books available on Amazon or kathydegrawministries.org **Support Kathy DeGraw Ministries:** - Give a one-time love offering or consider partnering with us for $15, $35, $75 or any amount! Every dollar helps us help others! - Website: https://www.kathydegrawministries.org/donate/ - CashApp $KathyDeGrawMinistry - Venmo @KD-Ministries - Paypal.me/KDeGrawMinistries or donate to email admin@degrawministries.org - Mail a check to: Kathy DeGraw Ministries ~ PO Box 65 ~ Grandville MI 49468

Every company has one. The little internal tool that Jane built back in 2021, and then Jane left. Nobody understands it, nobody will touch it. There are two unwritten rules around it: don't change it, it's working. And if you break it, you bought it. That's dark-matter enterprise software. For every app you can actually see, there are ten of these sitting in the shadows, frozen. Michael Booth thinks that just changed. He read my article on hyper-personal software and ran with it, writing about hyper-team software: small teams inside big companies finally building the tools that were never going to get built. We cover where this works, where it quietly goes wrong, and the guardrails that keep it from turning into a mess. Let's get into it. Episode sponsors Sentry Error Monitoring, Code talkpython26 Python in Production Talk Python Courses Links from the show Guest Michael Booth: github.com Talk Python AI Integrations: talkpython.fm/blog From Hyper-Personal to Hyper-Team Software: Small Team-Built, AI-Assisted Tools Inside the Enterprise: www.databooth.com.au What hyper-personal software looks like (MK's article): mkennedy.codes Databooth Site: www.databooth.com.au Wall Street just lost $285 billion because of 13 markdown files: martinalderson.com SaaSpocalypse is real but everyone is panicking about the wrong thing: www.reddit.com Warp Terminal: www.warp.dev Watch this episode on YouTube: youtube.com Episode #558 deep-dive: talkpython.fm/558 Episode transcripts: talkpython.fm Theme Song: Developer Rap