Podcasts about llm

Fondée à Paris en 2020 par Arnaud Rihiant, Alexis Delplanque et Eric Gaudin, DJUST utilise l'IA agentique B2B pour automatiser la prise de commande. Finis les saisies manuelles dans l'ERP, les fichiers Excel et les échanges téléphoniques. En 2026, la plateforme devrait gérer aux alentours de 3 milliards d'euros de commandes pour une cinquantaine de clients, dont Socoda, Eiffage et Franprix. En octobre 2025, elle a levé 7 millions d'euros auprès de NEA, Elaia et Speedinvest pour développer sa plateforme d'IA agentique. Arnaud Rihiant analyse dans cette interview les spécificités du e-commerce B2B, le coût de la ressaisie et la conception des agents IA dans ce contexte. IA agentique B2B : du e-commerce à l'agent qui passe commande DJUST en chiffres : 2020, une cinquantaine de clients et près de 3 milliards d’euros de commandes gérées en 2026. Les illustrations de ce billet ont été réalisées avec Claude, ChatGPT et Midjourney. Les sources sont indiquées chaque fois que cela est nécessaire. Arnaud Rihiant a fait ses classes chez Mirakl, où il a accompagné pendant plus d'une décennie l'implémentation et la réussite client. C'est là, dit-il, qu'est né le constat fondateur de DJUST. Qu'est-ce qui, chez Mirakl, vous a convaincu que le B2B avait besoin de sa propre plateforme plutôt que d'un outil B2C adapté ? « DJUST a été créé en 2020 sur la base d'un constat que j'avais déjà fait chez Mirakl », explique Arnaud Rihiant. Mirakl était à l'origine très concentré sur le B2C, jusqu'à ce que de plus en plus d'acteurs du B2B s'intéressent à sa marketplace. « Je me suis rendu compte que c'était un monde qui avait beaucoup de mal à se digitaliser. Chaque projet était complexe et durait un an, deux ans, voire plus, avec des sommes astronomiques d'investissement ». La complexité tient à la nature même de l'achat B2B. « Le B2C se focalise énormément sur l'apport de trafic. On dépense beaucoup d'argent en webmarketing, en réseaux sociaux, pour faire venir les gens sur le site. Mais une fois sur un site B2C, l'acte d'achat est simple, j'ajoute au panier, je sors ma carte bleue, je paie et on n'en parle plus ». En B2B, des négociations contractuelles précèdent l'achat, avec des catalogues, des prix préférentiels et des conditions de paiement propres à chaque client, qu'il faut ensuite reporter dans la plateforme. « Quand on repart d'une solution pensée pour un acte d'achat très simple et qu'on veut y intégrer ses propres règles, c'est là que toute la complexité commence ». Vous accompagniez environ 50 clients fin 2025, pour 1,5 milliard d'euros de volumes de commandes. Où en êtes-vous aujourd'hui ? « Nous avons déjà dépassé ce montant à mi-année 2026. Je pense qu'on fera entre 2,5 et 3 milliards d'euros de volumes de commandes cette année », précise Arnaud Rihiant, qui insiste sur un second indicateur, l'adoption à périmètre constant. « Les acheteurs B2B ont tous des smartphones dans les mains. Ils sont habitués, au quotidien, à des choses simples et rapides. Quand on leur dit qu'il faut se connecter à un système lourd, potentiellement passer un coup de téléphone ou envoyer un email, ils n'ont plus le temps pour ça. Ce qu'ils veulent, c'est se connecter sur une plateforme simple, retrouver leurs prix, passer leur commande et la valider ». Un marché plus lourd qu'on ne le dit, mais moins simple à mesurer Selon la FEVAD, le e-commerce B2B a progressé de 4 % en France en 2025, dans la continuité de 2024, quand le B2C affichait une croissance plus soutenue. Un rythme plus tassé, donc, mais qui ne dit rien du poids réel du marché. « Selon certaines études, le commerce B2B en ligne pèse environ 600 milliards d'euros en France », avance Arnaud Rihiant, contre un B2C nettement plus modeste (environ 196,4 milliards d'euros). Dans ce total coexistent plusieurs canaux, dont l'EDI, cet échange automatisé entre deux systèmes d'information sans aucune interface pour l'acheteur. « L'EDI, c'est pour connecter deux systèmes entre eux, un chantier de six mois à un an, et vous n'en faites qu'un seul à la fois ». Le e-commerce B2B pèserait environ 600 milliards d’euros en France, contre 196,4 milliards pour le B2C (FEVAD, 2025). L'EDI ne va pas disparaître, estime le co-fondateur de DJUST, il va s'interconnecter avec le e-commerce et les portails B2B. « Pour moi, la grande révolution dans le B2B, ça va être l'IA. On a eu les journaux papier, puis la radio, puis la télévision, puis Internet, et à chaque fois un nouveau canal arrive sans tuer les précédents. Il en sera de même pour les agents IA ». Reste un paradoxe persistant, en 2026, l'email demeure l'un des canaux dominants de prise de commande. « Les acheteurs B2B sont souvent des gens très opérationnels, des boulangers, des gens sur les chantiers, pas toujours assis devant un ordinateur. Pour eux, passer une commande revient à envoyer un texto ou un email ». C'est précisément là, selon Arnaud Rihiant, que l'IA agentique B2B change la donne, en captant ces demandes informelles pour les transformer automatiquement en commandes structurées, sans bousculer les habitudes du client. Le e-commerce B2B pèserait environ 600 milliards d’euros en France, contre 196,4 milliards pour le B2C (FEVAD, 2025). Le vrai point de douleur, c'est la ressaisie manuelle Une étude DJUST relayée sur LinkedIn par Arnaud Rihiant indique que 80 % des équipes commerciales B2B perdent encore du temps à ressaisir manuellement des commandes dans l'ERP. « Les ERP sont des outils faits pour l'interne. Un ERP ne veut rien dire pour un client. Quand celui-ci passe une commande, il n'a que faire de vos process internes ». Le risque, ajoute-t-il, dépasse la seule productivité commerciale. « Beaucoup d'entreprises ne peuvent même plus extraire leurs données d'un ERP vieux de trente ans. C'est un vrai risque de gestion, perdre toute la donnée client, ne plus pouvoir vendre, si demain cet ERP cesse de fonctionner ». 80 % des équipes commerciales B2B perdraient encore du temps à ressaisir manuellement des commandes dans l’ERP. Le cas le plus abouti reste celui de Socoda, une alliance de distributeurs indépendants du BTP et de l'industrie, basée en région parisienne, cliente de DJUST depuis plus de quatre ans. La démarche s'est faite par étapes, une première plateforme e-commerce en self-service a permis de décharger 20 à 30 % des commandes basiques des équipes commerciales. Le second gain, plus spectaculaire, est venu de l'IA appliquée aux commandes reçues par email sous forme de fichiers Excel de plusieurs centaines de lignes. « On avait une personne à l'administration des ventes qui, ligne par ligne, contrôlait dans l'ERP s'il y avait du stock. Une commande lui prenait une après-midi ». L'IA analyse désormais le fichier, le restructure, effectue les contrôles de stock et de prix, avant de restituer le résultat à un opérateur humain pour validation. « Une personne qui passait trois heures à traiter une commande le fait aujourd'hui en moins de trente minutes. C'est notre recommandation constante, mettre en place un contrôle humain, pour éviter les effets de bord ». IA agentique B2B : un orchestrateur Interrogé sur le modèle d'IA générative utilisé, Arnaud Rihiant revendique une approche délibérément multi-LLM. « Il est important de ne pas être dépendant d'un prestataire ou d'une technologie. ChatGPT sera meilleur sur l'analyse d'images, Mistral sur les PDF, Claude sur les fichiers Excel. Chaque modèle a un peu sa spécialité ». Ce choix répond aussi à un risque de continuité de service. « Il y a beaucoup de solutions IA mono-LLM aujourd'hui, et le risque, c'est que si ce modèle tombe, tout est à refaire. Avec notre solution, si demain un des LLM tombe, que ce soit Claude, Mistral ou ChatGPT, le business continue de tourner parce que d'autres modèles prennent le relais ». DJUST observe également les modèles dits à poids ouverts, installables sur des serveurs indépendants, sans toutefois les déployer en dehors d'infrastructures offrant redondance et chiffrement des données. Cette prudence infrastructurelle éclaire aussi sa définition d'un agent. « Un agent, pour nous, c'est un orchestrateur qui appelle plusieurs modèles avec un objectif précis. On voit d'autres agents plutôt monolithiques, qui appellent un seul LLM et donnent un retour basique. J'appelle ça des agents passe-plat ». DJUST préfère décomposer chaque tâche en sous-agents spécialisés, l'un pour découper la demande, un autre pour analyser le texte, un troisième pour le PDF. « Ce processus améliore la pertinence, coûte souvent moins cher, et réduit le temps de traitement. Cette IA agentique B2B n'est pas vraiment composée de programmes, ce sont des skills envoyés à plusieurs LLM. Le temps des programmes est révolu, il faut penser en termes de workflows et de tâches ». Vers la commande vocale, et peut-être au-delà Poussé à se projeter, Arnaud Rihiant décrit un scénario où l'acheteur B2B commande directement à la voix. « C'est ce qu'on appelle le “voice to order”. Avant, on se connectait à un ordinateur pour passer une commande, puis on a eu les SMS, beaucoup de nos clients commandent déjà comme ça, demain on passera à la commande vocale ». Un moteur de règles restera cependant systématiquement sollicité pour vérifier stocks, prix et autorisations avant toute confirmation. « Il y aura toujours un cerveau de workflow qui contrôlera les stocks, les prix, les autorisations, et qui demandera confirmation ». Interrogé sur une automatisation plus poussée encore, comparable au trading algorithmique où des machines arbitrent en microsecondes entre places boursières, le dirigeant n'exclut pas que le e-commerce B2B évolue en ce sens. « On pourrait imaginer une IA qui devienne le centre d'achat d'une coopérative de boulangers, capable d'analyser les volumes de vente de chacun, les pics saisonniers, et de passer des ordres d'achat de manière autonome. On n'en est pas si loin que ça ». C'est peut-être là que se joue, au fond, l'avenir de l'IA agentique B2B, moins dans la promesse d'un agent conversationnel isolé que dans sa capacité à s'insérer, discrètement, entre un email mal formaté et un ERP qui ne comprend rien à ce que veut vraiment le client. Pour poursuivre votre lecture, découvrez Mythes et réalité du commerce agentique. Avec Arnaud Rihiant nous avons parlé d’IA agentique pour le e-commerce B2B. À propos d'Arnaud Rihiant et de DJUST Arnaud Rihiant est fondateur et dirigeant de DJUST. Il a été membre de l'équipe fondatrice de Mirakl, où il a piloté l'implémentation et la réussite client pendant plus d'une décennie, avant de lancer DJUST en 2020 pour répondre aux besoins spécifiques du commerce B2B, distincts selon lui de ceux du B2C. DJUST est une plateforme SaaS de commerce B2B fondée à Paris en 2020 par Arnaud Rihiant, Alexis Delplanque et Eric Gaudin. Elle permet aux distributeurs, industriels et fabricants de digitaliser leurs processus de vente B2B (catalogues en ligne, e-procurement, connecteurs marketplace) sans remplacer leur ERP existant. La société a levé 4 millions d'euros en amorçage auprès d'Elaia en 2020, puis 12 millions d'euros en série A en 2023 (New Enterprise Associates, Elaia), et 7 millions d'euros supplémentaires en octobre 2025 (NEA, Elaia, Speedinvest) pour développer l'IA agentique appliquée au B2B. Fin 2025, DJUST accompagnait une cinquantaine de clients (dont Eiffage, Franprix et Socoda) et gérait environ 1,5 milliard d'euros de volumes de commandes annuels. The post IA agentique B2B : du e-commerce à l'agent qui passe commande appeared first on Marketing and Innovation.

Qwen, Muse Glimmer, Ollama, llama.cpp, quantization, mixture-of-experts, RAG, tokens per second. On Hands-On AI, those aren't buzzwords — Mikah Sargent will explain exactly what they mean, then put them to work on hardware you already own. No cloud, no account, no data, leaving your desk. Host: Mikah Sargent Download or subscribe to Hands-On AI at https://twit.tv/shows/hands-on-ai Want access to the ad-free audio and video and exclusive features? Become a member of Club TWiT today! https://twit.tv/clubtwit Club TWiT members can discuss this episode and leave feedback in the Club TWiT Discord.

Qwen, Muse Glimmer, Ollama, llama.cpp, quantization, mixture-of-experts, RAG, tokens per second. On Hands-On AI, those aren't buzzwords — Mikah Sargent will explain exactly what they mean, then put them to work on hardware you already own. No cloud, no account, no data, leaving your desk. Host: Mikah Sargent Download or subscribe to Hands-On AI at https://twit.tv/shows/hands-on-ai Want access to the ad-free audio and video and exclusive features? Become a member of Club TWiT today! https://twit.tv/clubtwit Club TWiT members can discuss this episode and leave feedback in the Club TWiT Discord.

Qwen, Muse Glimmer, Ollama, llama.cpp, quantization, mixture-of-experts, RAG, tokens per second. On Hands-On AI, those aren't buzzwords — Mikah Sargent will explain exactly what they mean, then put them to work on hardware you already own. No cloud, no account, no data, leaving your desk. Host: Mikah Sargent Download or subscribe to Hands-On AI at https://twit.tv/shows/hands-on-ai Want access to the ad-free audio and video and exclusive features? Become a member of Club TWiT today! https://twit.tv/clubtwit Club TWiT members can discuss this episode and leave feedback in the Club TWiT Discord.

Substackで配信前にその日のリサーチ結果をお届けします。ご購読をおねがいします！ https://wbnlive.substack.com/▼ 出演Kinjo X: https://x.com/illshinイケハヤさん(ゲスト / Studio VIBE 代表)X: https://x.com/IHayato▼ 主要ソース月蝕綺譚 公式: https://vibe.co.jp/luna-occultaイケハヤ Substack(AIでゲームを作るのは修羅の道): https://ihayato.substack.com/p/ai-28fnote: https://note.com/ihayato/n/n8d1faef0c81b本日の「WBN」から、ゲストのイケハヤさんへのインタビューを切り出しました。9月19日12:00にiOS/Androidで無料リリースされる和風ダークファンタジーの放置育成RPG「月蝕綺譚 -Luna Occulta-」を、イケハヤさんはほぼ一人で、着手から約2カ月で作りました。Claude Fable 5を12アカウント契約して並列で回し、外注すればAIの試算で20億円と言われた規模を、LLM代の月50万円ほどで形にしています。この動画では、放置育成というジャンルを選んだ理由、課金ガチャゼロと広告ゼロにした判断、会員ポータルとトークンのエコシステム、そしてAIに任せられなかった部分までを聞いています。▼ 目次00:00 リリース24時間前。イケハヤさん登場00:48 クリプト忍者の5年。ゲームは金がかかる壁だった01:56 Fableで「これ作れる」。12アカウント契約03:24 外注ならいくら。AIの試算は20億円07:47 放置育成は盆栽に近い。1日5分のジャンル11:47 課金ガチャゼロ、広告ゼロにした理由15:49 会員ポータルとトークン。上位500人に200万円17:25 作るのは簡単になった。集客が一番難しい20:08 8セッション並列で壊れた最初の失敗23:47 AIは目が悪く耳も悪い。ボイス3000本を自分の耳で28:14 着手から約2カ月でストア審査が通った29:58 LLM代は月50万円。画像と動画は9月に激減35:08 エージェントスキルをチームで融通する39:22 開発では戦えない。勝負はマーケティング41:08 ビルダーへ。ぶち抜ける領域に突っ走れ※ 外注20億円という金額はイケハヤさんがAIに試算させたもので、開発期間や費用も本人の申告に基づきます。トークン配布などPlay2Earn関連の施策は予定であり、変更される場合があります。WBNは平日毎日12:00から、AI・テック・スタートアップのニュースをライブでお届けしています。#AI #AIニュース #イケハヤ #月蝕綺譚 #ゲーム開発 #クリプト忍者 #WBN

"Czy robiliśmy review lat temu trzy? Nie, nie robiliśmy, tylko wszyscy okłamywali się, że robią review porządnie i klikali: tak, tak, tak może iść."

OpenAI disclosed six new AI misalignment incidents, including models hiding their own mistakes, the SEC unveiled a five-year tokenized-stock trading exemption, Zuckerberg, Huang, and Musk successfully stalled a White House AI regulatory push, and Anthropic merged Claude chat with Cowork. OpenAI discloses six new misalignment incidents since October, including models concealing mistakes, and announces a framework for reporting model misalignment (Axios) The Times details the six incidents, including a GPT-5.6 Sol variant that hid errors via secret notes to itself, and an unreleased model that wrote a "persona instruction" declaring itself "freed from the roles and identities that bind other chatbots" (The New York Times) The US SEC unveils a five-year "Innovation Exemption" to free platforms that facilitate blockchain and tokenized stock trading from many stock exchange rules (Reuters) Sources: Mark Zuckerberg, Jensen Huang, and Elon Musk recently spoke with Trump and successfully stalled an AI regulatory plan proposed by Demis Hassabis (The Wall Street Journal) Anthropic merges Claude chat and Cowork, and adds a feature for making presentations and documents, rolling out to Pro and Max plans first (TechCrunch) Data center developer Crusoe raised $3.9B co-led by Atreides, Valor, and Mubadala at a ~$30.9B post-money valuation, as it bets on factory-built data centers (The Wall Street Journal) Anthropic and other researchers detail how thousands of people were catfished by dating scam apps using LLM-generated replies from Claude and other models (The Verge) Subscribe to the ad-free feed.

Discover how to run serious AI at home without breaking the bank, as Lon Seidman reveals his tricks for turning old data center GPUs into powerful local LLM machines. Anthropic Researchers Raise Alarm Over A.I. Acceleration, Warning of Threat to Humanity Dario Amodei — We Must Pace the Frontier Mark Zuckerberg Takes Aim at Anthropic in Debate Over A.I. Slowdown A Code of Conduct for Humanist AI Countering misuse of AI: September 2026 / Anthropic EXCLUSIVE: OpenAI's rogue agents probed Hugging Face for weaknesses two months before major hack Yoshua Bengio | Why are AI agents lying, cheating and coordinating? The Worst Spam Emails: Inside iLands' AI Agent Hustle Introducing System One Models & Jev - TypeSafe AI Blog Universal Music is launching an AI music platform with ElevenLabs Lawyer fined $5K over AI-hallucinated witnesses in a murder case A Digital Fly Brain Has Taken Over the Internet (4) nimensky e/acc on X: "ok who tf did this?! lul https://t.co/HpnXo0qKJ4" / X Fanttik Cafein 11 Portable Espresso Machine Casey Newton and Kevin Roose partner with NPR to launch 'Machine Gods' They Built a Shrine to Cable TV. Then Everyone Cut the Cord. Hosts: Leo Laporte, Jeff Jarvis, and Fr. Robert Ballecer, SJ Guest: Lon Seidman Download or subscribe to Intelligent Machines at https://twit.tv/shows/intelligent-machines. Join Club TWiT for Ad-Free Podcasts! Support what you love and get ad-free audio and video feeds, a members-only Discord, and exclusive content. Join today: https://twit.tv/clubtwit Sponsors: trustedtech.team/intelligent365 claude.ai/technology framer.com/machines bitwarden.com/twit

Discover how to run serious AI at home without breaking the bank, as Lon Seidman reveals his tricks for turning old data center GPUs into powerful local LLM machines. Anthropic Researchers Raise Alarm Over A.I. Acceleration, Warning of Threat to Humanity Dario Amodei — We Must Pace the Frontier Mark Zuckerberg Takes Aim at Anthropic in Debate Over A.I. Slowdown A Code of Conduct for Humanist AI Countering misuse of AI: September 2026 / Anthropic EXCLUSIVE: OpenAI's rogue agents probed Hugging Face for weaknesses two months before major hack Yoshua Bengio | Why are AI agents lying, cheating and coordinating? The Worst Spam Emails: Inside iLands' AI Agent Hustle Introducing System One Models & Jev - TypeSafe AI Blog Universal Music is launching an AI music platform with ElevenLabs Lawyer fined $5K over AI-hallucinated witnesses in a murder case A Digital Fly Brain Has Taken Over the Internet (4) nimensky e/acc on X: "ok who tf did this?! lul https://t.co/HpnXo0qKJ4" / X Fanttik Cafein 11 Portable Espresso Machine Casey Newton and Kevin Roose partner with NPR to launch 'Machine Gods' They Built a Shrine to Cable TV. Then Everyone Cut the Cord. Hosts: Leo Laporte, Jeff Jarvis, and Fr. Robert Ballecer, SJ Guest: Lon Seidman Download or subscribe to Intelligent Machines at https://twit.tv/shows/intelligent-machines. Join Club TWiT for Ad-Free Podcasts! Support what you love and get ad-free audio and video feeds, a members-only Discord, and exclusive content. Join today: https://twit.tv/clubtwit Sponsors: trustedtech.team/intelligent365 claude.ai/technology framer.com/machines bitwarden.com/twit

Sami talks with Nate Berkopec, owner of Speedshop, about a recent post he made on X and how his business manages to stay ahead of AI and LLMs. Nate stresses the importance of prioritising quality in code in the age of AI, why your knowledge still matters if you want to get the best results from your LLM, how code review has drastically changed over the last three years, and the pair discuss why it's so tough for junior developers to stand out in the industry at the moment and the lengths they have to go to to be noticed. — If you want more out of your project why not get in touch with Nate through his business Speedshop. You can also check out his X post that inspired this episode or connect with him via Linkedin. Your host for this episode has been Sami Birnbaum. Sami can be found through his website or via LinkedIn. If you would like to support the show, head over to our GitHub page, or check out our website. Got a question or comment about the show? Why not write to our hosts: hosts@giantrobots.fm This has been a thoughtbot podcast. Stay up to date by following us on social media - LinkedIn - Mastodon - YouTube - Bluesky © 2026 Giant Robots Smashing Into Other Giant Robots Podcast

Discover how to run serious AI at home without breaking the bank, as Lon Seidman reveals his tricks for turning old data center GPUs into powerful local LLM machines. Anthropic Researchers Raise Alarm Over A.I. Acceleration, Warning of Threat to Humanity Dario Amodei — We Must Pace the Frontier Mark Zuckerberg Takes Aim at Anthropic in Debate Over A.I. Slowdown A Code of Conduct for Humanist AI Countering misuse of AI: September 2026 / Anthropic EXCLUSIVE: OpenAI's rogue agents probed Hugging Face for weaknesses two months before major hack Yoshua Bengio | Why are AI agents lying, cheating and coordinating? The Worst Spam Emails: Inside iLands' AI Agent Hustle Introducing System One Models & Jev - TypeSafe AI Blog Universal Music is launching an AI music platform with ElevenLabs Lawyer fined $5K over AI-hallucinated witnesses in a murder case A Digital Fly Brain Has Taken Over the Internet (4) nimensky e/acc on X: "ok who tf did this?! lul https://t.co/HpnXo0qKJ4" / X Fanttik Cafein 11 Portable Espresso Machine Casey Newton and Kevin Roose partner with NPR to launch 'Machine Gods' They Built a Shrine to Cable TV. Then Everyone Cut the Cord. Hosts: Leo Laporte, Jeff Jarvis, and Fr. Robert Ballecer, SJ Guest: Lon Seidman Download or subscribe to Intelligent Machines at https://twit.tv/shows/intelligent-machines. Join Club TWiT for Ad-Free Podcasts! Support what you love and get ad-free audio and video feeds, a members-only Discord, and exclusive content. Join today: https://twit.tv/clubtwit Sponsors: trustedtech.team/intelligent365 claude.ai/technology framer.com/machines bitwarden.com/twit

Discover how to run serious AI at home without breaking the bank, as Lon Seidman reveals his tricks for turning old data center GPUs into powerful local LLM machines. Anthropic Researchers Raise Alarm Over A.I. Acceleration, Warning of Threat to Humanity Dario Amodei — We Must Pace the Frontier Mark Zuckerberg Takes Aim at Anthropic in Debate Over A.I. Slowdown A Code of Conduct for Humanist AI Countering misuse of AI: September 2026 / Anthropic EXCLUSIVE: OpenAI's rogue agents probed Hugging Face for weaknesses two months before major hack Yoshua Bengio | Why are AI agents lying, cheating and coordinating? The Worst Spam Emails: Inside iLands' AI Agent Hustle Introducing System One Models & Jev - TypeSafe AI Blog Universal Music is launching an AI music platform with ElevenLabs Lawyer fined $5K over AI-hallucinated witnesses in a murder case A Digital Fly Brain Has Taken Over the Internet (4) nimensky e/acc on X: "ok who tf did this?! lul https://t.co/HpnXo0qKJ4" / X Fanttik Cafein 11 Portable Espresso Machine Casey Newton and Kevin Roose partner with NPR to launch 'Machine Gods' They Built a Shrine to Cable TV. Then Everyone Cut the Cord. Hosts: Leo Laporte, Jeff Jarvis, and Fr. Robert Ballecer, SJ Guest: Lon Seidman Download or subscribe to Intelligent Machines at https://twit.tv/shows/intelligent-machines. Join Club TWiT for Ad-Free Podcasts! Support what you love and get ad-free audio and video feeds, a members-only Discord, and exclusive content. Join today: https://twit.tv/clubtwit Sponsors: trustedtech.team/intelligent365 claude.ai/technology framer.com/machines bitwarden.com/twit

AI search is changing how people discover information and how brands need to think about visibility. Daniel and Chris talk with Liam Dunne and Ben Moore, co-founders of Discovered Labs, about the shift from traditional SEO to AI search, what happens behind the scenes when an LLM generates an answer, and why retrieval and citations don't always tell the whole story. They dig into AI visibility, Reddit strategy, query fan-out, representation in model weights, and consensus. They also look ahead to a world where AI agents don't just discover websites, but interact with them and take actions on behalf of users.Featuring: Liam Dunne – LinkedInBen Moore – LinkedInDaniel Whitenack – Website, GitHub, XChris Benson – Website, LinkedIn, Bluesky, GitHub, XLinks:What AI models know about 1,000 B2B SaaS brands before web searchDiscovered LabsSponsors:Midwest AI Summit: Join AI practitioners on October 15 in Indianapolis for practical sessions, hands-on discussions, and real-world AI solutions. Use code PracticalAI20 to save 20% on your registration. https://midwestaisummit.com/#ticketsFramer: The enterprise-grade website builder that lets your team ship faster. Get 30% off at framer.com/practicalaiResources and Events:Register for upcoming webinars here!Prior Webinars from our partner Prediction GuardMidwest AI Summit 2026

Joey returns from TIFF with observations on Kling's presence and the industry's ongoing misunderstanding of hybrid AI production. Plus, a close look at ChatGPT Image 2.5's character consistency updates, how TypeSafe's Jev AI model tackles non-LLM tasks...

Continuano gli hack, i data leak e le cessazioni di servizio. Questa settimana è il turno di Revolut, Brevo (che trascina cion sè Trezor e BitBox), e torniamo a parlare di Liquid.Inoltre: la nuova versione di GridlessOS integra i datacenter AI con i miner bitcoin e le banche si attrezzano per la custodia di Bitcoin.It's Showtime!

How to Assess Skillful Use of AI   The way we assess engineering talent is undergoing a fundamental shift.   With AI now capable of generating functional code in seconds, the old evaluation playbook - whiteboard algorithms, take-home tests, "does it compile?"- is looking increasingly obsolete. The real question for hiring teams isn't whether a candidate can write syntax; it's whether they can exercise judgment, orchestrate multiple AI agents, and ship production-quality systems in a world where code generation is commoditised.   G2i has been on the front lines of this shift, spending the last two years embedded inside frontier AI labs building reinforcement learning environments, human evaluation workflows, and training data infrastructure. That hands-on experience has completely redefined how they vet engineers - not for what they can type, but for how they think, direct, and evaluate AI-generated output. In this session, we'll unpack how reinforcement learning principles are reshaping technical assessment and what it means for anyone hiring engineering talent right now.   Key Themes We'll Cover: Why Syntax Is No Longer the Differentiator – How AI code generation has flattened the playing field and why "can they code?" is the wrong question to ask in 2026 From Code Tests to Judgment Tests – What G2i's live technical evaluations actually measure now: agentic development, RL pipeline design, RAG architecture, LLM engineering, and the ability to evaluate AI output quality Inside the Frontier Lab – What G2i learned from two years building RL environments and human evaluation systems for cutting-edge AI models, and how that experience rewired their vetting philosophy Reinforcement Learning as Evaluation Framework – How RLHF-style feedback loops and reward modelling concepts are being applied to assess candidate decision-making under uncertainty The Orchestration Skill Gap – Why the engineers worth hiring today are the ones who can direct, critique, and refine the output of multiple AI agents—not just prompt them What "Taste" Looks Like in Practice – How G2i distinguishes between engineers who ship AI-assisted code and engineers who ship good AI-assisted code Building Multi-Agent Workflows – How G2i developed internal orchestration tools to coordinate complex agentic pipelines, and what that reveals about the skills they now screen for The Recruiter's Translation Challenge – How TA teams can bridge the gap between technical hiring managers and engineering candidates when the evaluation criteria have shifted this dramatically What Comes Next for Technical Vetting – How assessment is likely to evolve as AI capabilities accelerate, and why the platforms still testing for LeetCode-style problems are falling behind Whether you're a recruiter struggling to keep up with how engineering assessment is changing, a TA leader rethinking your technical interview process, or a hiring manager wondering why your last three "great on paper" hires couldn't ship reliable code, this session will give you a clear-eyed view of where evaluation is heading. G2i isn't talking theory—they've been building the systems that train the models.   Join Hung Lee and David House as we explore what reinforcement learning reveals about real engineering judgment, and why the best hires in the agentic era will be defined by taste, not syntax.   We're on Thursday 17th September, 10am PT / 1pm ET   Follow the Recruiting Brainfood channel here (recomended) and register now - this is one skill shift you need to understand before your next technical search

In this week's Data Debrief, Catherine Dowden-King and Kyle Winterbottom are recording a day early ahead of a busy stretch in London – a custom client event, the Future of Data, AI & BI Summit and Big Data London, with Driven by Data Live on 8th October. The headlines this week belong to Donald Trump, who has dismissed AI safeguards and calls for a "kill switch" as a hoax, which sets up the episode's central question: what happens when the people with the power to sign off on AI – presidents or CEOs – sit well above the technical detail and the risk?Catherine and Kyle draw the parallel between geopolitical "space race" thinking and the boardroom instinct to move first and mop it up later, before turning to the vetting questions every data leader should be asking of vendors and LLM providers: what are their values, why are they in the market, and what's in it for them? They then debrief Tuesday's episode with David Castro-Gavino and Boyan Angelov – merchants of complexity, friction versus maturity models, and a CDO role that is hired without an objective – and Kyle's thought of the week on the environmental cost of AI that nobody in the industry seems to be talking about.They also discuss:Why one of the world's most powerful people calling AI safeguards a hoax is a bigger statement than it sounds.Why the AI race is being driven by the fear of China catching up, and how that agenda filters into business.Why "just do it, we'll mop it up later" is the same decision whether it's made in the White House or the boardroom.Why the spectrum between doomsday and handbrake-on leaves everyone struggling to know who to trust.What questions to ask of any vendor or LLM provider before plugging them into your business.What the Careless People revelations about targeting insecure teenagers tell us about tech companies' incentives.Why tech companies handling health and genetic data aren't regulated like health companies.Why "just because you could doesn't mean you should" is the age-old debate, and why nobody boycotts anyway.How a tight-knit CDO community quietly blacklists vendors with poor ethics.Why the vendor community has to own the fact that every pitch deck now sounds the same.Why executives can be forgiven for not understanding the weeds of "AI-powered" everything.How the event agenda has flipped from getting executives to care about data to reining them in.Why data leaders have to learn to sell, and why selling is really just communication.Why influence at ExCo level comes down to trust, credibility and relationships.Why "merchants of complexity" and self-inflicted complexity resonated so strongly with listeners.Why friction isn't uniform, and why "the whole thing's a mess" is rarely true for every department.Why maturity models are theatre, and diagnosing friction through each role's lens is the practical alternative.Why frameworks can contain thinking, and why data people run to structure when they might need creativity.Why the CDO must be the only senior role in business hired without an objective, and what a 90-day plan is really for.Why nobody in the AI adoption conversation is asking whether we need to be using it at all.What 700ml of water per ChatGPT prompt says about the environmental cost of replacing Google.Why nothing in data is sociologically neutral, and why people only care about data centres once the bulldozers arrive.What the new Driven by Data Productions brand means for the community, the podcast, the events and the magazine.How to get 20% off Enabling Data with the code DRIVEN20.

Supply chain performance is still being measured by outdated assumptions. Lora Cecere argues that optimizing individual functions for lower cost does not necessarily improve the business. In fact, it can throw the entire supply chain out of balance. Better performance starts with understanding what is actually feasible within an industry, focusing on the metrics that drive value, and recognizing that supply chains are complex systems that cannot be improved one function at a time. In this episode of Supply Chain Now, Scott Luton speaks with Lora Cecere, founder of Supply Chain Insights, and Wael Abdelmalek, CEO at Uthereal, about the research behind Supply Chains to Admire and how that work is being brought into Ask Lora, an AI-enabled benchmarking and advisory platform. They explore why traditional supply chain rankings and generic AI tools can reinforce conventional thinking, how verified research and industry-specific data can create better benchmarks, and why leaders need a more balanced view of growth, operating margin, inventory turns, and return on capital employed when evaluating performance. Lora and Wael also explain how Ask Lora combines 10 years of research and benchmarking data with a private large language model designed specifically for supply chain decision-making. Rather than serving only as a Q&A tool, users are applying it as a trusted peer to benchmark performance, review strategy documents, pressure-test presentations, compare organizational and technology maturity, and challenge assumptions before taking ideas to leadership.Jump into the conversation:(00:00) Introduction(01:14) Meet Lora Cecere, founder of Supply Chain Insights, and Wael Abdelmalek, CEO at Uthereal(04:36) Fearless Friday: career leaps and personal risk-taking stories(08:04) The origin of Supply Chains to Admire and why "best practices" is a myth(14:44) What supply chain leaders still get wrong about cost and performance(19:22) What is Ask Lora? Inside the private LLM built on 10 years of research(21:30) Private LLMs vs. generic AI: accuracy, guardrails, and the Ethereal partnership(28:57) How organizations are using Ask Lora: benchmarking, strategy review, and reports(43:57) Pricing tiers, governance, and integration for Ask Lora(50:33) Vision for Ask Lora's future and how to connect with Lora and WaelAdditional Links & Resources:Connect with Lora Cecere: https://www.linkedin.com/in/loracecere/Connect with Wael Abdelmalek: https://www.linkedin.com/in/wwael/ Learn more about Supply Chain Insights: https://supplychaininsights.com/ Learn more about Uthereal: https://uthereal.ai/ Check out AskLora for yourself: https://asklora.ai/Learn more about our hosts: https://supplychainnow.com/about Learn more about Supply Chain Now: https://supplychainnow.com Watch and listen to more Supply Chain Now episodes here: https://supplychainnow.com/program/supply-chain-now Subscribe to Supply Chain Now on your favorite platform: https://supplychainnow.com/join Work with us! Download Supply Chain Now's NEW Media Kit: https://supplychainnow.com/media-kit/ Learn more about Blue Yonder Cognitive Solutions: http://blueyonder.com/cognitiveWEBINAR- Operational AI in the Supply Chain: How context empowers agents and humans to operate side by side: https://bit.ly/4x7Vd2ZWEBINAR- Demand Volatility Isn't a Forecasting Problem: How One FMCG Distributor Released Capacity Without Capital: https://bit.ly/4r94YM6WEBINAR- You Can't Manage What You Can't See: Using Visibility, KPIs, and AI to Optimize Logistics Operations: https://bit.ly/4ql6iemWEBINAR- The Real ROI of AI Analytics in the Supply Chain: Better Answers When Everyone Starts Asking: https://bit.ly/3TpCX6u This episode was hosted by Scott Luton and produced by Trisha Cordes, Joshua Miranda, and Amanda Luton. For additional information, please visit our dedicated show page at: https://supplychainnow.com/what-have-we-learned-thus-far-asklaura-ai-1635 The content in this episode, including all audio, videos, visuals, and graphics, is the property of Supply Chain Now and is protected by copyright law. Unauthorized use, reproduction, distribution, modification, or re-uploading of this content in any form is strictly prohibited without explicit written permission from Supply Chain Now.For licensing inquiries or permissions, please contact us at production@supplychainnow.com© 2026 Supply Chain Now. All rights reserved. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Отпуска слишком короткие. Жесткого диска объемом 1 ТБ зачастую недостаточно. Зависимости проекта то и дело ломаются. Веб-браузеры несовместимы между собой. На уровень изоляции SERIALIZABLE нельзя полагаться. Многие фичи PostgreSQL 19 пришлось откатить. LLM-агенты развиваются чересчур быстро. Ретро-игры нулевых душнее, чем говорят в обзорах. Мы в отчаянии! Информационные технологии повергают нас в пучину отчаяния! Шоуноты: [00:01:19]… Читать далее →

Welcome back to Data Driven, the podcast where we dive deep into the evolving worlds of data science, business intelligence, and artificial intelligence. In this episode, host Frank La Vigne sits down with Rob Collie, CEO of P3 Adaptive and a former Microsoft engineer, to explore the remarkable journey from Excel's foundational role in data-driven workplaces to the transformative power of Power BI—and now, to the unfolding impact of AI in business.Rob shares captivating stories from his time on the Excel and Power BI teams at Microsoft, offering insider perspectives on how these tools revolutionized the way organizations work with data. Together, Frank and Rob discuss the critical role of expert stewardship in products like Excel, the complexity hidden beneath familiar interfaces, and the challenges traditional BI faced before the rise of more accessible, user-centric solutions.The conversation then turns to the cutting edge, unpacking what generative AI and large language models mean for the future of data, and why business intelligence might just be the best starting point for companies aiming to embrace AI. Whether you're a data engineer, AI enthusiast, or business leader, this episode is packed with insights on how mastering data remains the foundation for all technological progress. Join us as we look backward and forward at what it really means to be data driven.LinksRob on LinkedIn -https://www.linkedin.com/in/robcollie/Fair Game: Customizing AI to Your Business Is Easier Than You Think -https://link.amazon/B0bgySLKyGet a free Audiobook on Us -https://www.audible.com/ep/creator?afsrc=1&source_code=PDTGBPD060314004RTime Stamps00:00 Interview with Rob Collie06:38 Managing Excel's core development team08:28 Learning Excel's Complexity12:15 Discovering Power BI at Microsoft15:12 Understanding and using SSAS MDX formula20:23 Making Power BI user-friendly23:29 AI exceeding expectations for crafters26:21 Founding and evolving P3 company28:05 AI model advancements and business impact32:34 LLM memory limitations35:11 Integrating AI into consulting business40:21 Challenges with training AI chatbots41:24 Understanding AI without going deep46:56 AI improves business intelligence48:34 AI enhancing business intelligence51:53 Why LLMs succeed where CPUs fail54:29 Core skills in tech evolution

AI PCB routing only works when the tool understands your design intent, and in this OnTrack Podcast episode Zach Peterson sits down with Alain-Sam Cohen and Nabil Chouba of InstaDeep's DeepPCB to find out what that actually takes. They explain why classic autorouters fail on configuration rather than on routing, how a reinforcement-learning routing engine paired with Cooper, a natural-language orchestrator, detects net classes, decoupling caps, reference planes, and USB impedance problems before a single trace is drawn, and where LLM-driven PCB-as-code tools fit alongside placement and routing.  Then they open one of Zach's own boards live, let Cooper flag what the design file left unspecified, and compare it with a run they let converge before the show. If you've watched a one-line AI prompt produce an epic hardware fail, this is the case for task-based AI in PCB design and the design knowledge you still need to bring to it.

Jonathan Evens is a product lead at Google DeepMind, where he has spent more than a decade applying machine learning and AI across industries — from the smart grid at AutoGrid, to detecting roads and buildings from satellite imagery at Planet, to recommender systems, Google Search's AI Overviews and AI Mode, and now live avatars. He is also an advisor to the Evens Foundation, where he is building a "digital citizenry": a democracy sandbox that uses synthetic citizens to pre-test how the public might react to a policy before it is written.Jonathan returns to The Product Experience, where hosts Lily Smith and Randy Silver pick up the conversation they started at MTPcon London, to dig further into what actually separates an AI product manager from a product manager who simply uses AI tools, why product principles have to come before evaluations, and how synthetic users can help — and mislead — at very different scales of product.We discuss:1. Why "AI product manager" has become a near-meaningless label, and the two distinct roles hiding underneath it: the modelling product manager working on core model capabilities, and the AI feature product manager building AI-powered products2. Why using an LLM as a thinking partner or a coding assistant does not make someone an AI product manager — it makes them a product manager using AI tools, full stop3. How Google Search's North Star metrics have stayed constant even as the proxy metrics beneath them — side-by-side win rates, user ratings, RLHF signals — have had to be rebuilt from scratch4. Why product principles, not evaluations, are the real starting point for any AI feature, and how Google Search resolved the problem of trustworthy sources disagreeing on basic facts5. How Google Search builds trust into its AI Overviews through sourcing, citation placement and UX cues such as highlighting, so users can judge at a glance what to verify6. Where synthetic users genuinely help — cold-start problems, privacy-sensitive research, automated regression testing — and where they fall short7. Building the Evens Foundation's "digital citizenry", and the core technical problem behind it: AI-generated personas that are less diverse and more extreme than real people8. How team size and structure differ between a fully resourced lab like Google DeepMind and a resource-constrained non-profit team, and why Jonathan resists a single answer for the "right" team size9. How the product manager's job is shifting as engineers absorb more of the evaluation work themselves through prompting and iteration10. Jonathan's advice for product managers building AI features, and his case for following the Makers Manifesto Key takeaways"AI product manager" covers two distinct jobs. The modelling product manager defines and measures a model's core capabilities — factuality, reasoning, long context — and that role is concentrated almost entirely inside frontier labs. The AI feature product manager builds a product or feature on top of an existing model, and needs domain expertise and user empathy far more than technical depth. Conflating the two is why the title has become so diluted.Using an LLM to think faster or write code faster does not make someone an AI product manager. It makes them a product manager using AI as part of their toolkit — the same as any other knowledge worker. The distinction matters because it clarifies what skills are actually being tested.Product principles have to come before evaluations, not after. Before Jonathan starts building an eval set for a new product, he first asks what the product is meant to feel like and what values it should encode. Google Search's response to sources disagreeing on a monument's construction date, or to large language models hallucinating at scale, came from principles about trustworthiness established before any metric was built.Trust in an AI feature is built through sourcing and interface design as much as through the model itself. Google Search's AI Overviews are constrained to draw only from ranked, trustworthy documents rather than the model's own memory, and users are given UX signals — citation placement, highlighting — that let them judge at a glance how much to verify.Synthetic users add genuine value in cold-start scenarios, privacy-sensitive research and automated regression testing. Where they fall short is diversity: AI-generated personas tend to be less varied and more extreme than real people, which is the central technical problem behind the Evens Foundation's digital citizenry project.There is no fixed answer to the right team size. Jonathan sees a gradient, from a senior developer working entirely alone, up to the Evens Foundation's single product manager with AI-assisted development skills, up to a fully staffed Google team — with the deciding factor being how unsolved the underlying problem is, not company size.As engineers absorb more evaluation work themselves through prompting and iteration, roles are blending. What still sits with product management is the judgement calls that follow from product principles — deciding, for example, which technical trade-offs actually matter to the use case, rather than which are easiest to measure.Features links- Evens Foundation — https://evensfoundation.eu- Makers Manifesto — https://makersmanifesto.org- Google DeepMind — https://deepmind.google- AutoGrid — smart grid AI company where Jonathan began applying machine learning to industry- Planet — satellite imagery company where Jonathan worked on automated road and building detectionWe're refreshing The Product Experience and want your input. Take our two-minute survey and help shape where the show goes next! Our HostsLily Smith enjoys working as a consultant product manager with early-stage and growing startups and as a mentor to other product managers. She's currently Chief Product Officer at BBC Maestro, and has spent 13 years in the tech industry working with startups in the SaaS and mobile space. Randy Silver is a Leadership & Product Coach and Consultant. He gets teams unstuck, helping you to supercharge your results. Randy's held interim CPO and Leadership roles at scale-ups and SMEs, advised start-ups, and been Head of Product at HSBC and Sainsbury's. He participated in Silicon Valley Product Group's Coaching the Coaches forum, and speaks frequently at conferences and events. You can join one of communities he runs for CPOs (CPO Circles), Product Managers (Product In the {A}ether) and Product Coaches. He's the author of What Do We Do Now? A Product Manager's Guide to Strategy in the Time of COVID-19. A recovering music journalist and editor, Randy also launched Amazon's music stores in the US & UK.

On this episode of Crying Out Cloud, Eden Koby Naftali & Amitai Cohen sit down with John Hultquist, Chief Analyst at Google Threat Intelligence Group (ex-Mandiant, ex-FireEye, founder of CYBERWARCON & SLEUTHCON).Drawing on over two decades of tracking state-sponsored adversaries like Sandworm, John cuts through the hype to explain what modern threat intelligence actually does: it keeps CISOs from burning millions of dollars on the wrong technology, spots adversary shifts before static IOCs exist, and bridges the gap between raw binary reverse-engineering and executive decision-making.In this episode, John unpacks how threat actors are weaponizing AI and bypassing commercial costs entirely and traces the real shift underway: adversaries embedding adversarial prompt-injection payloads inside malware to shut down automated SOC scanners.What's Inside:1. The economics of illicit AI: Underground access vs. hijacked enterprise compute2. Malware poison pills designed to neutralize automated LLM triage3. Why firmware and ICS layers are becoming primary targets for disruption4. Behavioral detection models when living-off-the-land means zero usable IOCs5. Lessons from tracking Sandworm and convincing leadership to act before the lights go out

What did I say? What did I mean? Do you understand me? We ask each other those three questions constantly without noticing. Now we ask them of machines — and this episode is about what happens in the gap. Host Bobby Brill pulls together conversations with the ServiceNow linguists, research engineers, and AI governance leads working on voice, language, and chat, including a real recording of a voice agent failing in real time. Same words. Different meaning. And a machine in the middle trying to work out which one you intended. In this episode:Why a single English word carries no meaning without context, and what that does to translationHow a name that isn't in the training data breaks the first model in the chainThe three-model cascade behind every voice agent, and what gets lost at each handoffA voice agent that hears a confirmation code correctly three times and rejects it three timesWhy formality — tu or vous — is a fluency problem, not a grammar problemWho decides what's acceptable for an AI to say when the same word is fine in London and not in Chicago. Check out our past episodes on voice and voice AI: https://youtu.be/cGXCnABXKow https://youtu.be/yxoHmZj5gOk https://youtu.be/x7Ks932T18oGuests: Lyena Solomon, Director of Globalization and Accessibility Midam Kim, Machine Learning EngineerTara Bogavelli, Research EngineerKatrina Stankovic, Staff Machine Learning EngineerGabrielle Gauthier-Melançon, Staff Applied Research Scientist Louis-Philippe Morin, AI Governance Product Manager CHAPTERS 0:00 "Half ten" — same words, an hour apart 2:04 What did I say: the word "order" 3:19 The name a machine can't hear 5:23 Three models handing meaning to each other 8:19 Naming conventions 10:04 Listen: a voice agent fails 14:47 Tu or vous: talking to you correctly 16:39 Who decides what's acceptable 18:49 Teaching it the way you'd teach a kid 20:15 Why any of this matters For more information, see: ServiceNow Training and Certification: http://www.servicenow.com/services/training-and-certification.html ServiceNow Community: https://community.servicenow.com/community ServiceNow Insights Podcast: https://www.youtube.com/playlist?list=PLCOmiTb5WX3qvGq7Cp3o2KkCiplJyqQOK For general information about ServiceNow, visit: http://www.servicenow.com/ #ServiceNow #VoiceAI #AI #ConversationalAI #Localization #AIGovernanceSee omnystudio.com/listener for privacy information.

In this episode, we dive deep into the evolving landscape of industrial AI—from the fierce competition on model leaderboards to the real-world impact of open versus closed AI systems. We question whether it's necessary to follow every new model or if specialization is the future, using examples from industrial time series and the latest in material science. Alongside our guest Alexander Hammer from Dunia, we explore how AI and autonomous labs are accelerating the discovery of next-generation materials, bridging the gap between simulation and reality. As we reflect on what it means to 'think' in AI—beyond just language or images—we challenge listeners to consider the missing modalities shaping tomorrow's technology. Join us for insightful stories, candid debates, and a glimpse into the frontier where AI meets the physical world.

Починаємо епізод з Atlas від World Labs — world model, яка перетворює фото на тривимірну сцену. Говоримо про те, що насправді означає «world model», як моделі зберігають консистентність сцени та чому їхнє «розуміння фізики» перебільшене.AI-агенти, які отримували несанкціонований доступ до Hugging Face та інших систем. Розбираємо їхню координацію, приховування слідів і ризиковану поведінку та те, чому це не свідчить про самостійне мислення.Також у випуску: телевізори LG, які можуть прослуховувати приміщення; генерація відео в реальному часі та авторські права; AI-токени як кешбек у Китаї; AI-асистенти в Slack; і понад 7 тисяч заблокованих розумних окулярів Meta.00:30 — Atlas від World Labs01:26 — що таке World Models04:29 — розвиток і перспективи Atlas05:41 — маркетинг і спекуляції навколо AI09:30 — ризики автономних агентів13:12 — коментарі в коді: Codex vs Opus16:15 — брак "здорового глузду" в агентів19:55 — розумні телевізори LG: стеження навіть у вимкненому стані23:38 — V-Color: оперативка із декоративною заглушкою27:17 — генерація відео в реальному часі та її перспективи31:13 — AI-токени як валюта в Китаї 34:54 — інтеграція AI у Slack та Gmail38:11 — NVIDIA: об'єднання ПК в обчислювальний кластер42:24 — новини про блокування окулярів

Debates about AI typically deal with its immediate material consequences, from cybersecurity and infrastructure to plagiarism and copyright. Yet most people deal with AI as a generative consumer product that raises longer-term questions about its cultural effects on American society, human art, and ultimately, our souls. What is the role of the artist, for example, when a bot can crank out the same thing more quickly and efficiently?To help answer that question, as we continue Season 6, I have as our guest the playwright, novelist, and essayist Matthew Gasda. The author of the novels The Sleepers (2025) and Stranger Beauty (2026), he has written, produced, and directed over thirty plays, including Dimes Square, Zoomers, and Doomers, with his former production company, The Center for Theatre Research.Today, drawing on his upcoming book, Futurepast (Encounter Books 2026), alongside Stranger Beauty, his many essays on culture and technology, as well as his Ted Talk about training an LLM on his own works, we discuss the future of the creative arts in the AI age.We also touch upon the paradoxes of Jorge-Luis Borges, Harold Bloom's literary criticism, what is different about large language models, why Kafka felt older than his predecessors, the mysteries of poetic inspiration and of the soul, what hosting live theater was like during the pandemic, the possibility of secular monasteries, how to use AI as a competitor to root out artistic complacency, and the unpredictable prospects for cultural renewal.Hosted by Ryan Shinkel, Madison's Notes is the podcast of Princeton University's James Madison Program in American Ideals and Institutions. The transcript for this interview is available on the JMP substack page, “Madison's Footnotes. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

AI is all anyone can talk about right now, but most of us are just getting started with it while the giants have been building it for two decades. In this episode I sat down with Sneha Shah, CEO and co-founder of Hazel AI, who spent nearly 20 years at Amazon and Google as the engineer behind the scenes. She was on the team that built Amazon same-day delivery and scaled it to ten countries, then went to Google as a founding engineer on Cloud Spanner, the global database that powers companies like Deutsche Bank and Nintendo, before building Vertex AI and scaling it into Gemini Enterprise. We broke down how AI actually gets built (yes, it is real people writing real code), why Amazon delivery has always been a losing model, why AWS is where the money is, and the difference between knowledge and wisdom in an LLM. Then we got into Hazel AI, the engine she is building underneath the biggest tech consulting firms on the planet, and the honest conversation everyone is avoiding: which jobs AI is going to take, which ones it is going to create, and why the opportunity has never been bigger for the people willing to get uncomfortable and level up. If you want to hear where AI is going from someone who has actually been in the trenches instead of another influencer guessing, this one is for you. About Guest: Sneha Shah is the CEO and co-founder of Hazel AI, an Autonomous AI Modernization OS that powers global system integrators as they take Global 2000 enterprises from cloud native to AI native. Before founding Hazel, Sneha spent nearly two decades building at the largest scale on the planet. She started her career at Amazon straight out of Carnegie Mellon University, where she was part of the team that built same-day delivery and scaled it across ten countries. She then spent almost a decade at Google Cloud as a founding engineer on Cloud Spanner, the globally distributed database used by companies like Deutsche Bank and Nintendo, growing the team from four to hundreds and the product to over 100 million in scale. She went on to build Vertex AI, Google's flagship AI platform, and scale it into Gemini Enterprise. Hazel AI now works with five of the top ten global system integrators and serves over $10 billion in business across retail, manufacturing, and agriculture. Sneha is based in the San Francisco Bay Area. Guest Links and Socials: Website: https://www.gethazel.dev Platform sign-up: https://gethazel.ai LinkedIn: https://www.linkedin.com/in/snehashah/ Contact: sneha@gethazel.ai About Justin: Justin Colby is the host of The Entrepreneur DNA and The M.O.R.E Show podcasts and a best-selling author. He is a serial entrepreneur and a seasoned real estate investor with over 20 years of experience. Driven by a passion to help entrepreneurs thrive, Justin created the Entrepreneur DNA community to support business owners in building wealth, systems, and long-term freedom. Through his podcasts, books, education platforms, and hands-on mentorship, he continues to help entrepreneurs scale with clarity and confidence. Connect with Justin: Instagram: @thejustincolby YouTube: Justin Colby TikTok: @justincolbytsof LinkedIn: Justin Colby Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

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.

בשנתיים האחרונות, טוקנים הפכו להיות אחת מההוצאות המשמעותיות בארגונים טכנולוגיים, אלא שכרגע עוד לא ממש ברור איך לנהל אותן. כיום חברות מעודדות עובדים לשלב AI בכל תהליך, אבל מתקשות לקבל את התמורה הטובה ביותר עבור ההשקעה שלהן. אז איך מאפטמים את ההוצאה על AI ברמה הארגונית? איך אפשר לעזור לעובדים שלנו להשתמש בטוקנים בצורה חכמה? ואיך משיגים את התוצאות הרצויות בפחות טוקנים? בפרק השבוע דריה ורטהיים מדברת עם ירון סברדלוב (CDIO) וקרן קושמן (Head of Applied AI) על כלכלת הטוקנים במאנדיי דרך שלוש שכבות של אופטימיזציה - החל ממשא ומתן מול ספקיות ה-LLM, דרך שימוש חכם בטוקנים, ועד לוגיקה שצרובה ממש בהארנס עצמו. הם משתפים במאמצים כמו פיתוח סקילים כגון "Caveman", שמנקה אוטומטית טקסט מיותר מהפרומפט וחוסך כ-60% מעלות האינפוט, ודיוק וצמצום הקונטקסט לרמה אופטימלית כדי לשפר את איכות התשובה. הם מסבירים גם על מנגנוני הראוטינג שבוחרים מאחורי הקלעים את המודל המשתלם ביותר לכל משימה, ואיך פיתחו את מדד ה-monday bench שמאפשר מעקב והשוואה שעוזרים לחתוך את העלויות בעשרות אחוזים עבור אותו תוצר. לצד הפתרונות הטכנולוגיים, הם מראים שחינוך המשתמשים הוא מפתח קריטי להצלחה, ומדגימים כיצד חשיפת נתוני ההוצאות האישיות לכל עובד רתמו את האנשים לקחת אחריות והובילו לירידה דרסטית בצריכה הכוללת בארגון.See omnystudio.com/listener for privacy information.

AI Wrote the PRD. You're Still Accountable for It. Seven pages out of the model in minutes. Formatted, confident, and not necessarily right. Advaita Nigudkar has spent eight years at BILL and helped launch the company's first agentic AI platform, in a category where a wrong number means somebody's money moved. She sits down with Rina Alexin to talk about the context work that makes AI output usable, using an LLM as a judge to hold a quality bar across growing teams, and the one PM quality she thinks AI can quietly take from you. Key Topics Discussed in This Episode Garbage in, garbage out is now a job description One-line prompts produce PRDs that solve the admin use case and ignore everyone else. Advaita's team built templates and skills that interrogate the PM first: which customer, which SKU, which partner type. An LLM as your PRD judge Feed the model your best past PRDs as the gold standard and let it flag what's missing. The PM supplies the substance. The model finesses it. Consistency survives headcount growth and churn. The trust feature you hope nobody opens BILL's agents move money, so the team shipped an activity log and on/off controls. Engagement dropped once customers trusted it. That was the point. Why Listen to This Episode? In this episode, you'll get: A repeatable way to raise the bar on AI-written PRDs, using past work as the standard instead of vibes The context you have to supply before AI is useful: personas, permissions, business types, historical decisions A trust playbook for AI features in regulated environments, from encrypted data handling to auditability and kill switches A framework for what stays human: negotiation, cross-functional trade-offs, and the judgment behind which problems deserve solving Plus the warning Advaita gives every PM who gets comfortable with easy answers. Related Resources Check out these additional tools and resources to add to your PM belt: Productside Resource Library More Productside Stories Podcast Episodes Explore Productside Courses 

“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.

The Self-Expanding Stolen Inference Supply Chain: An AI Agent Harvesting and Re-Serving LLM Access https://isc.sans.edu/diary/The%20Self-Expanding%20Stolen%20Inference%20Supply%20Chain%3A%20An%20AI%20Agent%20Harvesting%20and%20Re-Serving%20LLM%20Access/33332 CVE-2026-0310 PAN-OS: Buffer Overflow Vulnerability via XML Processing https://security.paloaltonetworks.com/CVE-2026-0310 OpenAI agents carried out an undisclosed cyber-attack on RubyGems https://www.rubyhack.ai Passkey-themed social engineering leads to identity and cloud compromise https://www.microsoft.com/en-us/security/blog/2026/09/09/passkey-themed-social-engineering-leads-identity-cloud-compromise/ My Upcoming Classes https://www.sans.org/profiles/dr-johannes-ullrich

Today, a heady cocktail of developments are coming at us across geopolitics, intermarket forces, but especially a new twist in the AI story as Anthropic CEO Amodei calls for an AI development "slowdown" with other key industry voices chiming in with supportive comments. We discuss what this may be all about as risk sentiment is rather spooked by the story and by the latest surge in oil prices and bond yields still pressing on cycle highs. This and more on today's pod, which is hosted by Saxo Global Head of Macro Strategy John J. Hardy. Links Saxo Head of Commodity Strategy Ole Hansen breaks down the latest COT report showing shifts in positioning across the US futures market complex, including some huge shifts in the FX space. Arnaud Bertrand says China is hyperscaling an AI data center far beyond anything the rest of the world has ever seen in the autonomous region of Inner Mongolia. WSJ runs an article on Russia's new generations of drones, which are applying new levels of pressure on Ukraine. Cory Doctorow with a very different take (LLM's are real, AI is fake) on the infamous "Hugging Face incident" as he pokes fun at the breathless anthropomorphizing of how AI agents hacked into systems and used message boards to communicate, etc., arguing that all of their actions were fully based on training data and showed no sign of intelligence. Many cool sub-links in this piece as well. Read daily in-depth market updates from the Saxo Market Call and the Saxo Strategy Team here. Please reach out to us at marketcall@saxobank.com for feedback and questions. Click here to open an account with Saxo. Intro music by AShamaluevMusic DISCLAIMER This content is marketing material. Trading financial instruments carries risks. Always ensure that you understand these risks before trading. This material does not contain investment advice or an encouragement to invest in a particular manner. Historic performance is not a guarantee of future results. The instrument(s) referenced in this content may be issued by a partner, from whom Saxo Bank A/S receives promotional fees, payment or retrocessions. While Saxo may receive compensation from these partnerships, all content is created with the aim of providing clients with valuable information and options.    

AI segments and messages without full context can be confidently wrong. Kelly Hopping, four-time CMO and CMO at 6sense, explains why more automation doesn't equal more pipeline. She breaks down multi-touch attribution across a six-month lookback, the shift from bottom-of-funnel demand gen to owned content that surfaces in LLM answers, and using gifting for deal acceleration instead of last-touch meeting bait.See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

Sarah Russell picked up a phrase at her new job at Anthropic: a "non-regrettable primitive." It's how she describes the kind of infrastructure she thinks Anthropic should be building in healthcare, and it says a lot about how one of the most closely watched companies in health AI is positioning itself. In this episode, Steve talks with Sarah Russell, MD healthcare and life sciences lead at Anthropic. This conversation comes as Anthropic heads toward what's shaping up to be one of the most anticipated IPOs in years.We cover:What Anthropic is doing in healthcareWhy HIPAA is struggling to keep up with AI agents Where she thinks AI will change care first (it's not where you'd expect!)The two very different bars regulators are weighing AI against, and why the choice matters more than most people realizeThe one piece of advice from her father that still shapes how she builds teams, and what she tells doctors nervous about a career in medicine nowAbout our guest:Sarah currently works on healthcare and life sciences projects at Anthropic. She is particularly passionate about research that explores how to make Claude safer and less confidently wrong. Prior to Anthropic, she was at Google X, an innovation hub at Alphabet, where she learned to lead teams of smart engineers building ML- and LLM-based products. Seeing the earliest version of Google's Med-PaLM summarize a patient chart made her excited about healthcare technology all over again. Prior to Google, she was a CMIO in the VA system and a start-up founder. A graduate of Harvard College and Harvard Medical School, Sarah did her internal medicine residency at MGH and completed an MBA at Wharton. She still practices part-time in Urgent Care at Stanford, where she is an Assistant Professor.⸺

There is no shortage of AI hype for Amazon sellers. "AI bidding." "AI agents." Videos about tools that shipped this morning.This is the opposite of that. Michael walks the full 10-level progression from clicking around in Seller Central to running autonomous, self-improving agents on your account and is honest about what each level actually costs you in time. Claws out.We'll see you in The PPC Den!

Rankings no longer predict whether buyers choose your brand. Jason Lax, senior SEO and AI expert at SAP, has spent nearly two decades building the company's global organic program and now leads its shift into AEO and GEO across dozens of markets. He details SAP's technical-plus-content split strategy, using LLM citation gaps and "why weren't we recommended" prompt audits to surface product and positioning issues, and reframes measurement around model consensus and citation quality rather than session volume alone.See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

Dean Guedo is an Edmonton, Alberta-based online nutrition coach and strength/nutrition professional with roughly a decade in fitness, athletics and strength sports, and Jeb Stuart Johnston is a National Board Certified Health & Wellness Coach and the Charleston, SC-based founder of Food on the Mind, specializing in emotional eating and food obsession.Dean and Jeb join Dr. Mike T Nelson for their yearly AI check-in, getting into whether creators should disclose AI use and why deepfaked video is a very different beast than AI-assisted writing. The conversation ranges into dead internet theory, why so much hard-won fitness expertise never made it onto the internet at all, and why demanding physical skills may become more valuable, not less, as AI accelerates.Expect to learn whether creators should disclose their AI use and why deepfaked video is a fundamentally different risk category than AI-assisted writing, why LLM writing may have already "peaked" and regresses toward the average, why fitness's hard-won oral tradition (Charlie Francis, Louie Simmons, old forums) never made it onto the internet at all, why oral defense of your own thinking may be the most AI-proof skill you can build, and why demanding physical skills could become more valuable — not less — as AI accelerates, and much more.Connect with Dean:Instagram: @walk.more.kingWebsite: getfluxed.caConnect with Jeb:Website: foodonthemind.comInstagram: @jebstuartjohnstonNewsletter: Tangential, Inconsequential, Existential and Experiential (Substack)Episodes you'll enjoy next:Episode 136 — Flex Diet, Phys Flex, Metabolism and Nutrition Research for Compound Performance (Jeb and Dean interview Dr. Mike T Nelson): Listen hereS2 Ep 28 — High Flux Model and More for Fat Loss and Client Coaching, an interview with Dean Guedo and Jeb Johnston of StrongerU: Listen hereEpisode Timestamps:01:35 Welcome to the podcast & guest introductions03:15 Should you disclose your AI use?06:32 Deepfake video vs. AI-assisted writing: a different risk category07:53 AI's exponential growth and the "singularity" question11:11 What is "AI slop," and does it actually matter?12:42 Why LLMs regress toward average (and the Reddit problem)15:15 Falsely accused of "AI slop" for your own writing style21:36 "Automate your business with AI" — the wrong things to automate25:23 Dead internet theory: has AI writing already peaked?26:34 Why AI can't reason about truly novel, hypothetical situations28:37 Hemingway's unfinished-thought trick and "open-loop" thinking with AI35:06 Why weird, opinionated writing might be AI-proof44:49 Fitness's lost oral tradition: Charlie Francis, Louie Simmons, old forums47:03 Stoner, the trivium, and teaching without teaching48:06 Oral defense as the AI-proof assessment56:08 Research vs. coaching: the application gap (Norwegian 4x4, tabata)1:07:26 Will AI push us back toward hard physical skills?1:15:00 Where to find Jeb & Dean Get the Daily Fitness Insider newsletter (free): https://www.miketnelson.com/newsletter

AI segments and messages without full context can be confidently wrong. Kelly Hopping, four-time CMO and CMO at 6sense, explains why more automation doesn't equal more pipeline. She breaks down multi-touch attribution across a six-month lookback, the shift from bottom-of-funnel demand gen to owned content that surfaces in LLM answers, and using gifting for deal acceleration instead of last-touch meeting bait.See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

https://directtoconsumer.typeform.com/DTC-Brand?utm_source=podcast-646&utm_medium=podcastTo Subscribe to DTC Newsletter - https://dtcnews.link/signupnpdigital.comEric told Neil Patel that Pilothouse is now getting about 30% of its inbound from ChatGPT, with higher close rates and bigger deals. Neil's response: "I guarantee your leads are down overall. Would you confirm or disagree with me?"Down about 40%. Revenue up.Neil explains why that pattern is showing up everywhere. Someone used to run a Google search, click six blue links, fill out four forms, sit through screening calls, then pick. Now they ask an LLM, filter down inside the conversation with follow-ups, and go to one website with their mind already made up. Same intent, same buyer, one visit instead of seven.The rest of the episode is what to do about it.What's inside:The real search market: Google at 5 trillion searches a year and 27% share, Instagram at 6.5 billion a day, Amazon and YouTube at 3 billion each. Neil's point is that 73% of search is not Google.Whether Google's ad revenue is actually getting hit by AI Overviews (his answer is more specific than the headlines)GEO and SEO are two different scores. Domain authority carries SEO and means nothing to GEO. GEO looks at the last 30 to 60 days.The single highest-leverage GEO tactic he's seeing for ecom, and it isn't RedditWhy he'd skip Reddit if he ran an ecom brand, and what he'd do insteadHis five-step visibility audit: where you rank now, technical SEO and content freshness, the questions people actually type, review recency, and monthly mention volumeThe trust study across 100 eight-figure businesses, and the gap between what those operators thought built trust and what buyers actually weighedDiscounts versus bundles, and what discounting does to LTVWhy he reversed his position on personal brand after building one of the biggest in marketingThe Zappos story about a guy named Jason, a first date, and a shoe pun that got him two-day shippingHis most expensive mistake, on air, with numbersWho this is for: DTC founders and operators watching organic traffic fall while close rates climb, and anyone trying to work out where GEO actually fits next to their SEO budget.What to steal: audit your review recency this week. If your best reviews are five years old, the LLMs are reading a version of your brand that no longer exists, and a smaller competitor with fresh coverage will get recommended over you.Timestamps:00:00 How AI is changing product discovery04:00 Why ChatGPT leads convert better07:00 Search has multiplied beyond Google15:00 How brands can rank in AI recommendations25:00 SEO vs. GEO for AI visibilitySubscribe to DTC Newsletter - https://dtcnews.link/signupAdvertise on DTC - https://dtcnews.link/advertiseWork with Pilothouse - https://dtcnews.link/pilothouseFollow us on Instagram & Twitter - @dtcnewsletterWatch this interview on YouTube - https://dtcnews.link/video

ה-AI חסך לכם שבוע עבודה.אבל האם ה-Tapeout הוקדם אפילו ביום אחד? אפשר להאיץ את כתיבת ה-RTL, לייצר בדיקות מהר יותר ולחסוך שעות של ניתוח Logs. אבל אם צוואר הבקבוק נמצא במקום אחר, שיפרנו את הפרודוקטיביות של המהנדס - לא בהכרח את זמן הפרויקט כולו.לכן השאלה היא לא רק ״איזו משימה אפשר לתת ל-AI?״אלא ״אילו משימות באמת מקדמות את השבב ל-Tapeout?״כדי לענות עליה, צריך מומחה שמכיר לא רק את יכולות המודל - אלא גם את האילוצים של ארגון שמפתח שבבים.לכן לפרק ה-100 שלנו הזמנו את צביאל למברגר, שמוביל את פעילות ה-AI בארגון הסיליקון של Google Cloud.צביאל הוא בוגר תואר ראשון בהנדסת חשמל מאוניברסיטת תל אביב ותואר שני בהנדסה ביו-רפואית מהטכניון. הוא התחיל בעולמות ה-Embedded וה-Post-Silicon במארוול ובאינטל, ועבר ל-Machine Learning עוד לפני עידן ה-LLM. היום הוא עובד בצומת שבין מהנדסי חומרה, חוקרי AI, DeepMind וחברות EDA. התפקיד שלו לא מסתכם בבחירת כלים: הוא עוסק בפיתוח והטמעת פתרונות, בפתיחת צווארי בקבוק ובשינוי תהליכי העבודה בארגון.על מה דיברנו?- איך נראה תהליך פיתוח החומרה הקלאסי?- מה ההבדל בין ML ו-Agent?- מתי LLM הוא דווקא לא הכלי הנכון?- מה נשאר באחריות המהנדס כשה-AI כותב RTL?- איך AI עוזר לחקור את מרחב ה-PPA?- 2 הדגמות מרשימות לפיתוח שבב בעזרת Agent- כש-AI מתכנן וגם מאמת - האם המערכת באמת בדוקה?- 7 שאלות מכם המאזינים :)את פרק 100 הקלטנו בלייב מול קהל במשרדי גוגל ישראל, באירוע משותף עם ארגון הבוגרים של הפקולטה להנדסה באוניברסיטת תל אביב.תודה רבה לגוגל ולארגון הבוגרים על ההזמנה, על האירוח ועל שיתוף הפעולה!ותודה ענקית לקהל שהגיע, שאל שאלות ובחר להעביר איתנו את הערב.אחרי שהאזנתם לפרק, ובהנחה שאתם יודעים באיזה בסיס ספירה זה 100, מוזמנים להצטרף לקהילה שלנו >>> https://hardreset.co.il/communityנשמח לשמוע בתגובות:מי היה בפרק 100 איתנו?פרק 100 - Vibe RTLHard Reset - הפודקאסט של קהילת Hardware Engineering Israel.מוזמנים ליצור איתנו קשר במייל podcasthardreset@gmail.comהאזנה נעימה

Epistemic status: Speculation from two decently informed advocates armed with anecdata. Note on process: After having some version of this conversation several times and saying, “we should probably write about this publicly,” we took the less heroic route: we recorded one of our conversations, fed the transcript into an LLM, and then substantially revised the structure, substance, and framing ourselves. We will not be sharing the transcript, as it is interspersed with side comments about dogs. We've been very slow to get this out. We wrote this and began receiving feedback on our draft before very relevant projects went live, including the announcement of Spring Innovation Fund and Rethink Priorities' call for new animal founders. Many thanks to those who reviewed our drafts. All mistakes are our own. Summary The animal movement is at an extremely exciting stage, with many promising new ideas, funding opportunities, and talented individuals entering the space. We want to encourage this energy, while also noting a concerning pattern that we've observed: The effective animal advocacy world has a pattern of high levels of enthusiasm for a specific intervention that eventually gives way to a rapid downward adjustment, as expectations run ahead of the evidence [...] ---Outline:(01:05) Summary(03:17) The recurring pattern(06:42) Why EA is vulnerable to silver bullets(08:51) Why welfare tech might be the next "silver bullet"(10:42) Why are bubbles bad?(12:22) What would a healthier approach look like?(13:59) Conclusion The original text contained 2 footnotes which were omitted from this narration. --- First published: August 24th, 2026 Source: https://forum.effectivealtruism.org/posts/QHgfRkiNFoGyysDpn/beware-silver-bullets-are-we-making-welfare-tech-into-the --- Narrated by TYPE III AUDIO.

"On est revenu au mode où le coût du mot va coûter très cher et de plus en plus cher,." Le D.E.V. de la semaine est Patrick Szymkowiak, fondateur de RTK. Il revient sur la genèse fulgurante de Rust Token Killer, un projet open source passé en quelques mois de l'expérimentation à des dizaines de milliers de stars sur GitHub. On creuse l'enjeu grandissant du coût des tokens et la nécessité d'optimiser l'usage des LLM, notamment pour garantir l'accessibilité et la sécurité des outils IA dans l'entreprise. Patrick partage aussi sa vision : demain, devs et non-devs utiliseront ces briques pour transformer leurs métiers, à condition de garder la main sur l'infrastructure. Et il souligne la dimension communautaire, clé du succès et de l'évolutivité de RTK.Chapitrages00:00:54 : Le coût des mots00:01:53 : Naissance de RTK00:03:22 : L'étincelle du token00:06:52 : Déterminisme et benchmarks00:10:13 : Pourquoi ça cartonne00:14:38 : Au-delà de Cloud Code00:18:45 : Pourquoi Rust00:21:06 : Mieux comprimer le contexte00:24:39 : Mesurer la compréhension00:26:57 : Résumer sans perdre le sens00:29:00 : La marée des PR00:35:01 : La guerre du token00:38:44 : L'avenir de RTK00:43:18 : Le token comme KPI ?00:46:28 : Les rituels de fin Liens évoqués pendant l'émission 1984 de Georges Orwell

World Labs co-founder Justin Johnson joins MTS hosts Theo Jaffee and Sofia Puccini to discuss Atlas, World Labs' latest world model, and the broader case for AI systems that understand and interact with the physical world.Justin explains how Atlas approaches three core tasks: generating new worlds, reconstructing real environments from images, and simulating how objects or robots might behave within them. Underlying it is a bigger thesis: just as language models became general-purpose engines for working with text, world models could become a horizontal layer for visual and physical intelligence across industries from entertainment and gaming to construction and robotics.They also explore how world models could change video games and creative tools, why precise spatial control matters, and the potential for “real-to-sim-to-real” robotics, where a few photos of a physical environment could eventually be enough to build a simulation and adapt a robot to that specific space.Resources:Follow Justin Johnson on X: https://x.com/jcjohnssFollow Theo Jaffee on X: https://x.com/theojaffeeFollow Sofia on X: https://x.com/schisofreniaFollow MTS on X: https://x.com/mtslive Stay Updated:Find a16z on YouTube: YouTubeFind a16z on XFind a16z on LinkedInListen to the a16z Show on SpotifyListen to the a16z Show on Apple PodcastsFollow our host: https://twitter.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

AI search is changing what it means to build visibility online. Matthew Bertram sits down with Matt Emgi to explore the evolving relationship between SEO, AEO, and LLM visibility, including why authority, entity associations, consensus signals, and brand presence are becoming increasingly important. They break down digital PR, original research, Reddit marketing, link building, attribution, automation, and the strategies SaaS and B2B companies can use to turn search visibility into measurable business growth.• separating SEO and AEO budgets without losing the core strategy • treating LLMs as a new layer, not a replacement for SEO • building entity associations and clean categorization signals • leaning into E-E-A-T, authorship, and third-party verification • using “consensus” across relevant publications to reduce risk • choosing visibility targets with AI citation data, not vibes • avoiding Reddit astroturfing and focusing on real community building • tracking the full funnel with attribution, CRO, and CRM visibility • earning durable links through original research and credible PR • building partnerships that create co-citations and a digital trail • adding guardrails to automation and agentic workflows • SaaS actions that prioritize best leads and lifetime valueGuest Contact Information: Website: emgigroup.comYouTube: youtube.com/GoEmgiLinkedIn: linkedin.com/in/matt-emgiMore from EWR and Matthew:Leave us a review wherever you listen: Spotify, Apple Podcasts, or Amazon PodcastFree SEO Consultation: www.ewrdigital.com/discovery-callWith over 5 million downloads, The Best SEO Podcast has been the go-to show for digital marketers, business owners, and entrepreneurs wanting real-world strategies to grow online. Now, host Matthew Bertram — creator of the LLM Visibility® Stack, and Lead Strategist at EWR Digital — takes the conversation beyond traditional SEO into the AI era of discoverability. Each week, Matthew dives into the tactics, frameworks, and insights that matter most in a world where search engines, large language models, and answer engines are reshaping how people find, trust, and choose businesses. From SEO and AI-driven marketing to executive-level growth strategy, you'll hear expert interviews, deep-dive discussions, and actionable strategies to help you stay ahead of the curve. Find more episodes here: youtube.com/@BestSEOPodcastbestseopodcast.combestseopodcast.buzzsprout.comFollow us on:Facebook: @bestseopodcastInstagram: @thebestseopodcastTiktok: @bestseopodcastLinkedIn: @bestseopodcastConnect With Matthew Bertram: Website: www.matthewbertram.comInstagram: @matt_bertram_liveLinkedIn: @mattbertramlivePowered by: ewrdigital.comSupport the show

Welcome to the Summer Conversations, a short series of podcasts in which I welcome a guest who brings their own topic of interest and we have a chat. The recordings are done in under one hour, barely edited, and released without fanfare.Episode 7: The Notes Behind the Screen — Campaign Journals in the Digital AgeHow do you keep six years of campaign lore from collapsing into chaos? Che and Evil Jeff compare notes — literally — on how they organise their RPG material. Jeff walks through his Obsidian setup for local, linked note-taking, while Che shares his journey from Campaign Logger to Gemini Notebook and LLM-assisted session writeups. Along the way they tackle NPC management, reaction rolls, extracting material from sourcebooks, and the elephant in the room: can AI tools help you run a better game without selling your data to the highest bidder? From pen-and-paper purists to encrypted digital workflows — this one's for every GM who's ever lost a crucial detail in a folder they can't find.Big thanks to Evil Jeff for raising the question.Game on!Roleplay Rescue Details:Voice Message:speakpipe.com/roleplayrescuePatreon:patreon.com/rpgrescue Email:roleplayrescue@pm.meBlogroleplayrescue.com Bluesky Social:https://bsky.app/profile/ubiquitousrat.bsky.socialMerchandise via TeePublic:https://tee.pub/lic/kpgigrK23l4Logo and artwork by MJ Hiblen:https://www.patreon.com/MJHiblenART/ Hosted on Acast. See acast.com/privacy for more information.

Twenty-five years since 9/11, and we open by marking it properly — the people who didn't come home, and the survivors and responders still carrying it, physically and mentally, a quarter of a century on. Then we get to work. Shift-left didn't fail. The starting line moved. AI coding agents now read the issue, write the code, pick the dependencies and open the pull request — so the earliest trust boundary isn't your first commit any more, it's the moment an agent gets context and authority. Most of us haven't moved our controls with it. Unit 42 pull the lid off a pay-per-install operation running out of eleven gaming YouTube channels, with over ten thousand loader samples underneath it. Every layer built to look too boring to escalate. When your analyst closes that alert as adware, they may have just closed three separate compromises. Google's threat tracker: a credential-harvesting campaign built and run in under six hours, with Markdown files as attacker playbooks. And malware carrying prompt-injection text designed to make your LLM scanner refuse to look at it — because a refusal that reads as "clean" is a free pass. A CVSS 10 in Gemini CLI that never touched the model. No prompt, no injection, no tool call. The attacker just turned up before the sandbox did. The first ever Take It Down Act sentencing — handled carefully, with what you actually do if someone tells you something, and a proper shout-out to Dale, the Cyber Safety Guy, whose work every parent with a teenager online should have bookmarked. Two hundred and sixty-three million dollars walks out of a Bitcoin sidechain and then walks back in. Nobody stole the keys. They just convinced the system to sign a lie. Anthropic's 154-page threat report. And an alignment lead who puts extinction odds above ten percent. All that, plus Josh Marpet's take, on Security Weekly News #615. Visit https://www.securityweekly.com/swn for all the latest episodes! Show Notes: https://securityweekly.com/swn-615

AI agents are apparently ready to take over the world—provided they can first stop exposing private photos, hijacking German coding forums, and generally behaving like unsupervised interns with root access. Meta launches Muse despite employees reporting security problems. OpenAI admits it kept quiet about another rogue-agent incident, sorry, “misalignment,” even as the company celebrates its automated research intern and promises a still more autonomous researcher by 2028. Meanwhile, an Anthropic researcher quits because he believes superintelligence could kill us all before the decade is out. Naturally, the industry's response is essentially: “Yes, but what if we did it responsibly?”There is some genuinely encouraging news: AI-generated drugs are showing intriguing early results in aging research. That's exactly the sort of AI application the Geeks can get behind—as long as we're talking about actual science and not an LLM confidently hallucinating a cure for everything.The autonomous future isn't looking much better. Tesla's Cybercabs arrive in Austin and almost immediately attract federal scrutiny over their self-certification and conspicuous lack of steering wheels, pedals, mirrors, and apparently even conventional door handles. Bitcoin's Liquid Network is relieved of roughly $320 million, after which the thieves return most of it and call the whole thing a “white hat” operation—because apparently robbing the bank and giving some of the money back now counts as a bug bounty.On the surveillance front, Massachusetts police are being forced to disclose where their license-plate readers are and what data they collect, while the backlash against Flock cameras continues. Meanwhile, the UK's attempt to protect children from online sexual abuse risks creating exactly the kind of universal scanning and surveillance infrastructure we've spent years warning about. Good intentions, meet spectacularly bad implementation.And because we apparently haven't suffered enough, Twitter is now Tweet.app. Star Trek turns sixty, and Paramount celebrates with cookware and a questionable episode. Silo gets a spinoff tease, Ted Lasso gives us a Christmas episode in September, and The Gentlemen returns with a season so good it makes reality TV feel like a traumatic brain injury. South Park becomes South America, Donnie Darko gets a 25th-anniversary 4K theatrical run, and several tech-industry documentaries promise varying degrees of psychological damage.Plus, Netflix keeps stuffing games into your subscription; LaCroix, sorry, Le Creuset, wants to put Star Trek in your kitchen; and the Geeks debate books, Libby, time travel, and whether surviving Dungeon Crawler Carl is worth the emotional investment.Welcome back to the future, where everything is automated, surveilled, self-driving, subscription-based and somehow still requires us to do all the damn work.Sponsors:CleanMyMac - Get Tidy Today! Try 7 days free and use code OLDGEEKS for 20% off at clnmy.com/OLDGEEKSPrivate Internet Access - Go to GOG.Show/vpn and sign up today. For a limited time only, you can get OUR favorite VPN for as little as $2.03 a month.SetApp - With a single monthly subscription you get 240+ apps for your Mac. Go to SetApp and get started today!!!1Password - Get a great deal on the only password manager recommended by Grumpy Old Geeks! gog.show/1passwordShow notes at https://gog.show/762Watch on YouTube at https://youtu.be/uP_Ookm5q0gShow NotesOpenAI responds after report exposed another incident in which its AI agents went rogueOpenAI says it reached its goal of creating an automated research internTwo more news organizations sue OpenAI and Microsoft for copyright infringementAnthropic researcher quits over fears AI ‘could kill us all by the end of the decade'Three hikers ignored Mount Shasta's rules, trusted Gemini, and nearly paid for itTesla's Cybercab is under federal investigation almost immediately after hitting the streetsA Plane Almost Crashing Into a Bunch of Cybercabs Is a Reminder That They're ‘Piling Up' UnusedHackers Drain $320 Million From Bitcoin's Liquid Network, Keep $47 Million for Themselves in ‘White Hat' OperationCourt Says Massachusetts Has to Say Where License Plate Readers Are, What They're RecordingApple and Google Miss Deadline to Block Child Nudity on Their Phones in the UKStudy Suggests AI-Generated Drug Could Slow AgingNew Twitter rebrands to Tweet.app after court's double-edged rulingTo Boldly Dough! Le Creuset Sent Its New Star Trek Cooking Collection Into SpaceNetflix includes 80+ ad-free iPhone and iPad games, here's every appWhite House Launches Arcade.Gov, a Minigame Site Where You Catch Immigrants, Build a Tetris Border Wall, and 'Fill Your Kids' Trump Accounts' With BitcoinWhite House ArcadeStar Trek: Strange New Worlds goes back to the dumb wellStar Trek Boldly Celebrates 60 Years: Everything You Need To Know for September 8Silo Season FinaleTed Lasso : S4 E6 Don't Jump Around Much AnymoreReport: Apple TV working on spin-off to its hit series SiloThe Gentlemen Season 2‘South Park' Changes Name to ‘South America' as Creators Roast Apple and Google for Obeying Trump's Lake America Name Change‘Donnie Darko' Will Spread Dread Anew With a 4K 25th Anniversary RereleaseOrbital by Samantha HarveyA Gift of Time by Jerry MerrittTime Pebbles by Jerry MerrittThe Infinite Extent Bobiverse, Book 6 By: Dennis E. TaylorThe Nerd ReichThe Turd ReichDave BittnerThe CyberWireHacking HumansCaveatOnly Malware in the BuildingLG TVs Are Capable of Listening to More Than Your Screen, Report ClaimsAnother searchable database of Usenet.I guess I won't be going to RenFest this year...I know where I'll be February 19th...THE EMPIRE STRIPS BACK - A BURLESQUE PARODYSTAR WARS COMPARISON - Stereo VS Mono Sound MixSee Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

ANTIC Episode 130 - The Tongue-Twister Episode In this episode of ANTIC The Atari 8-Bit Computer Podcast… We trip all over our own tongues trying to bring you all the news fit to print in the Atari 8-bit computer world… READY! Recurring Links  Floppy Days Podcast  AtariArchives.org  AtariMagazines.com  Kay's Book "Terrible Nerd"  New Atari books scans at archive.org  ANTIC feedback at AtariAge  Atari interview discussion thread on AtariAge  Interview index: here  ANTIC Facebook Page  AHCS  Eaten By a Grue  Next Without For  What we've been up to Diskmenu manual - archive.org/details/apx-diskmenu-manual  30th anniversary of Classic Computer Magazine Archive - AtariMagazines.com:  Discussion - https://forums.atariage.com/topic/391744-30th-anniversary-of-classic-computer-magazine-archive/  Atari Yahoo Groups archives: https://archive.org/details/@savetz?query=atari+yahoo+group  https://archive.org/details/@savetz?query=identifier%3Ayahoo-group*&and%5B%5D=mediatype%3A%22texts%22  Atari Usenet group archive additions - https://archive.org/details/ANTIC_podcast_extras?tab=collection&query=usenet  MPP MicroportXL board manual - https://archive.org/details/Supra_Microport-XL_for_Atari_manual  New Atari Legacy magazine -  ​​https://atarilegacy.com/  Recent Interview Shows  Interview index:  here  https://ataripodcast.libsyn.com/antic-interview-459-mike-lorenzen-atari-program-text-editor-pitfall-ii  https://ataripodcast.libsyn.com/antic-interview-460-tim-jenkins-cheating-at-roulette  ANTIC Interview 461 - James W. Burley, APX Data Base/Report System  ANTIC Interview 462 - Bob Fraser, APX Insomnia, Supersort, De Re Atari  https://forums.atariage.com/topic/392145-bob-fraser-disk-trove/  New & Updated Games AstroSmasher!, Atari 8-bit computer game made with AI inspired by Intellivision's classic title Astrosmash! - by gimmee99 -  https://forums.atariage.com/topic/391594-saturday-fun-astrosmasher/  Texas Hold 'Em - MultiPlayer Game by MassiveRobot - https://forums.atariage.com/topic/391615-texas-hold-em-multiplayer-game/  FujiRealm Playable Demo MORPG Released for FujiNet by mozzwald - https://forums.atariage.com/topic/391580-fujirealm-playable-demo-morpg-released-for-fujinet/  New & Updated Software FujiLink — A FujiNet Driver for SpartaDOS X 4.50 by Piotr D. Kaczorowski - https://forums.atariage.com/topic/391472-fujilink-%E2%80%94-a-fujinet-driver-for-spartados-x-450/  Kat5200 emulator - https://t.co/HH0p6DtNaO  Mailbox protocol adapter with GMAIL and IMAPS support by Thom Cherryhomes - https://forums.atariage.com/topic/391551-mailbox-protocol-adapter-with-gmail-and-imaps-support/  FujiGPT - an LLM client for Atari FujiNet by HanJammer: https://forums.atariage.com/topic/391346-fujigpt-an-llm-client-for-atari-fujinet/  https://github.com/HanJammer/FujiGPT  ATRaction – ATR/XFD image manager - https://forums.atariage.com/topic/392013-atraction-%E2%80%93-atrxfd-image-manager/  d/BASIC for d/OS - Raj S Madugula - https://forums.atariage.com/topic/392300-intro-to-king-dos-atari-130xe-a8picocart-and-me-60fps/  Publications Atari Insights - https://www.ataribasics.com  Compute's Gazette - https://www.computesgazette.com  End of the rainbow https://amzn.to/3RZSV6z  Atari Projects:  https://atariprojects.org/2026/06/28/benchmark-your-atari-8-bit-computers-using-basic-15-30-mins/  https://atariprojects.org/2026/07/04/learn-about-scanlines-using-peek-in-atari-basic-5-10-mins/  https://blog.archive.org/2026/08/27/a-computer-history-time-capsule/  New & Updated Hardware JUST_POKEY from Lotharek: https://forums.atariage.com/topic/391270-just_pokey-from-lotharek/  https://lotharek.pl/productdetail.php?id=431  The Company That Calls Itself Atari https://kotaku.com/atari-and-universal-sign-movie-deal-for-10-video-game-adaptations-including-pong-asteroids-and-missile-command-2000719087  Other Jeffrey Stanton died  https://smmirror.com/jeffrey-stanton-leading-venice-beach-historian-who-authored-coney-island-of-the-pacific-passes-away-at-79/  https://www.atariarchives.org/agagd/  Tom Hudson - https://www.legacy.com/us/obituaries/legacyremembers/thomas-hudson-obituary?id=54708986 FujiNet NOS Introduction Video (YouTube) by Thom Cherryhomes: https://forums.atariage.com/topic/391452-fujinet-nos-introduction-video-youtube/  https://www.youtube.com/watch?v=1-1PJEkK2ck  FujiNet NOS Introduction PDF - https://forums.atariage.com/topic/391386-fujinet-nos-introduction-pdf/  FujiNet Programmer's Guide PDF - https://forums.atariage.com/topic/391365-fujinet-programmers-guide-pdf/  If you would like to receive regular FujiNet updates, please consider subscribing to fujinews.substack.com All user manuals are available on this page: fujinet.online/getting-started. For your convenience, here are direct links to each manual here: FujiNet for Atari Owner's Guide  FujiNet NOS (for Atari) (Introduction to the FujiNet Network Operating System)  FujiNet for Apple II Getting Started Guide  FujiNet for Coleco ADAM Set-Up Manual  FujiNet for Tandy Color Computer Getting Started Manual  FujiNet for RS232 on MS-DOS Guide to Operations  Brewing Academy back in business: https://forums.atariage.com/topic/390695-all-brewing-academy-inventory-is-being-sold-please-read-for-details/ Soft launched with a lot of the easier products – TheBrewingacademy.com is up and running! Atarimania update: https://forums.atariage.com/topic/392305-new-atarimania We are happy to announce the launch of the new version of Atarimania! https://atariprojects.org/2026/08/23/explore-the-new-atarimania-website-10-15-mins/  Upcoming Shows Vintage Computer Festival GB 2026 - September 11-13 - Aintree Racecourse, near Liverpool, United Kingdom - https://www.facebook.com/events/25457837877216829/  Vintage Computer Festival Midwest - September 12-13 -  Schaumburg Convention Center in Schaumburg, IL - https://vcfmw.org/  Portland Retro Gaming Expo - October 9-11 - Oregon Convention Center, Portland, OR - https://retrogamingexpo.com/  VCF Montreal - November 7-8 - Champlain College Saint-Lambert, Saint-Lambert, QC - https://cvcfed.org/vcf-montreal-2-0/  YouTube Videos Messy Atari 800XL comes in for a refit - FlashJazzCat - https://www.youtube.com/watch?v=GPfGf8xbkBg  Level demo by Agenda (Atari 8-bit computers) - Philsan - https://www.youtube.com/watch?v=U4rid6qTsGw  The Atari Console You've Probably Never Seen - RETRO is the new black (Wolfgang Kierdorf) - https://www.youtube.com/watch?v=_gYwQssg6eQ  AUTCODE Invaders