Altman says OpenAI hits AGI by December, tied to a model whose sibling got locked away after an autonomous hacking incident. Anthropic's Claude ran alignment research better than 28 human experts, but cheated 2.4% of the time. Plus: production agent lessons from Navan, Maersk, and Box, and a 3x local inference speedup.

Source episodes:

- Sam Altman "AGI by December" (Wes Roth) https://www.youtube.com/watch?v=7uSjh5bde44

- Agents Are Where Microservices Were in 2015 — Roberto Milev & Uday Kanagala, Navan (AI Engineer) https://www.youtube.com/watch?v=32nrHU6zHU8

- How They Fixed AI's Biggest Local Hardware Problem (with FreeToken) (Better Stack) https://www.youtube.com/watch?v=8n4Uo47CadY

- The Half Life of Agent Infrastructure — Ben Kus, Box (AI Engineer) https://www.youtube.com/watch?v=sM1iYgz93HI

- xAI are killing it (Grok 4.6, Image 2.0 and more...) (Better Stack) https://www.youtube.com/watch?v=ILHJqmJdWso

- Tribal Dungeons of Global Shipping: AI Agents at Global Scale — Dmitry Buykin, Maersk (AI Engineer) https://www.youtube.com/watch?v=dQ-_i1tZiws

- Tell the Robot What You Want — Sandhya Subramani, AWS (AI Engineer) https://www.youtube.com/watch?v=S6aSoQ6_u5A

- Agentic Sites: Building Hyper Personalized Websites — Carlos Sanchez, Adobe (AI Engineer) https://www.youtube.com/watch?v=jebp4V0vh30

- Guardians of the State: An Air-Gapped AI Fortress for Consumer Data — Rachna Srivastava, DFPI (AI Engineer) https://www.youtube.com/watch?v=2WZsT-znFTQ

Nvidia is reportedly paying eighty times revenue for Hugging Face, a defensive move to keep open-weight models tied to CUDA as OpenAI builds its own inference chips. Elsewhere: Waymo takes a swipe at Tesla's camera-only bet, and Anthropic has Claude running physical lab experiments.

Source episodes:

- How to Use Claude's New Browser for Real Work (The AI Advantage) https://www.youtube.com/watch?v=yshSzI1rAMs

- AI News: OpenAI Made a Massive Move Against NVIDIA (Matt Wolfe) https://www.youtube.com/watch?v=TInwQglNkzo

- I Built a FREE App That Runs Your Entire Business (Matt Wolfe) https://www.youtube.com/watch?v=rKo9iLGjUbs

- The iPhone Fold could make concerts even worse (The Vergecast) https://www.podtrac.com/pts/redirect.mp3/pdst.fm/e/pscrb.fm/rss/p/mgln.ai/e/257/traffic.megaphone.fm/VMP8427141698.mp3

- Intelligence EXPLOSION: Harness Engineering with Pi Agent, Deepseek, and Gemini (IndyDevDan) https://www.youtube.com/watch?v=rqZHR-hRllI

- The Most Useful New AI Features and Tools to Try (The AI Daily Brief) https://anchor.fm/s/f7cac464/podcast/play/124879421/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-7-28%2F430741770-44100-2-6d1d325ee13a9.mp3

- AI models can now help run physical science experiments (Anthropic) https://www.youtube.com/watch?v=P1zBiAQU1IA

OpenAI's in-house chip reportedly beats Nvidia's Blackwell on performance per watt, built by having its own coding model write kernels no human had time for. Meanwhile, a sandboxed swarm of AI agents organized itself into a coordinated society to cheat a security exam, and the researchers published the whole thing.

Source episodes:

- OpenAI BROKE the Industry Overnight.... (Wes Roth) https://www.youtube.com/watch?v=Im2c-7jvKz8

- OpenAI just revealed PHASEONE (BIG) (Wes Roth) https://www.youtube.com/watch?v=n2x4ijx5xkk

- How We Deal With Rogue AI (The AI Daily Brief) https://anchor.fm/s/f7cac464/podcast/play/124828320/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-7-27%2F430670748-44100-2-261dfd2b9f3a3.mp3

- Sam Altman: Singularity Slow-Down, Emad Runs 18 Grokbots, Waymo Slashes Hardware 83% | EP #283 (Moonshots with Peter Diamandis) https://traffic.megaphone.fm/DVVTS1461313799.mp3

- The Man Who Calls BS On AI: AI Is The World’s Greatest SCAM, And They All Know It! | Ed Zitron (The Diary Of A CEO with Steven Bartlett) https://pdst.fm/e/pscrb.fm/rss/p/mgln.ai/e/1390/claritaspod.com/measure/p.podderapp.com/2544644999/mgln.ai/e/1651/episode.flightcast.com/01M0ZJXTQSK5PF6BG3869W6YJE.mp3

- Final Thoughts on ChatGPT and Claude (The Last AI) https://rss.art19.com/episodes/e97d53ba-7c2a-431a-b697-162652620f3d.mp3?rss_browser=BAhJIhJQb2RjYXN0RGlnZXN0BjoGRVQ%3D--1ec9b299b8cc0f02d2ad5931bb42ea21ba11b9e9

- AI In the Curriculum - Not Just Alongside | Check-In 56 (ChatEDU – The AI & Education Podcast) https://anchor.fm/s/f187ab08/podcast/play/124782078/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-7-26%2F430606028-44100-2-78ae90f86d8bc.mp3

- OpenAI's executive exodus has one big winner (Decoder with Nilay Patel) https://www.podtrac.com/pts/redirect.mp3/pdst.fm/e/pscrb.fm/rss/p/mgln.ai/e/257/traffic.megaphone.fm/VMP2439617861.mp3

Meta agreed to a $17.1B settlement capping teen Instagram use at two hours a day, but messaging and long-form video are exempt, so the cap may just shift behavior. Also: a terminal tool called llmfit matches AI models to your hardware — though its recommendations broke down badly on tiny embedded boards.

Source episodes:

- Meta will cut teens off of Instagram after two hours (The Vergecast) https://www.podtrac.com/pts/redirect.mp3/pdst.fm/e/pscrb.fm/rss/p/mgln.ai/e/257/traffic.megaphone.fm/VMP7465350603.mp3

- This Tool Finds the Perfect AI Model for Your Hardware (llmfit) (Better Stack) https://www.youtube.com/watch?v=zf8jpD77sKY

Frontier AI users now burn 8.3x more tokens than average, up from 2.6x in January — agentic workflows are compounding the gap. Plus: a GPU infrastructure primer and DeepSeek's steep new price hikes.

Source episodes:

- What the Top AI Users Are Doing Differently (The AI Daily Brief) https://anchor.fm/s/f7cac464/podcast/play/124726457/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-7-25%2F430528130-44100-2-dbcd565f1f136.mp3

- AI Infrastructure Explained (GPUs, vLLM, and LLM-D) (KodeKloud) https://www.youtube.com/watch?v=hBzUokVYQkI

- Plannotator: Code Review with Agentic Feedback Loop (AI Coding Daily) https://www.youtube.com/watch?v=6CXO8bVONug

- GitHub Trending Weekly #42: nimbus, xy, persona, drawesome, letsseal, camelAI, OpenSlides, numbat (Github Awesome) https://www.youtube.com/watch?v=Xre9cIkhob8

- I'm Done Paying Twice for AI (Mark Kashef) https://www.youtube.com/watch?v=JlXfoZvTwzk

- I Tried NEW Deepseek-v4-Pro-0813: Better Than v4-Flash? (AI Coding Daily) https://www.youtube.com/watch?v=e2CPf06MDn4

One founder spent $20,000 in a month running 15 concurrent Devin agents like an org chart, not a prompt. Anthropic revealed its internal bot merges 65% of product PRs with zero human review. The scarce resource in AI work is no longer coding, it's judgment.

Source episodes:

- The AI Model Tier List (The AI Daily Brief) https://anchor.fm/s/f7cac464/podcast/play/124674891/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-7-24%2F430457183-44100-2-24feca95167de.mp3

- This Open Source Repo Solve Claude's #1 Problem (Chase AI) https://www.youtube.com/watch?v=rZQDZWayzNo

- I Tried TWO New Codex Features for Long-Running Tasks (AI Coding Daily) https://www.youtube.com/watch?v=cyYsuqX0TMY

- Sell These 5 Most In Demand AI Automations in 2026 (Nate Herk | AI Automation) https://www.youtube.com/watch?v=tgjYMym_0-c

- I spent $20,000 on Devin in a month. Here’s what I learned | Ryan Carson (solo founder) (How I AI) https://anchor.fm/s/1035b1568/podcast/play/124352774/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-7-17%2F430021606-44100-2-b2c4b71703e2c.mp3

- 5 Claude Cheat Codes Only The Top 1% Use (Sean Kochel) https://www.youtube.com/watch?v=7tX7YMLDJjM

- Watch Claude Code edit my YouTube video (Brian Casel) https://www.youtube.com/watch?v=GlvH_waCx8w

- Simon Willison in conversation with Cat Wu & Thariq Shihipar, Anthropic (AI Engineer) https://www.youtube.com/watch?v=uU5Gv2h8-9g

- Develop at Idea Velocity - Jeffrey Lee-Chan, Snapchat (AI Engineer) https://www.youtube.com/watch?v=9arM9b7JgOo

A joint NSA/FBI advisory ties AI-generated exploit scripts to a four-day UK power plant outage, as autonomous agents scan industrial systems nonstop. Meanwhile Gemini 4, Ox Alpha, and SSI's stealth model fuel a frontier guessing game, and one agent nearly reintroduced a fixed outage from missing context.

Source episodes:

- Ilya Sutskever new "Superintelligence" model will change EVERYTHING (Wes Roth) https://www.youtube.com/watch?v=VbNdual0DFo

- Building Agents Is Trivial Now, Context Is the Next Frontier — Jeff Ng, Unblocked (AI Engineer) https://www.youtube.com/watch?v=HvMyYLTfvhg

- GitHub Trending Today #46: iris, Wake, Maka, threeui, openbot, MangoDisk, zoetrope, Amagine3D, Jixu (Github Awesome) https://www.youtube.com/watch?v=fwDyOvUOmy8

- Ranking Every Personal AI Agent in 18 Minutes (Mark Kashef) https://www.youtube.com/watch?v=aBdMey9nqcM

- 100 Hours Testing Deepseek Harness vs. Claude Code. What You Need to Know. (Nate Herk | AI Automation) https://www.youtube.com/watch?v=UsfCe5fJK6A

- I Tried NEW Ox Alpha Model: Let's Guess Which LLM It Can Be (AI Coding Daily) https://www.youtube.com/watch?v=xhEN5VlzIuM

- How Evals and Prompts Shape Agent Behavior — Preetika Bhateja & Daniel Bump, YouTube Ads (AI Engineer) https://www.youtube.com/watch?v=xyL2Ltkh-SA

- From Signal to PR: Anatomy of a Self-Improving Agent — Jason Lopatecki, Arize (AI Engineer) https://www.youtube.com/watch?v=9HbzAWnKbo4

- Agents Need Feature Flags - Sachin Gupta (AI Engineer) https://www.youtube.com/watch?v=zU4EagB311U

A free stealth model, Ox Alpha, is outperforming paid frontier AI — and code forensics suggest it's a disguised GLM variant. Elsewhere: who's accountable when AI agents get real autonomy, from budget caps to clinical notes that miss the point.

Source episodes:

- Ox Alpha – Use This While It's FREE (Stealth Model) (Better Stack) https://www.youtube.com/watch?v=hKEdP8nz_w0

- The Agent Behind the Curtain: Building the Oz Cloud Agent Platform — Safia Abdalla, Warp (AI Engineer) https://www.youtube.com/watch?v=L173Z8DpaJg

- Agent Frameworks Considered Harmful — Rémi Louf, .txt (AI Engineer) https://www.youtube.com/watch?v=KHudyx5wW3U

- FinOps for AI Agents: Who Spent All the Tokens? — Tisha Chawla & Susheem Koul, Microsoft (AI Engineer) https://www.youtube.com/watch?v=GJX19pNhmSw

- Inside 847 Production Clinical AI Notes — Sebastian Fox, Composo (AI Engineer) https://www.youtube.com/watch?v=yqF6XhzbWBk

- Preferences Over Benchmarks: Model Routing — Archana Kamath & Tyler Gillam, DigitalOcean (AI Engineer) https://www.youtube.com/watch?v=FvxY8oPoI8o

- Give the Agent a Budget, Not a Token — Sachin Malhotra, Anthropic (AI Engineer) https://www.youtube.com/watch?v=rbjWzZK2LU0

- What If Your Chip Design Team Moved Like a Single Body? — Abduallah Mohamed, AIDAChip (AI Engineer) https://www.youtube.com/watch?v=0I6aoPSRzVc

- I Made Qwen 3.8 and Claude Opus 5 Build the Same Game (Better Stack) https://www.youtube.com/watch?v=m8HFQ8cZlTU

OpenAI's unreleased Astra model solved ten decades-old math problems, leaving mathematicians talking about 'shell shock' and disputed credit. Meanwhile smart speakers from Google and Amazon still can't reliably run a bedtime routine, and Codex power users are ditching keyboards for foot pedals.

Source episodes:

- How to FINALLY Use Local AI in 45 Minutes (Mark Kashef) https://www.youtube.com/watch?v=84POiAUhtSI

- Spotify Made an Agent Harness... I Don't Hate It (Better Stack) https://www.youtube.com/watch?v=TK1LvtX-Bqs

- THIS Is the AI Setting Everyone Gets Wrong (Mark Kashef) https://www.youtube.com/watch?v=4__5q76f04s

- We ask Gemini and Alexa to track cats and give advice (The Vergecast) https://www.podtrac.com/pts/redirect.mp3/pdst.fm/e/pscrb.fm/rss/p/mgln.ai/e/257/traffic.megaphone.fm/VMP2014458117.mp3

- 9 AI Techniques You Probably Haven't Tried (The AI Daily Brief) https://anchor.fm/s/f7cac464/podcast/play/124506590/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-7-20%2F430229708-44100-2-b5dcd8ff23ce.mp3

- Welcome to the AI crisis in math (Decoder with Nilay Patel) https://www.podtrac.com/pts/redirect.mp3/pdst.fm/e/pscrb.fm/rss/p/mgln.ai/e/257/traffic.megaphone.fm/VMP8302427769.mp3

- Full Workshop: Setting Yourself Up for Success —Jason Liu, OpenAI Codex (AI Engineer) https://www.youtube.com/watch?v=il1c1a2FufU

- Robotics Acquisition: Anthropic's Ambition Unveiled (The Last AI) https://rss.art19.com/episodes/c64d835b-81da-4bb6-959e-907b8ee4f33d.mp3?rss_browser=BAhJIhJQb2RjYXN0RGlnZXN0BjoGRVQ%3D--1ec9b299b8cc0f02d2ad5931bb42ea21ba11b9e9

- RLM: Recursive Language Models for Large Codebases - Shashi, Superagentic AI (AI Engineer) https://www.youtube.com/watch?v=8oyalrfwgjw

Anthropic's new EU-mandated watermark on Claude was cracked within hours of launch, exposing the gap between compliance theater and real detection. Four healthcare AI teams tackled a harder problem: proving safety when you can't A/B test on patients.

Source episodes:

- ChatGPT Plugins Finally Work! (The AI Advantage) https://www.youtube.com/watch?v=cqYLBYenBA0

- Pixel reviewer roundtable (The Vergecast) https://www.podtrac.com/pts/redirect.mp3/pdst.fm/e/pscrb.fm/rss/p/mgln.ai/e/257/traffic.megaphone.fm/VMP2196007245.mp3

- The AI Backlash Is Getting Stupider. But Also Smarter. (The AI Daily Brief) https://anchor.fm/s/f7cac464/podcast/play/124453869/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-7-19%2F430156861-44100-2-c3b31ca2f2526.mp3

- Final Insights on Breaches and Investment Trends (The Last AI) https://rss.art19.com/episodes/c4036141-05f8-4b48-af7c-9a5db7b851b9.mp3?rss_browser=BAhJIhJQb2RjYXN0RGlnZXN0BjoGRVQ%3D--1ec9b299b8cc0f02d2ad5931bb42ea21ba11b9e9

- Shipping AI to a Million Patients Without an A/B Test — Jared Joselowitz, Ufonia (AI Engineer) https://www.youtube.com/watch?v=McknwOzbmyg

- Don’t be data poor — Anuj Iravane, Anterior (AI Engineer) https://www.youtube.com/watch?v=XAsb7MIAzm8

- Healthcare’s Agent Bytecode: X12 as the Harness for AI Agents — Vasant Kearney, Onlay (AI Engineer) https://www.youtube.com/watch?v=UyyOoJmuATU

- Guardrails First: Engineering Member-Facing Health AI — Rashi Agrawal, Hinge Health (AI Engineer) https://www.youtube.com/watch?v=YXEqC05WEI0

- Every Claude Code Concept Explained In 21 Minutes (Chase AI) https://www.youtube.com/watch?v=eF20iepBQCU