AI Models Will Automate Their Own Research | Ryan Williams
26 min
Listen through the publisher, then discuss the episode here or arrange an in-person Roundtable. Your city determines where to meet.
Roundtables are designed for groups of four, with a shared subject and room for everyone to speak. Come for discussion, with no dating expectation.
MIT Professor Ryan Williams breaks down the AI-generated proofs from OpenAI and Anthropic claiming to refute the exponential time hypothesis, explaining where the models succeeded via cross-literature synthesis and where a genuinely novel piece emerged. He also discusses how AI math breakthroughs translate to recursive self-improvement, predicting that labs are likely already using these capabilities internally to improve their own models, and shares why Navier-Stokes could be the first Millennium Prize problem to fall. MTS chronicles the singularity, live. Monitoring the situation of tech and AI. Follow MTS: Spotify |https://open.spotify.com/show/4HUNNmV1pjV7CW6Lphm1rM Apple |https://podcasts.apple.com/us/podcast/mts/id1891088763 – Brought to you by: Mongo DB - Anyone can prototype an agent. Lovable - Turn ideas into software people love with Lovable. Start today at https://lovable.dev/ Factory - A 24/7 system turning signals into production software. Visithttps://factory.com/ Neon (by Databricks) - Scale your startup on Neon. The Postgres backend for apps and agents. Built by Databricks.https://neon.com/mtshttps://neon.com/mts (https://neon.com/mts%E2%81%A0) Head over to http://getvcx.com (http://getvcx.com/) to learn more Kong - The AI connectivity platform. Larridin - Finally prove what your AI investment is worth. Ship 5x faster. https://blitzy.com (https://blitzy.com/) AdQuick - Make your brand a billboard. Clickhouse - The real-time database that keeps up with all of it. https://clickhouse.com/ Horizon3 - Don't assume you're secure. Prove it. Go hack yourself with AI. https://horizon3.ai/nodezero/ SRS Acquiom - Delivering shareholder rep, payments, and escrow solutions that streamline how deals get done. https://www.srsacquiom.com/mts – Timestamps: (00:00) Introduction (00:47) Exponential Time Hypothesis Explained (06:04) Inside OpenAI’s AI Math Proof (08:54) How Labs Should Release Breakthroughs (12:31) When AI Solves Millennium Problems (15:23) Why AI Solves Certain Math (19:05) AI Automating Its Own Research (21:47) Advice For Future Math Students –
Episode information is provided by the publisher. Listening takes place on the publisher’s service.
What NVIDIA is trying to build next… You're probably hearing a lot about AI, DeepSeek, NVIDIA and more right now. If you want the big picture (and to start from the beginning), watch this Huge Conversation with NVIDIA CEO Jensen Huang. In the last few years, NVIDIA has skyrocketed to become one of the world's most valuable companies. That's because, beginning in the 90s, they led a fundamental shift in how computers work, now unleashing the current explosion of what’s possible with technology. A huge amount of the most futuristic tech you’re hearing about - in AI, robotics, gaming, self-driving cars, breakthrough medical research - relies on new chips and software designed by him and his company. During the dozens of background interviews I did to prepare for this, what struck me most was how much Jensen Huang has already influenced all our lives over the last 30 years, and how many are saying it’s just the beginning of something even bigger… We all need to know what he’s building and
Launch, Scale & Monetize Your Newsletter with Beehiiv! Start today by visiting www.beehiiv.com?via=fellow (https://www.beehiiv.com/?via=fellow) Run Your Agency on Autopilot – White Label CRM, Funnels, and Automations in One Platform for Success. Claim 2-week free trail now www.gohighlevel.com/protrial?fp_ref=fg (https://www.gohighlevel.com/protrial?fp_ref=fg) Mark Tilbury firmly believes that anyone who learns the skill sets of a millionaire can become a millionaire, ANYONE. Mark Tilbury left school at 16 with no qualifications and no money. Now he runs a multi-million dollar business & has grossed over 50 million. He has the house of his dreams and most importantly, the freedom to spend time with his family. Now Mark wants to help you become financially free as well. Follow the podcast and turn on notifications! Follow @MarkTilbury on Tiktok, YouTube, Twitter & IG Disclaimer: All content rights belong to Mark Tilbury. This Podcast is fan-made. No copyright infringement intended. -----
Google’s Gemini Neon edged out Fable on a coding benchmark, but its best pitch may be as a second opinion. NLW shares how Neon helped him find the story in an oil episode that Claude and ChatGPT kept turning into a Wikipedia entry, and what happened when he pushed back on its ideas. In the headlines: Anthropic closes its first day of trading up 17%, and Chinese AI labs warn that new U.S. chip restrictions could delay model releases. LATEST EPISODE: How We Got Here - 'How GTA 6 Explains America" Brought to you by: Gartner - AI strategy, trends and insights - https://www.gartner.com/ai Granola - The AI notepad for people in back-to-back meetings. Try it free granola.ai/brief (http://granola.ai/brief) Newsletter: https://aidailybrief.beehiiv.com/ Interested in sponsoring the show? [email protected]
AI news from Anthropic's first day of trading, with Gemini Neon in Antigravity, Project Genie, Microsoft's MAI models, AI chips for China and Meta Muse. Anthropic priced its IPO at 83, opened near 111, and topped out above 114 before drifting back into the low 90s. Employees hold an estimated 232 billion dollars in stock, and Ashton Kutcher became a billionaire through his firm's early stake. Google stories included a hands-on account of Gemini Neon in Antigravity reviewing a whole project and then directing Claude Code as its product manager, a live Project Genie build of a South African safari, the Gemini Mac app's computer use, and Gemini for healthcare. Other topics were building an AI operations function from scratch at Verstela, gamified employee onboarding, Microsoft's Labs Canvas and MAI Code 2, House documents showing about 880,000 NVIDIA and AMD chips approved for export to China, and a Reuters report that Meta Muse surfaced private Signal messages through device permissions,
Kraken (https://kraken.pxf.io/c/6563010/687155/10583) Ledger (https://shop.ledger.com/?r=59fe6e05c254) NordVPN (https://go.nordvpn.net/aff_c?aff_id=143753&offer_id=612&url_id=14830) Spotify (https://open.spotify.com/show/4gu4MrqbGVZeNNNbAuvSmR) Follow Crypto News Today on X: @CryptoNewsHQ (https://x.com/CryptoNewsHQ) Published companion Deep Dive: AI's $81 Billion Explosion — Is Wall Street Abandoning Bitcoin? (https://cryptonewstoday.io/podcast/episodes/cefa139e-23c4-4424-b40a-238636763f2e) November 20, 2026: Bitcoin breaks below $60,000 after Anthropic's debut, but a completed IPO does not guarantee returning crypto buyers. Hong Kong publishes a broader licensing bill, Christine Lagarde warns about disconnected tokenized markets, and Wagyu's wrapped Monero token exposes the risk of buy orders left waiting through a crisis. Listening guide: Why Bitcoin ETF flows and the Bitcoin price can tell different stories (https://cryptonewstoday.io/podcast/guides/bitcoin-etf-inflows-and-price).
Palo Alto Networks has agreed to acquire the artificial intelligence startup Applied Compute in an all-stock transaction valued at $4.5 billion. Founded by former OpenAI researchers, Applied Compute specializes in helping businesses train and deploy customized open-weight AI models using their own proprietary data. This strategic purchase enables the cybersecurity giant to expand beyond traditional security offerings and embed protection directly into corporate AI infrastructure. Alongside the acquisition announcement, Palo Alto reported strong quarterly financial results that pushed its shares up significantly and led the company to raise its annual revenue forecast. Ultimately, the deal highlights a broader industry trend where major tech firms are rapidly absorbing specialized artificial intelligence capabilities to meet growing enterprise demand.
Malo Bourgon, CEO of the Machine Intelligence Research Institute (MIRI), explains why the organization pivoted away from technical alignment research toward communications and Washington, D.C. governance to buy time before superintelligence arrives. He argues that AI capabilities are scaling more predictably than fundamental alignment breakthroughs, leaving humanity unprepared if superintelligence arrives within one to five years, and contends that frontier labs will not reliably slow down on their own due to competitive pressures and deceptive alignment risks. -- MTS chronicles the singularity, live. Monitoring the situation of tech and AI. Follow MTS: Spotify |https://open.spotify.com/show/4HUNNmV1pjV7CW6Lphm1rM Apple |https://podcasts.apple.com/us/podcast/mts/id1891088763 – Brought to you by: Mongo DB - Anyone can prototype an agent. Lovable - Turn ideas into software people love with Lovable. Start today at https://lovable.dev/ Factory - A 24/7 system turning signals into productio
Asana Chief Product Officer Arnab Bose discusses how enterprises deploy human-agent teams in production, why organizational alignment is the primary bottleneck over raw model intelligence, and how self-driving organizations will materialize by 2027. He also shares concrete customer results from FedEx and Washmen and explains how Asana’s own R&D team is expanding rather than shrinking as AI accelerates product development. MTS chronicles the singularity, live. Monitoring the situation of tech and AI. Follow MTS: Spotify |https://open.spotify.com/show/4HUNNmV1pjV7CW6Lphm1rM Apple |https://podcasts.apple.com/us/podcast/mts/id1891088763 – Brought to you by: Mongo DB - Anyone can prototype an agent. Lovable - Turn ideas into software people love with Lovable. Start today at https://lovable.dev/ Factory - A 24/7 system turning signals into production software. Visithttps://factory.com/ Neon (by Databricks) - Scale your startup on Neon. The Postgres backend for apps and agents. Built by Datab
Kyle Chan, a fellow at the Brookings Institution and author of the newsletter High Capacity, breaks down the shifting US-China AI landscape. He discusses Trump’s APEC summit with Xi Jinping, China’s push for an indigenous semiconductor supply chain, Beijing’s pullback from open-source frontier models like Qwen 4 Max over cyber risks, and how Nvidia’s ecosystem will change without Jensen Huang. MTS chronicles the singularity, live. Monitoring the situation of tech and AI. Follow MTS: Spotify |https://open.spotify.com/show/4HUNNmV1pjV7CW6Lphm1rM Apple |https://podcasts.apple.com/us/podcast/mts/id1891088763 – Brought to you by: Mongo DB - Anyone can prototype an agent. Lovable - Turn ideas into software people love with Lovable. Start today at https://lovable.dev/ Factory - A 24/7 system turning signals into production software. Visithttps://factory.com/ Neon (by Databricks) - Scale your startup on Neon. The Postgres backend for apps and agents. Built by Databricks.https://neon.com/mtshtt
Waste-reduction policies should make responsible choices easy for households.
Share your view, then see how readers voted. No sign-in required. This poll is shared across stories.
Do you agree or disagree?
Common Quandary
A Common Quandary to discuss
Another conversation to explore, separate from this news story.
The restaurant gives free birthday cake. Is moving your birthday by a week close enough?
A restaurant offers a spectacular free dessert on birthdays. Your birthday was last week, but nobody took you out then. When the server asks whether anyone is celebrating, your friends point at you and wait for your answer.
How much flexibility can a harmless celebration tolerate before it becomes a lie?