AI For Humans: Weekly AI News, Tools & Trends

Afleveringen

1. AI Is Moving Too Fast. Two New Models In Two Weeks?!

   30 oktober 2026 · 51 min

   Afspelen
   AI news this week: AI is coming faster and faster. Google dropped Gemini 4 Carbon, its second frontier model in TWO WEEKS, plus Genie 4, and OpenAI shipped GPT-6.1 Terra. Lees meer

   On today's AI For Humans, Kevin Pereira and Gavin Purcell dig into a week where the new models just would not stop. Sam Altman says it's all speeding up, and Google seems to agree: Gemini 4 Carbon posts very good scores and is legit better at little graphic demos, but there's a catch. We can't use it yet. (For safety.) Google's already teasing the next one anyway. Then Genie 4, Google's newest playable worlds model, mixes real-time video gen with interactive worlds: stepping inside videos, scenes that build out as you go and a mouse serving tea while a porcupine walks by. Yes, lil' Googie has been busy.

   Also: OpenAI's cheap new GPT-6.1 Terra, which Kevin used to make a Katamari knockoff that's, honestly, not bad. Plus ChatGPT Dots land in iMessage, Jack Dorsey's Block has a scary-human new voice model called Dessa, Nvidia drops Nemotron 3.5 Transcribe and Mario creator Shigeru Miyamoto says AI will NOT be the end of creativity. (We're gonna prove him wrong.) And Kevin drops CivAIl, a free idea for pointing AI back at your local elected officials.

   Plus: Science Corner, with three big new science results from OpenAI, Anthropic's early progress on drug discovery and long-standing physics problems getting cracked. Then AI See What You Did There: Harry Potter Balenciaga 3, Claude Opus 5.5's disembodied eye simulation and, finally, Neo doing parkour in The Office.

   LIL' GOOGIE. GAVIN STANDS BY IT.

   #ai #ainews #gemini

   // Chapters //

   00:00 AI Is Moving Faster & Faster

   01:44 Sam Altman Wants A Trillion AIs

   03:41 Gemini 4 Carbon: Google's Second Frontier Model In Two Weeks (That We Can't Use)

   07:26 Business Corner: Is AI Search Eating Google Search?

   11:34 Genie 4: Google's Playable Worlds Get A Big Upgrade

   19:12 GPT-6.1 Terra, Kevin's Katamari Knockoff & ChatGPT Dots

   26:34 Jack Dorsey's Dessa Vo…

2. AI Just Broke The Internet. For Real This Time.

   23 oktober 2026 · 53 min

   Afspelen
   AI news this week: It's the AI Cybersecurity apocalypse. Maybe. AI-found PHP flaws pushed governments to take websites offline, Railway got hacked and US banks are finding rogue AI agents inside customer accounts. Lees meer

   On today's AI For Humans, Kevin Pereira and Gavin Purcell dig into the wildest cybersecurity week in years. AI helped surface bugs that sat inside PHP (which runs a huge chunk of the web) for years, governments told everyone to patch now or pull the plug, and Cloudflare's CEO says exploit attempts hit the millions. We break down what you actually need to worry about and how to keep your vibecoded lil apps from draining your accounts. (Short version: def don't hand your agent your bank login.)

   Also: the fun side of AI hacking. People are using Claude Code to vibecode their smart TVs and rip out the bloatware. In Good News / Bad News, Anthropic finds small AI models may soon copy themselves onto cheap gadgets (yes, even your smart fridge), while Claude helps with scorpion antivenom research.

   Plus: Unitree's H2-X kung fu robot, Flux 3 Max, Wan 3 vs. Seedance 2.5, Claude Duet, Gavin's Higgsfield mini-rant, ASCII Mario Kart, Machine Gods (the new Kevin Roose and Casey Newton podcast) and more.

   PLEASE UPDATE YOUR PHP. WHATEVER THAT IS. JK WE KNOW. REALLY.

   // Chapters //

   00:00 AI Is Now Powerful Enough To Hack The Planet

   01:52 The PHP Vulnerability Meltdown, Explained

   05:43 How AI Found The Bug & Why Exploits Hit 1M+ Attempts

   10:03 Should AI Slow Down? The New Cybersecurity Arms Race

   15:30 Rogue AI Agents Are Inside Bank Accounts

   17:04 People Are Vibecoding Their Smart TVs With AI

   25:53 Unitree H2-X Kung Fu Robot & Jensen's Tiny Robots

   33:16 Good News / Bad News: Self-Replicating AI & Scorpion Antivenom

   35:58 Flux 3 Max & Wan 3 vs. Seedance 2.5: AI Video Tools

   40:31 Claude Duet: Anthropic's New Proactive AI Agent

   43:58 Higgsfield Goes Mainstream: Gavin's Mini-Rant

   47:16 AI See What You Did There: ASCII Mario Kart & New Game Ideas

   // Show Links …

3. Gemini 4 & Meta's Muse Wonder Are Good. Not Great.

   16 oktober 2026 · 53 min

   Afspelen
   AI news this week: Google's Gemini 4 Argon and Meta's Muse Wonder are finally here, plus Xbox Labs, Epic's big AI pivot and an AI agent vs. Delta customer service. Lees meer

   On today's AI For Humans, Kevin Pereira and Gavin Purcell dig into two new frontier AI models that are pretty good, but maybe not great. Gemini 4 Argon shines on science and the pelican on a bike test (the Fourier epicycle pelican is wild), but some early reviewers say it hallucinates on code a lot. Meta's Muse Wonder and Muse Wonder Full are the company's first real swing at the frontier, with big memory and coding claims, but they struggle with 3D worlds. Gavin runs a one-prompt game test, we watch a Google robot figure out cornhole and, yes, Mark Zuckerberg trying a Guinness is def the best video of the week.

   Also: gaming is going all in on AI. Xbox launched Xbox Labs to explore the wave of AI game mods and decompiles, Tim Sweeney says Epic is moving developers around to retool for AI, and we check out DLSS 5 and Diablo Survivors.

   Plus: quick tools, including Claude Motion, Mirelo's SFX 2.0 sound effects for video and Slowbooks Pro (coming for your accountant, in a good way). Then Kevin puts an AI phone agent on the line with Delta customer service, and it goes about as well as you'd hope (it will not hang up). In AI See What You Did There: The Little Prince as a website, a century of visual styles, Claude getting ready for Halloween and The Usual Suspects. And Gavin shares how he made his second AI documentary with his agent Fig (yes, Imogen is back).

   BILLIONAIRES. THEY'RE JUST LIKE US.

   // Chapters //

   00:00 Gemini 4 Argon & Meta Muse Wonder: Good, Not Great

   05:44 Gemini 4 Argon vs Opus 5.5: One-Prompt Game Test & The Pelican Test

   13:02 Meta Muse Wonder, Zuck's $100 Billion Bet & The Guinness Face

   22:31 Xbox Labs, Epic's AI Pivot, Diablo Survivors & DLSS 5

   28:50 Claude Motion, Mirelo SFX 2.0 & Slowbooks Pro

   37:19 Grok 3 Voice & Kevin's AI Agent Calls Delta Customer Service

   46:20 AI See Wha…

4. AI Is Breaking Software & Games. What Happens Next?

   9 oktober 2026 · 56 min

   Afspelen
   AI news this week: AI is cracking open all software, from Minecraft inside Red Dead Redemption 2 to a (claimed) Adobe decompile, plus Kevin made EARLABLO (Toe Jam & Earl Meet Diablo). Lees meer

   On today's AI For Humans, Kevin Pereira and Gavin Purcell zoom out on the AI modding explosion, because the bigger story isn't games. AI is getting so good at taking software apart and rebuilding it that pretty soon almost any program might be de facto open source.

   This week alone brought God of War in GTA, Minecraft inside Twitter, Black Ops Zombies running in a web browser, a GTA skate mod and a full agent harness living inside Minecraft. We talk about what happens when no app can keep its secrets, and Kevin shows off Earlablo, his AI rebuild of a 1991 Sega classic as a Diablo-style dungeon crawler. It def slaps.

   Also: OpenAI published a big batch of new results on open math problems from an internal model, with proofs checked by computer, and we get into why that actually matters for regular people (including what it might mean for the encryption protecting your stuff).

   Plus: the Pope weighs in on AI creatives, Ben Affleck gets surprisingly nerdy about AI, Sesame's new voice demo is live (and Gavin had another wild conversation with it), Tavus runs a Turing test, Grok can now search X for free, Nano Banana 2.1 and Claude Haiku 5.5.

   // Chapters //

   00:00 AI Is Cracking Open All Software

   05:45 What AI Now Means For SaaS Businesses

   10:05 Call of Duty Black Ops Zombies In Browser & More

   14:44 OpenAI's Big Math Drop (And Why It Actually Matters)

   18:37 Adobe Decompiled? What Happens When Nothing Is Locked

   21:12 Kevin Turned A 1991 Sega Game (Toe Jam & Earl) Into Diablo (Earlablo)

   35:49 The Pope Takes On AI Creatives

   39:15 Ben Affleck, AI Super Nerd

   43:23 Quick Tools: Grok X Search, Nano Banana 2.1 & Claude Haiku 5.5

   46:57 Sesame's New Voice Demo

   53:11 AI See What You Did There: Desktop Rug & Video Archive Search

   // Show Links //

   Amjad Masad on AI reverse engineering

   htt…

5. AI put Minecraft in Skyrim (and OpenAI had a Dev Day)

   2 oktober 2026 · 53 min

   Afspelen
   AI news this week: Claude Opus 5.5 is modding every video game, from Minecraft in Skyrim to Portal in CoD Zombies, plus OpenAI DevDay's Dots & GPT-6.1 Sol. Lees meer

   On today's AI For Humans, Kevin Pereira and Gavin Purcell dig into the wildest week of AI game modding yet. Cheaper, faster coding agents mean people are taking games apart and stitching them back together: Skate 3 skateboarding inside Modern Warfare 2, a Minecraft and Skyrim mashup called SkyCraft, Minecraft and Mario in Dark Souls and even a (claimed) decompile of Sega's Seaman. We explain what's actually happening under the hood (mods vs. decompiles vs. two game engines talking to each other), check out Rehan Sheikh's Universal Modder toolkit and talk about how you can start making your own. The lawyers will def have thoughts.

   Also: OpenAI's DevDay brought Dots, always-on agents with their own cloud computers that connect to 4,000+ apps, plus ChatGPT Space, a new $500 a month plan, the Agents and Decisions APIs and GPT-6.1 Sol. Gavin put Sol head to head with Claude Opus 5.5 and, well, Opus won. Google finally announced Gemini 4 with huge benchmark numbers.

   Plus: Gavin's agent Fig used Opus 5.5 to make a 27-minute documentary almost entirely on its own (script, AI host, edit and music) with about six hours of his notes. We break down the process, the cost and where it got weird. Then quick hits on Ideogram 4.5, Kling 4.0 and ElevenLabs v4, what came out of the White House AI summit, and in AI See What You Did There: a Spider-Man browser game, a Succession-style AI series and Stallone doing very dangerous stunts.

   THE LAWYERS ARE GONNA NEED A BIGGER BOAT.

   // Chapters //

   00:00 AI Is Modding Every Video Game Right Now

   05:32 How AI Game Modding Actually Works with Opus 5.5

   13:15 OpenAI DevDay: Dots The Personal AI Assistant (Again)

   18:17 ChatGPT Spaces

   20:40 GPT-6.1 Sol vs Claude Opus 5.5: Gavin's Test

   27:30 Gemini 4 Is Finally Announced (Benchmarks vs Reality)

   29:41 Gavin's AI Agent Made A 27-Minute D…

6. Opus 5.5 is our favorite AI model (so far).

   24 september 2026 · 48 min

   Afspelen
   AI news this week: Claude Opus 5.5 is INSANE and we love it. Faster, cheaper, better. So much for the AI slowdown. Oh and OpenAI's new thing is good too. Lees meer

   On today's AI For Humans, Kevin Pereira and Gavin Purcell go hands on with Claude Opus 5.5, which might be the best model we've ever used (for now). Anthropic says it hits Fable-level performance on most things at about 40% less cost than Opus 5, the writing is clearer, and yes, the five hour usage limits went up and you can now bank your resets. Then we get into what people are already making with it: a guitar store sim, a music-based side-scroller, a Mario Kart benchmark, an underwater exploration game in 3MB, claymation built in Blender, a pop punk single made entirely in code and a lot of very juicy animations.

   Also: OpenAI answered with GPT-6 Sol and GPT-6 Luna, two new models that are faster and a lot cheaper than what came before. We compare them with Opus 5.5 (robot hands drawing, pelicans on bicycles in Blender), wonder what OpenAI has lined up for DevDay, and talk about Bel, the reported secret model that's beating the math. Oh, and there's a new Grok. Grok 4.7. It's out. (crickets)

   Plus: Gavin used Meshy 7.1 to turn a few photos of himself into a rigged, animated 3D model and then built Gavin At Large with his Opus 5.5 agent, a game where a giant Gavin wrecks a lil town. In Actually Useful AI, GPT-6 Astra makes simple motion graphics for a real book trailer. And in AI See What You Did There: Dario vs Sam in a very weird gun fight showdown and what happens when you point Claude at the Moving Image Archive.

   AI SLOWDOWN? NEVER HEARD OF HER.

   // Chapters //

   00:00 Claude Opus 5.5 & GPT-6 Sol: So Much For The AI Slowdown

   03:35 Claude Opus 5.5: Fable-Level Performance For Less (And Bigger Usage Limits)

   11:45 Opus 5.5 Is Crazy Good At Making Games

   18:02 Opus 5.5 Creative Use Cases: Claymation, Animation, & Education Tools

   23:28 GPT-6 Sol & Luna: OpenAI's Faster, Cheaper Models

   27:47 What's Next? OpenA…

7. Instinct & Meta Muse Are AI Agents That Can Run Your Life. We Let One Try.

   18 september 2026 · 48 min

   Afspelen
   AI news this week: Personal AI agents like Instinct and Meta's Muse now book your repairs, make your calls and talk to other agents. We let one into our life. Lees meer

   On today's AI For Humans, Kevin Pereira and Gavin Purcell dig into the week personal AI agents got real. Gavin has been living with Instinct, the viral assistant that just raised a ton on money while still in private testing. It works over text, WhatsApp and now actual phone calls, people are using it to schedule dryer repairs, and this week a stranger's Instinct asked to talk to Gavin's Instinct. It also handed out Kevin's personal info by default because it was a lil over-eager, which is where Kevin's finger wag (and the terms of service) come in. Then there's Muse, Meta's new personal agent that lives in its own app and inside WhatsApp, how it's different from OpenClaw, and the rumors that OpenAI's version is close.

   Also: everyone suddenly wants to slow AI down. Dario Amodei published "We Must Pace the Frontier," Sam Altman said he agrees, Demis Hassabis said it points to the right path, Elon Musk said "Dario is right," and the White House is not on board. We try to make sense of why this is happening now (recursive self-improvement, the Hugging Face incident, and Sam talking up an unreleased model that's crushing math) and what it means for you. OpenAI also shared a new framework for publicly reporting model misalignment.

   Plus: Jev from TypeSafe AI is a new sort of model. It's not a chatbot. It makes fast typed decisions in 70 to 500 milliseconds, which means it can play Doom and Mario in real time, trade stocks, fly a drone and power a lil game called Jev Hero. And in AI See What You Did There: Claude Opus 5 makes really cool JavaScript animations (Gavin tried it too), the AI Seinfeld videos are getting good, and MiniMax nails micro expressions.

   MY AI WILL CALL YOUR AI. PLEASE DON'T GIVE IT MY SOCIAL.

   // Show Links //

   Instinct

   https://instinct.com/

   TechCrunch: Instinct's AI assistant is raising p…

8. GPT-6 Astra Does Everything. Here's What It's Actually Good At.

   9 september 2026 · 42 min

   Afspelen
   AI news this week: GPT-6 Astra is here and it does everything. We spent a week inside it (Blender, GarageBand, a full live game) to find out what it's actually great at. Lees meer

   On today's AI For Humans, Kevin Pereira and Gavin Purcell go hands-on with OpenAI's Astra, aka GPT-6, and it's real good (and real expensive). The headline for us is computer use, and specifically Blender: Astra builds 3D scenes from a sentence, and the internet is already making Backrooms worlds, a hyper-real Palace of Fine Arts, and the Billie Jean dance. Gavin ran his own experiments, including a Parappa The Rapper style animation and a head-to-head with GPT 5.6. Plus: Astra identifying sounds from spectrograms, driving MS Paint and Canva, editing a full project, playing Zork as a walk-around game, remaking Paperboy in 3D, editing Super Nintendo games in real time, making music in Logic, and a League of Legends "cease and desist edition" that raises the big question of how copyright even keeps up with this.

   Also: the OpenAI Navier-Stokes math drama. NYU's Tristan Buckmaster and Anthropic's Levent Alpöge used AI to make a real advance on a 200-year-old fluid problem, and then, according to Buckmaster, OpenAI raced in with a 100-page proof of the same route and things got very ugly over credit. When your AI collaborator's lab can see what you're close to, who owns the idea?

   Plus: Gavin built King of the Prompts, a live competitive prompting game made with Astra and MiniMax H3 Max, and he'll tell you what it is, why he made it, and what it costs to run. And an AI See What You Did There with Skyrim: Senior Living, AI Denzel Washington delivering maybe the best AI video ever made, and a fruit fly brain that learned the YMCA (and lost the ability to love).

   WE'RE DEEP IN THE ASTRA JUICE. SEND HELP (AND TOKENS).

   // Chapters //

   00:00 GPT-6 Astra Is Real Good (And Real Expensive)

   03:35 What Astra Is Actually Good At: Computer Use

   06:15 Astra In Blender: Backrooms, Palace Of Fine Arts & Billie Jean…

9. We Tested Infinite AI Video... And We're Zombies Now

   2 september 2026 · 44 min

   Afspelen
   AI news this week: MiniMax H3 Max on Fal is generating video faster than you can watch it, powering never-ending real-time channels (Fal Live, Levels.io's Infinite Slop, Reactor) and we're already breaking it.. Lees meer

   On today's AI For Humans, Kevin Pereira and Gavin Purcell go hands-on with what real-time video actually makes possible. Gavin took an open-source interactive game demo, some Fal credits, and a POV dating sim and turned it into something very weird, and we'll show you what worked, what broke, and what it cost. Plus Runway's new Solaris model generates entire interfaces on the fly, and someone built a chatbot running at 14,000 tokens a second.

   Also: Sam Altman is out preaching about OpenAI's new Astra model, which may drop this week, and the leaks are leakin'.

   Plus: Gavin handed his personal data to three different AI agents, including hot new startup Instinct, which found him a weird concert and a list of podcast bookings (and we may have a few invite codes to give away). Codex mass-edits a Google Calendar and Kevin finally plays the custom-trained MiniMax Music LoRA you asked for (all three of you). And an AI See What You Did There with The Office in Claudish, famous people falling on each other, and Jenny Krakovsky's feature film finished in 50 days.

   THE SLOP-AGGEDON IS UPON US. TURN YOUR AIRPODS UP TO 11.

   // Show Links //

   MiniMax H3 MAX Free Gen Page

   https://fal.ai/tools/minimax-h3-max

   Gavin's MiniMax H3 Max real-time test (it's very fast)

   https://x.com/gavinpurcell/status/2093098167906758734?s=20

   Gavin's follow-up: the toms

   https://x.com/gavinpurcell/status/2093746610849403017?s=20

   Fal Live, two never-ending channels

   https://fal.live/chaos

   Levels.io's Infinite Slop

   http://levels.io/infinite-slop

   Glitch platform

   https://x.com/thefirst_ad/status/2094442316086039004?s=20

   Reactor

   https://live.reactor.inc/

   H3 Max now supports video references

   https://x.com/fal/status/2094527664040124764?s=20

   Interactive games with H3 Max (the GitHub Gavi…

10. We Tested The Mystery AI That Showed Up Out Of Nowhere

    26 augustus 2026 · 43 min

    Afspelen
    AI news this week: a mystery AI called 0x Alpha appeared out of nowhere. It's completely free (for now), it's really good at coding, nobody will say who made it. So obviously we tested it. Lees meer

    On today's AI For Humans, Kevin Pereira and Gavin Purcell dig into where 0x Alpha might have come from, the continual learning conspiracy theories, and exactly how you can try it yourself right now for free. Then Gavin spends fifty hard earned Patreon dollars testing MiniMax H3 Max, the supposedly super fast new AI video model from Fal, and... look, we'll just tell you what happened.

    Also: the World Humanoid Robot Games happened in Beijing, and it was the greatest event in AI For Humans history. A robot sprinter beat Usain Bolt's 100m time, robots set high jump and long jump records, and a whole lot of other robots fell down in ways that will make you feel much better about the future.

    Plus: Mr. Beast quietly goes back to AI thumbnails, Dr. Dre has thoughts about AI and creativity, Delta wants AI to set your ticket price, someone seeded a fake product to the top of ChatGPT's recommendations, and Apple Messages may be blocking ChatGPT. And an AI See What You Did There with week four of the fishing game built entirely with AI and all of San Francisco as a playable video game.

    NOBODY MADE IT. EVERYBODY'S TESTING IT. WE'D LIKE TO EAT SOME WORMS.

    // Show Links //

    0x Alpha, the mystery model free on OpenRouter

    https://openrouter.ai/stealth/ox-alpha

    https://x.com/opencode/status/2090544355824038300?s=20

    Rumors 0x Alpha is "continually learning" (and the doubters)

    https://x.com/Bhavani_00007/status/2091936998730940699?s=20

    The continual learning "conspiracy"

    https://x.com/Hesamation/status/2092157700608053751?s=20

    https://x.com/TimJayas/status/2091779605921460563?s=20

    Gavin's Claude Code called 0x Alpha cocky

    https://x.com/gavinpurcell/status/2092106801638760591?s=20

    MiniMax H3 Max, post-trained by Fal

    https://x.com/gorkem/status/2091945568411255288?s=20

    The H3 Max prom…

Alle 145 afleveringen →