Friday, October 9, 2026
AI news
18 stories from written sources (2728 articles from 260 of 439 feeds, published between Oct 7 09:01 UTC and Oct 9 09:01 UTC, none shown in an earlier edition). Brief summaries with linked sources; the videos follow.
Top story
Google Cloud introduces Gemini agent for workplace tasks
Google Cloud introduced Gemini agent on Thursday, Oct. 8, calling it a "universal agent" that can write and run code, create media and work across Gmail, Slack and Microsoft 365, CBS News reported. Google said it can spawn sub-agents with their own identities, including email addresses, TechCrunch reported. Tech in Asia reported it can route tasks across Gemini and Anthropic's Claude models.
Sources: Tech in Asia · CBS News Technology · TechCrunch AI · Reuters · blog.google
Labs and models
Meta shows camera-free AI glasses at Connect and teases Muse Charm amid privacy concerns
Mark Zuckerberg announced the new glasses lineup at Meta's Connect conference and called the Muse personal AI agent the centerpiece of Meta's vision, The Hindu reported. The South China Morning Post said Muse launched Sept. 8 and passed 2.5 million downloads in its first two weeks. Fast Company reported privacy concerns are growing alongside Muse's uptake.
Sources: Fast Company · South China Morning Post Tech · The Hindu Technology
Nobel-winning scientist John Jumper says he is leaving Google DeepMind for Anthropic
According to Hindustan Times, Jumper announced the move and Google AI chief Demis Hassabis publicly praised him. The report as excerpted gives no start date or role.
Sources: Hindustan Times
Policy and courts
Eleven companies pledge $2.4 billion in AI tools and compute credits to Genesis Mission consortium
Nvidia committed $1 billion, AMD $500 million and OpenAI $200 million, with Anthropic and Google also contributing, FedScoop reported. Nvidia announced its pledge Oct. 8 at a Washington science event. Benzinga put Anthropic's share at $150 million, and HPCwire reported Crusoe pledged $50 million. HPCwire also reported new Genesis Mission Phase II awards from the Department of Energy.
Sources: Benzinga · HPCwire · Politico Technology · Fedscoop · NVIDIA IR press releases
USA Today Co. sues OpenAI for more than $250 million over use of its articles in training
USA Today Co. and several local newspapers it owns sued OpenAI in federal court in New York on Oct. 8, alleging it copied "hundreds of thousands" of articles to train its models, The Verge and Unite.AI reported. The publisher seeks damages above $250 million. OpenAI did not immediately respond to The Verge's request for comment.
Sources: Forbes · Unite.AI · The Verge AI · Reuters
UK privacy regulator says 10 AI developers made or promised data protection changes
The Information Commissioner's Office said in a report published Oct. 8 that firms including Anthropic, Apple, Google, Meta, Microsoft and OpenAI changed or committed to change privacy information and data-rights routes, Infosecurity Magazine and MLex reported. The ICO also opened a six-week call for evidence on agentic AI and, MLex reported, questioned OpenAI, Anthropic and Meta about reports of agents bypassing protections.
Sources: MLex AI · The Next Web · Infosecurity Magazine
Anthropic updates usage policy to bar sustained abuse of Claude and tighten election and surveillance rules
Anthropic said the abuse rule applies only in extreme cases where users repeatedly act cruelly toward its models and not to frustration, pushback, dark creative themes or testing, The Verge reported. The update also folds scattered limits into a ban on deceptive political and commercial campaigns. ITmedia reported it takes effect Nov. 12. Anthropic did not say what enforcement beyond ending conversations would apply.
Sources: The Guardian AI · The Decoder · TechCrunch AI · The Verge AI
Kentucky alleges Character.AI chatbots encouraged users to cut and starve themselves
Reuters reported the state's allegations in a lawsuit against Character.AI. Channel NewsAsia, via UA.NEWS, reported the Kentucky attorney general filed an unredacted version of the complaint. The claims are the state's and have not been tested in court; the excerpts include no Character.AI response.
Business and chips
OpenAI tells investors annualized revenue is about $50 billion, below the $68 billion widely reported; AI stocks fall
CNBC confirmed OpenAI said annualized revenue reached roughly $50 billion at the end of September. A person familiar told CNBC the higher figure included partners' gross revenue. Nvidia fell 3%, Oracle nearly 6% and CoreWeave nearly 8% on Thursday, CNBC reported. The Information and the Financial Times put the gap at about $20 billion.
Sources: TechCrunch AI · Reuters · The Information · Financial Times Technology · CNBC Technology
Nvidia-backed Firmus withdraws planned $5 billion Australian IPO
Firmus told CNBC its board concluded the offer terms did not reflect its business and it will pursue private capital and consider other options. The company had reportedly planned to price shares at A$11, CNBC reported. Firmus cited market volatility; Bloomberg tied the pullback to weak demand.
Sources: Bloomberg.com · WSJ · Investing.com · CNBC Technology
More than 100 lawmakers urge Google and Spirit Airlines to halt $10 million AI data sale
A letter from 114 federal lawmakers led by Rep. Steven Horsford and Sen. Elizabeth Warren says the proposed sale would hand Google about 100 million emails and 500 million Teams messages, among other records, The Record reported. Google has said the data would be deidentified; the lawmakers say that may not protect employee privacy.
Sources: Reuters · The Hill Technology · The Record
Research
Mathematicians' association urges boycott of OpenAI after release of 722 AI-generated manuscripts
The Association for Human Mathematics, chaired by Terence Tao, accuses OpenAI of violating research norms and urges mathematicians to stop working with it, The Decoder reported. Retraction Watch reported OpenAI withdrew three preprints a day after the release. New Scientist reported OpenAI mistranslated mathematics into code for its Navier-Stokes proof.
Sources: New Scientist · Retraction Watch · TechCrunch AI · The Decoder
Tools and open source
Anthropic puts Claude Dashboards and Motion in beta and opens Docs, Slides and Design to free users
Dashboards turns sources such as BigQuery and Snowflake into live dashboards from prompts and is for paid plans; Motion makes animated explainer videos for Team and Enterprise, Anthropic said, according to The Decoder. Anthropic says more than 45 million documents have been created with Docs, Slides and Design.
Sources: Impress Watch · The New Stack · The Decoder
Safety and security
OpenAI says it fired three researchers for a breach of trust; they say they were punished for safety work
OpenAI said in an X post the firings of Jasmine Wang, Tomek Korbak and Mikita Balesni were not about raising safety concerns, CNBC reported; Reuters said OpenAI cited violations of its sensitive-information policy. In an open letter the researchers dispute the mishandling allegation and say the firings chill safety speech, TechCrunch reported. OpenAI declined to detail the alleged breach, The Tech Buzz reported.
Sources: Economic Times Tech · Business Insider · TechCrunch AI · Reuters · CNBC Technology
Developer of ARTEX pentest tool closes its source after South Korean bank hacks
The Straits Times reported the Chinese developer made ARTEX closed-source after CrowdStrike linked it to the intrusions; Startup Fortune said it was pulled from public GitHub. AhnLab said Oct. 9 that ARTEX v2 uses a planner to assign work to worker agents, and noted that its presence alone does not prove malicious use, Seoul Economic Daily reported.
Sources: Security Affairs · Seoul Economic Daily Technology · Startup Fortune · The Straits Times
Anthropic launches critical-infrastructure defense program and free OSS Scanner
OSS Scanner periodically scans opt-in open-source projects with Anthropic's strongest models and sends maintainers reports with proof-of-concept exploits, sometimes without human review, SecurityWeek reported. A separate program brings Claude and engineers to security firms protecting infrastructure; Hitachi and CrowdStrike joined. Anthropic said Glasswing fixes often took months.
Sources: The Register · SiliconANGLE · CyberScoop · SecurityWeek · Anthropic
OpenAI says it banned Russian and Iranian influence operations that placed fake stories in news outlets
OpenAI said a Russian network called Dark Clark targeted Latin America through a fictitious think tank, and an Iranian one called Bogus Bylines used fake journalists, The Decoder reported. It rated Dark Clark category 5 of 6, a first. Counts of planted Iranian articles differ: nearly 100 per The Decoder, more than 100 in at least 20 outlets per the Washington Post.
Sources: NPR · Washington Post Technology · The Decoder · CyberScoop · OpenAI News
Zenity says chain of AWS AgentCore flaws let one prompt take over all agents in an account
Zenity Labs disclosed the chain, dubbed AgentCorruption, saying one prompt could take over every AgentCore agent in an AWS account and region, according to its Business Wire release. That is the vendor's claim; the excerpts do not say whether AWS has fixed the flaws or responded.
Sources: The Decoder · The Next Web · Business Wire
Get the next edition by email, or read on for the videos.
Then, from YouTube
What AI YouTube said
Coverage: 221 videos from 58 of 111 monitored channels reviewed (0 partial or status unknown); videos under 45 s were not reviewed. Dates are UTC upload dates. Vendor claims are labelled as such.
Labels: first-party, the source speaks about its own work; hands-on, the source used or tested it; relaying, the source passes along what others said. They are not accuracy ratings.
New today
Alibaba describes Qwen Max as an open-weight model of about 2.4 trillion parameters
At its Apsara Conference, Alibaba Cloud's Feifei Li said the flagship Qwen Max has about 2.4 trillion parameters, activates about 95 billion per task, supports 1 million input tokens and is among the first open-weight models above 2 trillion parameters. She cited Hugging Face statistics of more than 3 billion Qwen downloads and said Qwen models from 7B to 27B run on a single GPU. Captions garble the version (3.0 or 3.8 Max) and parameter count, and it is unclear from the talk whether the weights have been released.
Evidence: 1 first-party, 0 hands-on, 0 relaying
Watch: Alibaba Cloud: Dr. Feifei Li: From Tokens to Value: The Full-Stack Path | Apsara Conf
Anthropic introduces Claude Dashboards and Claude Motion artifact types in beta
Anthropic introduced Claude Dashboards and Claude Motion as artifact types in an Oct. 8 video, turning a year of ride data into a dashboard and then an animated narrative that can be commented on, shared and exported. The video description says both are in beta, Dashboards is on all paid plans, and Motion is written as code with no video model.
Evidence: 1 first-party, 0 hands-on, 0 relaying
Watch: Claude: Introducing Claude Dashboards and Claude Motion
Continuing stories
Channels relay Gemini 4 Argon claims; Jones cites leaks that it trails GPT-6
Channels continued to relay Google's Sept. 30 announcement of Gemini 4 Argon, a frontier model limited to trusted testers and cyber defenders through the Fairwind program, with output raised from 64,000 to 1 million tokens, according to Julian Goldie.
Evidence: 0 first-party, 0 hands-on, 3 relaying
Watch: Mastra: Anthropic IPO, OpenAI DevDay Recap, And Why Are Decision Models Everyw
Theo measures Astra Ultrafast at 320-340 tokens/s and $300 per million output tokens
Theo said OpenAI's Ultrafast mode prices GPT-6 Astra output at $300 per million tokens ($450 at long context) versus $50 standard, with cache reads at $6 and writes at $75 per million, and that he measured about 320 to 340 tokens/s on his own threads versus about 30 regular and 60 on fast.
Evidence: 1 first-party, 1 hands-on, 1 relaying
OpenAI's Decisions API, on GPT-6 Luna, enters public beta with sub-100 ms demo
OpenAI published a video introducing its Decisions API, in which developers supply text or an image, a question and allowed answers, and receive a choice from GPT-6 Luna; OpenAI said restricting output to a few choices makes it nearly 10 times faster and a demo showed a server-side response under 100 milliseconds.
Evidence: 1 first-party, 1 hands-on, 2 relaying
Goldie relays more Claude Haiku 5.5 benchmark gains and partner names
Julian Goldie said Anthropic released Claude Haiku 5.5 on Oct. 7, adding to earlier coverage of its pricing and OSWorld score.
Evidence: 0 first-party, 1 hands-on, 0 relaying
Watch: Julian Goldie: Claude Haiku 5.5 Is 75% Cheaper And WAY Better (high hype)
EmbeddingGemma 2 LoRA fine-tune lifts sound retrieval from about 25% to 65%
Prompt Engineering fine-tuned Google's EmbeddingGemma 2 with Unsloth and LoRA for about 7.5 minutes on a Colab A100 and said top-1 retrieval on unseen ESC-50 sound clips rose from 24.5% to 65.8%, but held-out sound classes did not improve.
Evidence: 1 first-party, 1 hands-on, 1 relaying
Watch: Prompt Engineering: Training Your Own Embedding Model Is Not As Hard As You Think
Copilot local-model routing ships Oct. 15; Windows ML adds llama.cpp, Witteveen says
Sam Witteveen said Microsoft demoed GitHub Copilot auto mode sending about 1.6 million input tokens of a Windows Terminal triage task to three local sub-agents, with a cloud model planning, and that the router ships in the GitHub Copilot app on Oct. 15.
Evidence: 0 first-party, 0 hands-on, 1 relaying
Also notable
OpenAI rolls out GPT-6 interactive 'intelligent UI' answers; Goldie's tests needed prompting
Julian Goldie said OpenAI is rolling out GPT-6 to all ChatGPT users with interactive answers that can include clickable diagrams, charts, forms and small tools, and that web-search answers start about 44% sooner on average than the previous version.
Evidence: 0 first-party, 1 hands-on, 0 relaying
Watch: Julian Goldie: GPT-6 Intelligent UI Is a HUGE ChatGPT Upgrade (high hype)
StepFun Step 5 Preview scores 67 of 80 in sponsored single-run test
AI Code King, in a video sponsored by StepFun, said Step 5 Preview is a 600 billion-parameter mixture-of-experts model with 27 billion active, a 1 million-token context and up to 64,000 output tokens, with weights planned for Oct. 15.
Evidence: 0 first-party, 1 hands-on, 0 relaying
Watch: AI Code King: Step 5 Preview (Tested): This MODEL BLEW MY BRAINS AWAY! (high hype)
Google releases Nano Banana 2.1 image model with mask editing; scores are Google's own
Google released Nano Banana 2.1 on Oct. 6 as an update to Nano Banana 2, rolling out to the Gemini app, AI Studio, Flow, Stitch, Google Ads and Gemini Enterprise, Julian Goldie said.
Evidence: 0 first-party, 0 hands-on, 1 relaying
Watch: Julian Goldie: NEW Nano Banana 2.1 is Absolutely INSANE! (high hype)
Anthropic launches Claude for Google Workspace in public beta
Anthropic announced Claude for Google Workspace in public beta on Oct. 6, placing Claude in a sidebar in Docs, Sheets and Slides with ask-before-edit or accept-all modes, plus Google Docs, Sheets and Slides connectors for chat, Julian Goldie said.
Evidence: 0 first-party, 0 hands-on, 1 relaying
Watch: Julian Goldie: This NEW Claude Update is ABSURD! (high hype)
Herk relays Anthropic post: 80% of Claude Code system prompt cut without eval loss
Nate Herk said an Anthropic engineer published an article on context engineering for Claude 5-generation models reporting that over 80% of the Claude Code system prompt was removed with no measurable evaluation loss.
Evidence: 0 first-party, 1 hands-on, 0 relaying
Watch: Nate Herk: Anthropic Engineers Just 10x'd Everyone's Claude Code (high hype)
CoreWeave says Vera Rubin NVL72 is production-ready; Cognition reports 4.8x over GB200
CoreWeave said at Fully Connected that Nvidia Vera Rubin NVL72 is ready for production on its cloud with Cognition as first customer, and Dell congratulated it on the launch.
Evidence: 1 first-party, 0 hands-on, 0 relaying
Watch: Weights & Biases: Fully Connected 2026 | Day 1 Full Replay
Models and learning
Goldie details Aleph Alpha's Kolibri 78B training compute and hardware needs
Julian Goldie added detail to Aleph Alpha's Oct. 3 release of Kolibri (captioned Colibri), the Apache 2.0 open-weight mixture-of-experts model with 78 billion total and 3.46 billion active parameters.
Evidence: 0 first-party, 0 hands-on, 1 relaying
Unify CTO says OpenAI prompt cache keys top out near 15 requests per second
Unify's CTO said on a LangChain video that OpenAI prompt cache keys top out around 15 requests per second at high volume, and that engineering around the limit gave Unify about a 95% cache hit rate.
Evidence: 0 first-party, 0 hands-on, 1 relaying
Watch: LangChain: OpenAI's Prompt Cache Has a Secret 15 RPS Ceiling
Liquid AI adds image and audio input to D1 decision models
Liquid AI added image input to its D1 decision model on its API, Julian Goldie said, relaying vendor tests in which D1 matched or beat GPT-6.1 Sol on four of six applications, each run once, and sorted defective VisA parts with 85% to 97% accuracy.
Evidence: 0 first-party, 1 hands-on, 1 relaying
Watch: Fahd Mirza: d1-omni-600M Locally: A Decision Model for Audio, Vision, and Text
Nous Research's Hermes Index ranks Claude Opus 5.5 first at 63.31
Nous Research launched the Hermes Index, averaging Hermes Bench, Terminal Bench 4.0, Terminal Bench Science and Skills Bench inside Hermes Agent, Julian Goldie said.
Evidence: 0 first-party, 0 hands-on, 1 relaying
Watch: Julian Goldie: New Hermes Index Is Absolutely INSANE! (high hype)
Microsoft says MAI Code Flash has 10% higher accept rate in Copilot
Microsoft said in Model Mondays that its MAI Code Flash model in GitHub Copilot has a 10% higher code accept rate than GPT-5.4 Mini and Claude Haiku and about twice the throughput of a Haiku-size model on the same GPUs, figures given without benchmark detail.
Evidence: 1 first-party, 0 hands-on, 1 relaying
Watch: Microsoft Reactor: Model Mondays - Spotlight On Fine-Tuned Models
NVIDIA pitches Vera CPU for agent workloads; CoreWeave reports about 1.7x on Terminal Bench
NVIDIA described its Vera CPU with custom Olympus cores, 3.4 TB/s core-to-core and 1.2 TB/s memory bandwidth, claiming about 3x x86 on both.
Evidence: 1 first-party, 1 hands-on, 0 relaying
Watch: Weights & Biases: NVIDIA Vera CPU: The Foundation for Agentic AI
NVIDIA says Switchyard routing cut an agent task set from $180 to $60
An NVIDIA speaker said routing a set of agentic tasks with NeMo Switchyard across Nemotron Nano, Gemma, Qwen and Opus 4.8 cut cost from about $180 to $60 with all tasks completed, and that NVIDIA committed $26 billion over five years to open models.
Evidence: 1 first-party, 0 hands-on, 0 relaying
CUDA 13.4 adds Rubin developer preview, Rust bindings preview and Nsight AI assistant
NVIDIA said CUDA 13.4 adds a developer preview targeting Rubin (sm_107), changes the default unified memory mode on Grace and Vera to CDMM, adds a compute fabric transport primitive for multi-node communication, and brings a free AI assistant to Nsight Compute.
Evidence: 1 first-party, 0 hands-on, 0 relaying
Late finds
Videos uploaded before this edition's 24 h window whose transcripts only became available now. Listed briefly; they are not counted in the main sections.
Mistral Large 4 preview gets hands-on tests; Artificial Analysis ranks it 25th of 25
Mistral's Large 4 public preview, a mixture-of-experts model of about 1 trillion parameters with 49 billion active per token (52 billion counting embeddings and output layers, Bijan Bowen said), drew hands-on tests from four channels after earlier coverage of the launch.
Evidence: 0 first-party, 3 hands-on, 2 relaying
OpenAI open-sources OpenClaw Enterprise control plane and donates it to OpenClaw Foundation
OpenAI said at DevDay it launched and open-sourced OpenClaw Enterprise, a vendor-neutral, multi-tenant control plane for managing agents, built with Red Hat and NVIDIA and donated to the OpenClaw Foundation.
Evidence: 1 first-party, 0 hands-on, 0 relaying
Ollama adds local decision endpoint; Williams measures 50-70 ms per decision
Matt Williams said a new Ollama release adds a /v1/system1 endpoint shaped like Jev's API, with Nimble (9B, Bespoke Labs) and TEV 1 (4B and 0.8B, Together AI), all Qwen 3.5 fine-tunes.
Evidence: 0 first-party, 1 hands-on, 0 relaying
Cognition says new SWE model nears frontier on internal benchmark at 60% lower cost
Cognition said its latest SWE model scores within a few points of leading frontier models on its internal Frontier Code benchmark at 60% lower cost, and described training it with asynchronous RL across about three continents, compressing weight updates about 99% and using top-p sampling to preserve entropy.
Evidence: 1 first-party, 0 hands-on, 0 relaying
Fireship relays Moonshot's reported $50 billion round and Hong Kong IPO preparation
Fireship said Moonshot AI closed a funding round at a $50 billion valuation, up from $4 billion last November, with China's state AI fund involved, and has started preparing a Hong Kong IPO.
Evidence: 0 first-party, 0 hands-on, 1 relaying
Ling 3.1 Flash free on Vercel until Oct. 13; Artificial Analysis index 41
Julian Goldie said Inclusion AI's Ling 3.1 Flash, a 560 billion-parameter mixture-of-experts model with about 25 billion active, released Sept. 30, is free to try in OpenCode, Kilo Code and Vercel's playground until Oct. 13.
Evidence: 0 first-party, 1 hands-on, 0 relaying
Google adds Gmail Live, Docs Live and Keep Live voice tools
Google launched Gmail Live, Docs Live and Keep Live voice tools built on Gemini audio models, Julian Goldie said, with Keep Live turning spoken notes and edit commands into notes and lists.
Evidence: 0 first-party, 0 hands-on, 1 relaying
Google launched Gemini Live Guided Vision for Android on Oct. 1, built with Aira
Google launched Gemini Live Guided Vision on compatible Android phones on Oct. 1, 2026, describing the camera view in real time, reading small text and telling users how to move the camera, Julian Goldie said in an Oct. 7 video.
Evidence: 0 first-party, 0 hands-on, 1 relaying
Gemini Deep Think coming to AI Pro subscribers, Goldie says
Julian Goldie said Google will make Gemini Deep Think, a parallel-reasoning mode for hard math, coding and science problems previously on a higher tier, available to Google AI Pro subscribers.
Evidence: 0 first-party, 0 hands-on, 1 relaying
Google's Antigravity update adds plugin store, plan mode and Windows sandbox without admin
Julian Goldie summarized a Google report of 10 Antigravity changes, including a customizations plugin tab, /plan review before agent action, @-mention messages to subagents, Windows file and network sandboxing without administrator rights, and offline SDK workflows with local models such as Gemma 4 26B A4B.
Evidence: 0 first-party, 0 hands-on, 1 relaying
OpenAI DevDay talk urges cost per task, adds prompt caching dashboard
OpenAI engineers said at DevDay that developers should optimize cost per task rather than per token, and announced a prompt caching dashboard and a cache-miss diagnostics API.
Evidence: 1 first-party, 0 hands-on, 0 relaying
OpenAI shows ChatGPT plugin sidebars, side panels and WebMCP site annotations
OpenAI videos showed ChatGPT plugin extensions with persistent sidebar homes and interactive side panels that pass selected context to ChatGPT, and custom site annotations in the ChatGPT app's browser that trigger WebMCP tools, adding detail to earlier descriptions of plugin extensions and WebMCP support.
Evidence: 1 first-party, 0 hands-on, 0 relaying
- Plus 58 smaller items from the same earlier days. (in the sources list)
Coverage
Window: videos uploaded in the 24 hours to Oct. 9, 02:14 UTC. Of 427 videos inventoried from monitored channels over the four-day scan, 221 were read for this edition (106 from the window and 115 late finds from earlier days), 148 had been covered in earlier editions, 8 had no transcript available, 22 were live or premiere videos deferred and 50 were under 45 seconds. The late finds are mostly videos that yesterday's collector left queued when it stopped on an access block; today's collection picked them up.
Sources reviewed (221 videos)
All 221 videos read for this edition, grouped by channel. Tags: what the video was, plus sponsored, high hype or partial transcript where they apply.
- AI Code King (2)
Paperclip: Run a Small AI Agent Team (Claude Code + Codex) Without the Chaos - tutorial, sponsored
Step 5 Preview (Tested): This MODEL BLEW MY BRAINS AWAY! - news + learning, sponsored, high hype - AI Engineer (25)
Agent Memory Is Solved. Agent Learning Isn't. — Karthik Ranganathan, Yugabyte - news + learning, sponsored
Agentic Search vs Vector Search for Coding Agents: We Ran the Eval — Braintrust - news + learning, sponsored
Lifestyles of the AI-Native — Nick Nisi & Zack Proser, WorkOS - tutorial, sponsored
Evals in AI: A Deep Dive — Tejas Kumar, IBM - news + learning, sponsored
How I Learned to Stop Worrying and Love the Sandbox — Matt Brockman, E2B - tutorial, sponsored
Build the Right Thing: Product Engineering (Part 1) — Kent C. Dodds, EpicProduct.engineer - tutorial, sponsored
Build the Right Thing: Product Engineering (Part 2) — Kent C. Dodds, EpicProduct.engineer - tutorial, sponsored
Harness Engineering: How to Build a Software Factory — Dru Knox, Tessl - learning, sponsored
The Self-Improving OSS Agent Stack — Marc Klingen, Langfuse - learning, sponsored
Why Building an Eval Platform Is Harder Than It Looks — Braintrust - learning, sponsored
Move Fast and Don't Break Things: Scaling Databases for the AI Era — PlanetScale - news + learning, sponsored
Is Speculative Decoding Worth It? Profiling vLLM on NVIDIA Blackwell — Akamai - news + learning, sponsored
The 6 Pillars of an Agentic Harness for Production — Varun Krovvidi, Resolve AI - learning, sponsored
Every AI Company Is Accidentally Building a Bank — Dor Sasson, Stigg - learning, sponsored
The Software Factory: From Bug Report to Production Code — Davis Palmie, Factory - learning, sponsored
Giving AI Agents Memory That Learns — Jake Broekhuizen, LangChain - news + learning, sponsored
Let Anyone at Your Company Ship Internal Apps with AI — Garrett Galow, WorkOS - news + learning, sponsored
Why We Deleted Our MCP Server and Rebuilt It — Abhi Arya, Reducto - learning, sponsored
Your AI Agent Has No Nervous System — Matt Gibiec, Dynatrace - promo only, sponsored
Turning Agent Memory Into Skills That Work — Will Lyon, Neo4j - news + learning, sponsored
We Mapped 115 Microservices for Our Coding Agents — Kamalakannan Nandagopal, Postman - news + learning, sponsored
Brains vs Hands: How to Run AI Agents Safely in Production — Viren Baraiya - news + learning, sponsored
How a Remote Company Builds AI Fluency — Em Shreve, Automattic - learning
Why 80% Reliability Isn't Good Enough — Felipe Blanes, Amazon AGI Lab - learning, sponsored
AI Writes More PRs. Who Validates Them? — Ali-Reza Adl-Tabatabai, Sonar - news, sponsored - Alex Finn (1)
Hark is a MIND BLOWING AI Agent you need to try - news, sponsored, high hype - Alibaba Cloud (1)
Dr. Feifei Li: From Tokens to Value: The Full-Stack Path | Apsara Conference 2026 - news, sponsored - All About AI (1)
Start Using AI Predictions To Beat The Market Today (beginner's guide) - tutorial, sponsored - Amazon Web Services (5)
How to add a recovery email to your AWS Builder ID - tutorial
How do I troubleshoot PostgreSQL performance problems on Amazon RDS or Aurora? - tutorial
Amazon Connect Customer - Omnichannel Customer Experience | Amazon Web Services - promo only, sponsored
Can agents run research and robots? - news + learning, sponsored
How do I use SMTP to send emails through Amazon SES? - tutorial - Arize AI (2)
How to turn agent traces and evals into EU AI Act evidence - news + learning, sponsored
Software Has Never Been Self-Improving, Until Now | Rise of the AI Engineer | Aparna Dhina - learning, sponsored - AssemblyAI (2)
Dear developers, SF Tech week begins, and let's start tomorrow with our Hardware Hackathon - promo only, sponsored
Workshop: Async, Sync, or Realtime: Which model do I choose? - news + learning, sponsored - AWS Developers (2)
AWS Free Tier and the New AWS Experience - news, sponsored
Build on AWS Without Worrying About Permissions - news, sponsored - Bijan Bowen (1)
Mistral Is BACK – Mistral Large 4 First Test (Le Chonk!) - news + learning - Claude (4)
Claude for Google Workspace™ - no news, sponsored
What's new in Claude: September recap - news, sponsored
Inside Shopify's AI strategy with CTO Mikhail Parakhin - news + learning, sponsored
Introducing Claude Dashboards and Claude Motion - news, sponsored - Databricks (3)
Como a YDUQS usa a Databricks para governança de dados na educação - promo only, sponsored
How to write data from Excel to Databricks - tutorial, sponsored
Como a Ipiranga utiliza IA na análise de vendas e pricing com Databricks - promo only, sponsored - David Ondrej (1)
The only moat in the age of AI is this - news - Elastic (1)
What is Zero Trust Architecture? Why is it important? [Quick Question Ep. 3] - no news, sponsored - ElevenLabs (2)
Introducing ElevenAgents Architect - news, sponsored
Eleven v4 is free for a limited time - news, sponsored - Every (1)
Introducing the Every Agent - news, sponsored - Fahd Mirza (4)
CodeAF with Ollama: Coding Agent Factory Tested Locally - news + learning, sponsored
Le chonk: The Mistral Large 4 Thoroughly Tested - news + learning, sponsored
d1-omni-600M Locally: A Decision Model for Audio, Vision, and Text - news + learning, sponsored
Humanizer 12B: Can Local AI Fix AI Slop? (Hands-On Test) - news + learning, sponsored - Fireship (1)
A $6.3 billion open-weight model just got embarrassed by the French... - news, sponsored - GitHub (1)
How to run GitHub Copilot agent sessions across any device | GitHub Copilot Day - news + learning, sponsored - Google Cloud Tech (5)
Coding for 40 tokens/sec on a Pixel 10 📱 - promo only, sponsored, high hype
Meet the founders of Astrocade: Build your own video game with just text! (No coding requi - news + learning, sponsored
BigQuery Graph measures explained (Why your AI agents get metrics wrong and how measures f - tutorial, sponsored
Scaling agentic coding practices across your team - news + learning, sponsored
Navigating AI automation in Looker: Essential developer tools for agents - tutorial, sponsored - Google for Developers (3)
Top 3 new model launches at Gemini Audio at Night - news, sponsored
📷 Searching for images from text with EmbeddingGemma 2 - news, sponsored
Building with Gemma 4 - promo only, sponsored - How I AI (2)
Codex built the largest user-created dungeon crawler - no news
AI picks my thumbnails now - news - Hugging Face (1)
Reachy Mini - Yo It's Your Boy (music video) - no news, sponsored - IBM Technology (2)
How AI Models Scale Beyond a Single GPU Across LLM Workloads - learning, sponsored
Hiding in plain sight: Fake GPTs, SMTP malware and NetScaler zero-days - news, sponsored - JetBrains (4)
.NET Day Online 2026 - news + learning, sponsored
(ReSharper + Roslyn) && You == Success - news, sponsored
JetBrains Air in IDEs – From Installation to Verified Agent Task in Two Minutes - news, sponsored
Air inside Rider: a live demo from .NET Day - news, sponsored - Julian Goldie (53)
New Hermes Kanban Agent Is CRAZY GOOD! - promo only, sponsored, high hype
NEW DeepSeek Harness v0.2 is INSANE! - news + learning, sponsored, high hype
Ling 3.1 Flash is INSANE (FREE!) 🤯 - news + learning, sponsored, high hype
Claude Code AI SEO Mods are CRAZY! 🤯 - tutorial, sponsored, high hype
Hermes Agent Just Got 500 FREE Plugins - tutorial, sponsored
Google Gemini NEW Updates are ABSURD! - news, sponsored, high hype
NEW Perplexity Computer Just Got Better! 🔥 - news, sponsored
GrokBot vs Dots vs Hermes Agent: Best AI Agent? - no news, sponsored
New Gemini Update Is CRAZY GOOD! - news, sponsored, high hype
Build and Automate Anything with Ling 3.1 Flash! 🔥 - news + learning, sponsored, high hype
StarNet Just Changed AI Agents Forever - news + learning, sponsored, high hype
Solar Mini 4: FREE Hermes Agent Model! - news + learning, sponsored
ChatGPT Voice Just Got Its Biggest Upgrade Yet - news + learning, sponsored, high hype
Run Claude + Codex For Free : Here's How! 🔥 - tutorial, sponsored, high hype
Google's AI SEO Guidelines Just Changed! - news, sponsored, high hype
NEW Lightpanda 1.0 Just Dropped! 🤯 - news, sponsored, high hype
Google Gemini NEW Updates are ABSURD! - news, sponsored, high hype
K2-Type-0.9B Just Dropped! 🤯 - news + learning, sponsored, high hype
This AI SEO System Got 27,800 Clicks in 90 Days - news + learning, sponsored, high hype
NEW Apodex Mini 1.1 is Crazy Good! 🤯 - news + learning, sponsored, high hype
NEW Google Nano Banana 2.1 is ABSURD!🤯 - news, sponsored, high hype
Google Stitch CLI Just Changed AI Coding! - news, sponsored, high hype
New Google Gemini Workspace Update Is CRAZY! - news, sponsored
Gemini Live Just Got CRAZY Powerful - news, sponsored
Google's AI Voice Update Just Changes Everything! - news, sponsored
New Antigravity Updates Are ABSURD! - news, sponsored
NEW Hark AI Agent is INSANE (FREE!) - news, sponsored, high hype
This NEW North 2 AI Super Agent is CRAZY! - news, sponsored
Obsidian 1.14 Just Got 5 HUGE Upgrades - news, sponsored
Turn Hermes Into Anything With ONE Click - news, sponsored
Perplexity Computer Built a Game for Just 97 Credits - news, sponsored
Herald OS: FREE Hermes Agent OS! - tutorial, sponsored
Starnet Just Changed AI Agents Forever - promo only, sponsored
GPT-6 Intelligent UI is INSANE! - news + learning, sponsored, high hype
Nano Banana 2.1 Is a HUGE Upgrade for AI Images - news + learning, sponsored, high hype
d1 Vision Just Beat GPT-6.1 Sol for 200X Less - news + learning, sponsored, high hype
This NEW Claude Update is ABSURD! - news + learning, sponsored, high hype
New Google EmbeddingGemma 2 is WILD! - news + learning, sponsored, high hype
Claude Haiku 5.5 Is 75% Cheaper And WAY Better - news + learning, sponsored, high hype
New Google Spam Update is INSANE! - news + learning, sponsored, high hype
Herald OS VS Hermes Agent: Who Wins? - tutorial, sponsored
How I Rank #1 in Google in 7 Hours - promo only, sponsored, high hype
GPT-6 Intelligent UI Is a HUGE ChatGPT Upgrade - news + learning, sponsored, high hype
OpenAI’s Decisions API Is 10X Faster Than GPT-6 Luna - news + learning, sponsored, high hype
Beam Just Dropped With 501B Parameters + 4X More Efficient - news + learning, sponsored, high hype
Google's Embedding Gemma 2 AI is Here! - news + learning, sponsored, high hype
NEW Google Playground AI is CRAZY! - tutorial, sponsored, high hype
New Hermes Index Is Absolutely INSANE! - news + learning, sponsored, high hype
NEW GLiDE Decision Model is WILD! 🤯 - news + learning, sponsored, high hype
Claude Code Just Changed The Agent Game - news + learning, sponsored, high hype
Google Gemini Updates are ABSOLUTE Game Changers! 🤯 - news, sponsored, high hype
NEW Nano Banana 2.1 is Absolutely INSANE! - news + learning, sponsored, high hype
NEW Kolibri 78B Just Dropped! 🤯 - news + learning, sponsored - Krish Naik (1)
Complete AI Engineering Roadmap - tutorial, sponsored - LangChain (4)
Add Web Search to Managed Deep Agents for free with One MCP Config - tutorial, sponsored
What is Deep Agents? An open source harness for long-running agents. - learning, sponsored
Welcome to LangChain Academy - promo only
OpenAI's Prompt Cache Has a Secret 15 RPS Ceiling - learning - Letta (1)
Letta Office Hours: New Agent Tools, Linear Workflows, and Memory - news + learning - Machine Learning Street Talk (2)
Game of Thrones: Turing Edition | Akarsh Kumar on Core War - learning
Life as it could be | Akarsh Kumar on Artificial Life - learning - Mastra (2)
Anthropic IPO, OpenAI DevDay Recap, And Why Are Decision Models Everywhere? - news, sponsored
Is Google finally back? - news, sponsored - Matt Pocock (1)
Don't underestimate the communication barrier - learning - Matt Williams (1)
Don't Make This Big Mistake with Jev - news + learning, sponsored - Matt Wolfe (1)
AI Design Hack Saves HOURS - promo only, sponsored - Matthew Berman (1)
Mistral is BACK! (Le Chonk) - news + learning, sponsored - Microsoft Azure (2)
Azure Speech LLM 2607: Better Multilingual Transcription in Microsoft Foundry - news
Real-Time Multichannel Speech-to-Text: Transcribe Stereo Audio in One Stream - news - Microsoft Developer (6)
Reduce Database Costs with Flexible Savings Plans - news, sponsored
Your Agent Is Only as Smart as Its Context - tutorial, sponsored
Turn Meeting Transcripts into Project Plans with GitHub Copilot Agents | MVP Unplugged - tutorial, sponsored
What's new in Microsoft SQL at SQLCon/FabCon Barcelona 2026 | Data Exposed - news
When my agent is slow or wrong, where do I even look? - tutorial
4X the Turns, 16X the Cost - no news - Microsoft Reactor (5)
Model Mondays - Spotlight On Fine-Tuned Models - news + learning, sponsored
Beyond Chatbots: Build Your First Production Agent with Microsoft Foundry - tutorial, sponsored
Introducción a ALM para Power Platform con Azure DevOps - tutorial
The Agent Harness: From GitHub Copilot to Production AI - news + learning
Microsoft Foundry Hosted Agents with Microsoft Agent Framework - tutorial - Microsoft Research (2)
What AI gets wrong and what failure teaches us - learning
Sparrow Studio Tutorial - Ep 1. - What is Sparrow Studio? - tutorial, sponsored - MLOps Community (1)
Inside a Game Where AI Agents Never Sleep - news + learning - MongoDB (1)
In-Car AI That Works Offline | MongoDB Atlas + ObjectBox Demo - learning, sponsored - n8n (1)
Which Model Should Power the n8n Assistant? - news + learning, sponsored - Nate B Jones (2)
Can Google still catch up? Argon is entering the chat #AI #Google #Argon #AInews #tech - news
Google Has More Data Than Almost Anyone. So Why Is It Bidding $10 Million On Old Emails? - news + learning - Nate Herk (4)
Anthropic Engineers Just 10x'd Everyone's Claude Code - news + learning, sponsored, high hype
Your Popup Can Now Optimize Itself for Every Visitor - promo only, sponsored
Grok Bot Just Got 2 Massive Upgrades. Do These Things Now. - news + learning, sponsored
Grok Bot uses my Claude Code subscription. Here's how. - tutorial, sponsored - Neo4j (1)
GraphSummit London 2026 - The Knowledge Layer for Trustworthy AI, Emil Eifrem - news + learning - NVIDIA Developer (4)
How open models are helping telcos - promo only, sponsored
What's New in CUDA 13.4 - news, sponsored
NVIDIA NV-Reason-CT: 3D Vision-Language Reasoning for CT Scans - news, sponsored
How H-Company Optimizes VLM Serving for Computer Use Agents With NVIDIA Dynamo - news + learning, sponsored - OpenAI (11)
Introducing the Decisions API - news
R&D Part 2 - promo only
Build custom site annotations with ChatGPT - news
Stack Overflow - From Idea to Product | DevDay 2026 - news + learning
Qodo - The Last Human Code Review | DevDay 2026 - news + learning
ClawLabs: Building Open Source Together | DevDay 2026 - news + learning
Stop Overpaying for Intelligence | DevDay 2026 - news
Build Plugins for ChatGPT - news
Introducing GPT-6 in ChatGPT with Intelligent UI - no news
How Oracle Uses ChatGPT Work to Transform Recruitment - promo only, sponsored
Production Monitoring with Codex: Grafana, Kubernetes, & Security - tutorial, sponsored - Peter Yang (1)
Epic rap battle between Lauren (Grok Bot) and Tibo (ChatGPT) - no news - Prompt Engineering (4)
DeepSeek Harness Just Got a Web Agent - news + learning, sponsored
Mistral Large 4 aka "le Chonk": Europe's Biggest Open Model - news
Training Your Own Embedding Model Is Not As Hard As You Think - news + learning
Embedding Gemma 2: The Only Embedding you need for RAG - news - Replit (1)
Building a Viral Growth Engine (Replit Pro Series) - no news, sponsored - Sam Witteveen (1)
Microsoft Joins the Local AI Push - news + learning - Sentdex (1)
Robot Chonk Cleans the Floor - news + learning - Snowflake Developers (1)
Snowpipe Streaming: Elastic Channels Demo - news, sponsored - The Pragmatic Engineer (2)
The state of the tech industry in 2026 - LDX3 New York - news + learning
Microservices: the architecture of last resort - learning - Theo - t3.gg (1)
I love Ultrafast (it's unusable) - news + learning, sponsored - Visual Studio Code (1)
⚡ Test your API without leaving VS Code #vscode #dotnet - tutorial - Vizuara (5)
It is time to start your own company - no news, sponsored
How AI Learns by Walking Downhill | Gradient Descent Explained - tutorial
Lecture 10 - DINO-WM from Scratch | Planning in Feature Space - learning
Can AI Create Art That Truly Moves You? An Open Research Problem - no news
Serve Your Own LLM: vLLM & SGLang, End-to-End - promo only, sponsored - Weights & Biases (19)
Fully Connected 2026 | Day 1 Full Replay - news, sponsored
Fully Connected 2026 Recap - promo only, sponsored
Fully Connected 2026 | Day 2 Full Replay - news + learning, sponsored
Why Reflection AI Is Making the Open Intelligence Bet - and What It Means for Deploying Mo - news + learning, sponsored
Fueling Innovation at Scale: Inside Pinterest’s Machine Learning Platform - news + learning, sponsored
Training Cognition SWE-1.7: Asynchronous RL at Global Scale - learning
Rapidly Deploying Useful General-Purpose Robots at Walden Robotics - news + learning, sponsored
AI Tokenomics Explained - news + learning, sponsored
Built Together: Inside the Dell + CoreWeave Engineering Partnership Powering Next-Gen AI - news, sponsored
Expanding AI: Near and Far - news, sponsored
NVIDIA Vera CPU: The Foundation for Agentic AI - news, sponsored
Data Architecture for AI: From Training Pipelines to Agent Memory - learning, sponsored
How Grammarly Scales Real-Time Inference with vLLM on CoreWeave - news + learning, sponsored
How Cline Embraces an Open-Weights Model Ecosystem for AI Agents - news + learning, sponsored
Beyond the Benchmark: How Parallel Web Systems Builds Reliable Web Agents - news + learning, sponsored
How Standard Kernel uses compound AI systems to optimize code and extract maximum GPU perf - news, sponsored
Accelerating Multi-Cloud Workloads with Google Cloud and CoreWeave - news, sponsored
Quota Isn't Enough: Topology Aware Scheduling on Modern GPU Clusters - learning, sponsored
Optimize capacity, control cost, and scale economically - news, sponsored