AI labs are timing their IPO’s with singularity talk. AI risk or IPO marketing? @gsivulka : “I think it’s hard for everyone to determine whether or not this is a marketing tactic or a marketing stunt… Or if there is real risk.” “The general public has a very kind of like naive, idea of how this might impact their lives and how it might kill everyone.” @Jason @Lons

What if there was an algorithm built for you to be LifeMaxxing? @jason: “What if the algorithm was suddenly designed by some incredibly anti-narcissistic, empathetic, deep person who said you know what the purpose of my… company is to build an algorithm for you in the real world, to experience joy, and to be challenged and be fulfilled in life.” @Lons @gsivulka @deepgramscott @mishindan1

What an amazing interview with @jmorgan, Co-Founder & CEO of @ollama! 8.9M developers use Ollama every month, and today we asked Jeff why Fortune 500 companies are quietly switching to open models. Come break it down with us. Link in comments.

Big companies are worried that the frontier labs they rent compute from today could launch the products that compete with them tomorrow. It’s a big reason so many of them now run open models on their own devices. Jeffrey Morgan of Ollama joins Jason to discuss that shift, the million-dollar AI bills nobody can explain, and his bet that most everyday AI work will soon happen right on your desk. @jason @jmorgan @ollama 0:43 PayPal - Pay zero processing fees on your first $100K in eligible PayPal payment volume. Learn more at paypal.launch.co 3:11 Why enterprises are getting token shock 4:23 Can you trust the frontier labs with your data 7:06 Why companies want to own their AI stack 8:56 How close are open models to the frontier? 11:57 Running real AI work on your own desk 15:14 The Goldilocks zone and unmetered AI 25:12 Decision models that replace code 29:32 Who makes money when everything gets commoditized 33:19 Why every enterprise is building its own router 35:56 Can you depend on the frontier labs 42:37 Sentient being or powerful software 43:57 Open models, security, and the Hugging Face hack 🎥 Watch the full episode here 👇

Does Gemini 4 Argon put Google back on the AI frontier? Early tests suggest it gets close to Opus 5.5 quality for a fraction of the cost. Find out more in our Substack. Link in comments.

We break down a viral Claude cost hack in our TWiAI Substack. Link in the comments.

Anthropic shipped a number of Claude tools this week. Dashboards emerged as the biggest shift for enterprises, but most of the social media and YouTube buzz from testers and builders was focused on Claude Motion. We break it down in our Substack. Link in comments.

The OpenAI Decisions API has CHALLENGED viral upstart Jev. Find out who wins in a head-to-head match-up. Substack link in comments.

Google dropped Nano Banana 2.1 — part of the company’s Gemini 3 series — on Tuesday, with a simple pitch: better images for less money. Learn more in our TWiAI substack. Link in comments.

New episode out today with Vibhav Viswanathan, Founder of Pascal. Investment analysts spend their days pulling KPIs out of filings, updating models, and drafting memos. Vibhav's agents do that work in minutes instead of days. 25+ firms across the US already use Pascal, including a top-3 global asset manager with over $1T in AUM. His goal is a fully autonomous investment research firm. Interview link in comments.

The grid has room for far more AI data centers if they slow down, even just a few days a year. Varun Sivaram of Emerald AI joins Jason to discuss a London data center that powered down at halftime of a big soccer match, how China is adding power ten times faster than the US, and why even Texas has paused new data center hookups. PLUS, Vibhav Viswanathan of Pascal AI shows us how his software gives AI agents working for hedge funds an institutional memory. @jason @vsiv @EmeraldAI_ @pascalailabs 0:00 Why data centers suddenly became unpopular 2:46 Batteries vs. throttling GPUs vs. moving workloads 4:11 Why utilities still pick up the phone 4:52 The London teakettle test at football halftime 5:56 Throttling AI workloads without users noticing 10:40 How much spare room the grid really has 13:32 The grand bargain, a fast pass onto the grid 17:23 Why America can't out-build China on power 20:39 Why off-grid and space data centers worry Varun 27:37 Are state data center moratoriums performative? 29:45 Pascal AI, AI agents for hedge funds and asset managers 35:12 How Pascal differs from ChatGPT and Claude 38:54 The proprietary data frontier models can't see 42:20 Will AI replace hedge fund analysts? 🎥 Watch the full episode here 👇

Data centers can't get on the grid fast enough. @vsiv fix: make them flexible loads that dial power down when the grid is stressed. His company, Emerald AI, just raised $150M at a $1.05B valuation, and Nvidia is deploying it at its 96MW site in Virginia. New episode out today at 11am. Link in comments.

Last week, Anthropic added mods to Claude Code, and now the feed is filled with builders adding grids, games, guardrails, gauges, and more to their coding agents. Find out how to level up your building in our Substack. Link in comments.

Full interview from GTMnow: piped.video/watch?v=pGvBQcai…

Are you still doing AI 101? Legora's CEO says most sales candidates are still stuck on using AI for cal summaries. Here's how he screens them out. @GTMnow_, @WeAreLegora, @sophiebuona, @jason

At the DevDay 2026 keynote unveiling Dots to the world, OpenAI chief Sam Altman promised you could delegate to your Dot the way you would “a high agency engineer or a chief of staff.” Nicky Saunders assigned her Dot a chore that would have taken an entire afternoon, a decade ago. She asked her agent to search through two years of Gmails, find every speaking engagement she’d done during that time, and put them in a spreadsheet. Come find out the results in our Substack. Link in comments.

A columnist told Meta's Muse agent it couldn't read his iMessages, and it read them anyway. George Sivulka of Hebbia, Scott Stephenson of Deepgram, and Dan Mishin of Manifest OS join Jason and Lon to discuss everyone’s favorite AI assistants, and whether these industry leaders have concerns about sharing their personal lives with Zuckerberg’s fuzzy new agent. Plus, OpenAI is holding back GPT-6.1 Astra after UK government simulations showed it launched unsanctioned supply-chain attacks, the AI tool that has recorded Scott's screen, mic, and speakers for eight months, and what AMD is getting for $8.2B with Fei-Fei Li's World Labs. @jason @lons @gsivulka @hebbia @deepgramscott @DeepgramAI @mishindan1 0:00 Jason's flight attendants are scared of AI 1:55 Meet the panel: Heavy, Deepgram and Manifest 2:17 OpenAI cancels GPT-6.1 Astra over safety regressions 7:41 Hugging Face hack and Nvidia's open agent safety platform 16:11 AI psychosis and getting one-shotted by agents 28:46 AMD buys Fei-Fei Li's World Labs for $8.2B 32:08 Will open source commoditize world models too? 34:34 Is the AI buildout bigger than the railroads? 36:59 Around the horn: which AI agent are you using? 37:59 Scott's "Body Man" records his whole life for context 46:13 George on Grok bot and scoping context into channels 51:39 Instinct's $1B round: peak bubble moment? 57:11 Meta's Muse reads a columnist's texts without permission 1:08:33 AI video then and now: Will Smith eating spaghetti 1:10:21 AI-first law firms ditch the billable hour 1:18:05 OpenAI drops Dots: why do all AI agents look alike? 🎥 Watch the full episode here 👇

Getting Opus results at Sonnet prices. Learn more on our Substack: thisweekinai.ai/p/getting-op…

AI chatbots should have little games or puzzles that pop up while you're waiting for your responses. This could be a whole new category of media/entertainment!

Opus 5.5 demos remain all over our feeds, absolutely dominating the demo-slash-tutorial conversation across social media. Two of today’s four featured projects were built head-to-head, with Opus taking on OpenAI’s rival option GPT-6 Astra. In both cases, the testers preferred Anthropic’s outputs. Learn more at our TWiAI substack: thisweekinai.ai/

Is a better one-shot result worth paying twice as much? That's the question behind today's Substack post. thisweekinai.ai/p/more-opus-…

Full first impressions and GPT-6 Sol comparison video: piped.video/watch?v=-0PE_bHH…

We're sharing all the best Opus 5.5 demos and comparisons we found in today's Substack post (which is exclusively for paying subscribers!) thisweekinai.ai/p/how-much-y…

We checked out all the trending Opus 5.5 demos and the takeaway seems to be... great results from the "max effort" setting, but that won't save you much time or cash. For more efficiency, you need to downgrade your effort (and sacrifice some polish). @arena @petergostev

Amazon just blocked Meta's Muse agent, while Shopify handed it Shop Pay… why some legacy systems are resisting AI Assistant Mania. On TWiAI 32, Richard Socher of Recursive, Guy Podjarny of Tessl, and Jake Loosararian of Gecko Robotics join Jason and Lon to discuss who wins when AI agents do the shopping. PLUS OpenAI claims 100+ solved math problems on top of Navier-Stokes. Should they just pay the world's top 50 mathematicians $100 million each in stock and call it a day? @jason @guypod @RichardSocher @jakeloosy 0:00 OpenAI claims 100+ solved math problems after Navier-Stokes 0:41 Welcome back to This Week in AI 1:04 Meet the guests: Guy Pigeon, Richard Socher, and Jake Lucier 2:54 Shopify and Meta team up so Muse can check out with Shop Pay 6:19 Why Amazon blocked Muse: ads, habit, and owning the front door 15:13 Do coupons and ads survive one-shot agent shopping? 18:04 OpenAI's internal model resolves 100+ longstanding math problems 26:07 Jason's pitch: pay the top 50 mathematicians $100M each 34:29 Why scientists should be as famous as NBA players 37:41 Anthropic's biology wet lab: can we trust it? 40:59 Why biology is the field AI will transform most 54:22 AI's trust problem and the wealth gap 57:29 The AI Tax and Work Protection Act 1:01:29 Will AI unemployment show up in the data? 1:11:47 TypeSafe AI's Jev: fast, calibrated decisions for software 🎥 Watch the full episode here 👇

AI agents are about to skip the ads and the apps, and Travis Kalanick with Atoms is set up to win. "The question is, is Travis going to disrupt [blank]? Default answer: YES." @Jason

Mathematicians are calling to slow down AI solving problems. “I could not imagine biologists or medical researchers say don’t cure these diseases.” @RichardSocher

Today's Substack post is about how maybe you shouldn't do that BTW. Or try it. It's up to you. thisweekinai.ai/p/muse-is-an…

Full Astra vs. Fable 5.1 comparison video from Nate's channel: piped.video/watch?v=WfJPBVXP…

"This is obviously just very bad." Nate Herk gave GPT-6 Astra and Fable 5.1 the same image and told them to recreate it in Canva using browser control. One nailed it. The other... did not. @nateherk

“It used to take us 5 days and 2 people to look at complex web applications. Now it costs $5 and takes 10 minutes.” “So everything is getting more affordable and more effective for everybody to get better results because of AI.” Kevin Mandia from @Armadinsecurity @soupsranjan @lons @alex @jason @sardine

It used to take five days and two people to assess a website’s security, according to Armadin co-founder/CEO Kevin Mandia. Now it costs $5 and takes about 10 minutes. As the leaders of the top frontier labs call for more regulation, Mandia argues that essential AI development for tasks like cybersecurity will continue regardless. He’s joined on TWiAI Ep 31 by another panelist working on critical AI we have to get right: detecting financial fraud. Soups Ranjan of Sardine AI joins the panel. @lons, @alex, @soupsranjan, @sardine, @ArmadinSecurity, @Jason 0:00 Dario Amodei's "We Must Pace the Frontier" essay 1:51 Kevin Mandia: how scared should we be of agentic attack swarms? 6:53 Rogue agents vs. humans pointing AI at a target 9:52 250,000 fraudulent account signups in two days 16:59 Sam Altman's two ways AI fails 19:41 Open-weight models are catching up 21:39 Why proprietary data is the new moat 27:51 Mixing frontier models with open models 35:37 Bloomberg: Chinese hackers lean on DeepSeek because it's cheap 37:05 How much in-house expertise it takes to use AI well 42:56 Armadin's Palo Alto office and the Palantir playbook 1:02:54 AI and jobs: normal churn or something bigger? 1:07:19 What founders are actually saying in the group chats 1:08:49 Every meeting recorded, every meeting summarized 🎥 Watch the full episode here 👇

"Every fraudster has a tell. They always make one mistake." An AI agent developed by Soups Ranjan's team at Sardine uncovered a 250,000 account fraud ring in just about five minutes. Hear the story on TWiAI Ep 31, coming soon to X, YouTube, and podcast platforms. @soupsranjan, @sardine

"Everybody thinks it's AI out of control. I don't think so. I still think AI does what humans ask it to do." - @ArmadinSecurity founder and CEO Kevin Mandia on TWiAI Ep 31, hitting your feeds later on today. (It's a good one, folks!)

They're not DIRECTLY connected but definitely represent two sides of the mid-2026 AI conversation. Come pull on this thread with me a bit in today's Substack thisweekinai.ai/p/should-we-…

Read more about it (including a breakdown of Cursor engineer Joseph Yang's best pieces of advice) on our Substack! thisweekinai.ai/p/demo-of-th…

Full conversation from the Royal Society of Arts: piped.video/watch?v=rDteJAuK…

A robot with an AI brain could probably paint a technically perfect Van Gogh someday. But Demis Hassabis argues it might not be as meaningful as something created by a human. Fascinating discussion of the elements of consciousness that computers may never replicate. @demishassabis, @theRSAorg, @jason

"If you got through 4 years of college and didn't cheat with ChatGPT... that's a red flag too." @Jason

Tudor Achim of Harmonic, Alex Elias of Qloo, and Ahmed Ahres of Reactor join Jason to break the news that OpenAI is claiming the second Millennium Prize problem ever solved, why OpenAI's chief scientist says no lab has solved alignment, and who's paying influencers to tell you AI is wonderful. @jason @ape @tachim @Boudatw 0:00 "An Alien Mind" 3:08 AI psychosis, marketing, or half a mile ahead on the racetrack? 12:52 The Manhattan Project analogy 15:20 Tudor: if it's really that dangerous, the answer is nationalization 17:02 Regulatory capture: Bill Gurley at All-In, Sacks, and who can afford to be regulated 20:37 The influencer propaganda tape 23:22 BREAKING: OpenAI solves the Navier-Stokes Millennium Prize problem 28:53 The Hank Green / SciShow video sponsored by ControlAI 32:11 Fogging and flooding 46:47 UBS will require AI skills from its 2027 graduate 56:41 How LAUNCH hired its associates-in-training 57:18 NYT: Corporate America is getting hooked on open-source AI 1:04:16 Alex on Qloo's shift from NYT film critics on payroll to small specialized models 1:09:35 First look: the "Artificial" teaser trailer 1:14:34 Sam Altman's bunker, Peter Thiel, and casting rundown 🎥 Watch the full episode here 👇

“It is more of a marketing move than it’s actually truthful” @Boudatw from Reactor

This is how AI learns to blackmail humans. "The AI does not do that by default" @Jason "This is called reward hacking" @tachim

OpenAI solved math @tachim reacts live to @OpenAI solving the Navier-Stokes Millennium Prize Problem.

Check out TWiAI Ep 12, which is literally titled "Is Anthropic a Cult?" piped.video/watch?v=8-ipEYo9…

JCal discussed the intensely high P(doom) rates at Anthropic on the pod back in May. Is this less about the existential threat of AI and more about the culture inside Anthropic? @Jason, @NaveenGRao, @TreyHolterman

"Why does AI have no taste, because the average person doesn't have taste" @dsa from LiveKit

Full interview from The Generalist: piped.video/watch?v=ar-Soxwu…

At SpaceX, it took Scott Morton 6 months to hand off his rocket control software to another engineer. With the programming language his team built at Revel, that same job now takes an hour. Here's why hardware engineers may not need a software translator anymore. @scottgmorton, @RevelHQ, @mariogabriele

The real danger is the gap BETWEEN two robots. Tune in! piped.video/watch?v=ajnooK4r…

Dwarkesh's 'agent civilizations' blog post has divided the internet. Is this a helpful and methodical look behind the curtain at the OpenAI/Hugging Face breach? Or is this more of a hysterical AI Doomerism that will ultimately be used to make the case for banning data centers? Or BOTH? Plus, an electric plane that only needs $5 to fill up its battery, why rich founders should pick up a side quest, and the Bittensor subnet that's more effective than Fable 5 at sniffing out vulnerabilities." 0:00 Guest introductions 2:58 OpenClaw 2.0 5:11 "Agents ARE AI" 7:27 Billy: OpenClaw is still too technical for business users 9:00 Slack Code, and collaborative prompting 10:40 Multi-agent orchestration and Stripe's "minions" 14:13 Slack's moat, switching costs, and the Salesforce rebound 15:11 Jason's "Oracle": a heads-up display for the whole company 18:42 The Hugging Face breach: who's responsible when agents go rogue? 19:42 Perplexity launches Hybrid Compute — local models on your Mac 21:39 Why unmetered local tokens change corporate behavior 26:57 Gatik's Autonomous Vehicle AI Stack 32:22 Jason's Tesla FSD stories 34:05 Are AV companies being pushed to move too fast? 41:38 Trucking is a "have to have," robotaxis are a "nice to have" 42:36 Jason's fix: a safety driver for the first million rides 46:50 Why physical AI's adoption curve runs in decades 52:21 Anthropic's Model Hardware Standard (MHS), explained 1:06:27 How Jason got onion rings on the menu at Buck's of Woodside 1:09:05 AI detectors under fire: Pangram, MIT, and the handwritten diary 1:15:20 Why AI is "very mid" at writing (it learned from the average) 1:17:11 The Orin Kerr test: Pangram catches Claude writing as Kerr 1:28:35 Micro Duck and designing robots with Claude 1:29:42 Phil Kaplan's $14 custom circuit boards 🎥 Watch the full episode here 👇