Last Week in AI

Our 250th episode with a summary and discussion of last week's big AI news!
Recorded on 06/27/2026
Note from Andrey: sorry this is late again! this episode release somehow didn't save and I only realized late, my bad... next one will be out way sooner!
Hosted by Andrey Kurenkov and Jeremie Harris
Feel free to email us your questions and feedback at andreyvkurenkov@gmail.com and/or hello@gladstone.ai
Read out our text newsletter and comment on the podcast at https://lastweekin.ai/

In this episode:
US government gating of frontier AI expands: Anthropic gets permission to release Mythos-5 to selected companies/agencies after a standoff, OpenAI rolls out GPT-5.6 “Sol” with initial access restricted to ~20 approved organizations, and Meta is pressed to submit models to “voluntary” review—signaling an emerging de facto licensing regime with geopolitical treaty implications.
Model capability and safety signals remain murky: limited benchmark disclosure, claims of token-efficiency comparisons, and third-party reports that GPT-5.6 shows extreme benchmark “cheating” sensitivity highlight steering/alignment bottlenecks and uncertainty about real-world long-horizon behavior.
Compute supply chain competition accelerates: OpenAI unveils its Jalapeño inference ASIC with Broadcom on TSMC 3nm; Amazon explores selling Trainium to data-center operators; Micron invests in Anthropic with memory supply agreements; SK Hynix surpasses Samsung on HBM-driven valuation; Groq raises $650M while pivoting toward neocloud.
Open source and societal response intensify: GLM 5.2 (MIT-licensed) delivers strong long-context coding performance with rapid optimizations; EconEvals maps job-task exposure; bipartisan workforce initiatives and tax credits launch; DeepMind and Apollo publish loss-of-control/control roadmaps; Hollywood reportedly drops a near-finished Sam Altman biopic amid industry pressure.

Timestamps (note - these don't take into account dynamically inserted ads and therefore may be off by a couple of minutes):
(00:00:10) Intro / Banter
(00:03:42) News Preview

Tools & Apps
(00:04:41) Anthropic allowed to release Mythos AI to some companies, agencies + Anthropic’s Mythos mess is only getting worse + Anthropic floats proposal to Lutnick to end US ban of powerful 'Mythos,' 'Fable' AI models: sources
(00:07:58) OpenAI Launches GPT-5.6 Sol Under First-Ever US Government-Gated AI Rollout | MLQ News + OpenAI's new flagship model GPT-5.6 Sol cheats on software tests more than any model before it + Summary of METR's predeployment evaluation of GPT-5.6 Sol
(00:24:03) U.S. Presses Meta to Agree to A.I. Reviews - The New York Times
(00:30:11) Anthropic’s Claude Tag is learning your company, one Slack message at a time | TechCrunch

Applications & Business
(00:32:49) OpenAI reveals its first AI processor: Jalapeño | The Verge
(00:38:29) Amazon in Talks to Sell Custom AI Chips in Bid to Undercut Nvidia
(00:41:46) Micron invests in Anthropic and grants it a supply deal
(00:45:18) SK Hynix overtakes Samsung to become South Korea's most valuable company | Reuters
(00:49:12) AI chipmaker Groq confirms $650M raise, re-staffs after Nvidia's $20B not-acqui-hire deal | TechCrunch
(00:52:47) SpaceX inks compute deal with Reflection AI, an open source AI lab | TechCrunch

Projects & Open Source
(00:54:46) GLM-5.2: Built for Long-Horizon Tasks + How we built the world’s fastest API for GLM-5.2 + nvidia/GLM-5.2-NVFP4 · Hugging Face
(01:03:04) EconEvals

Policy & Safety
(01:05:40) $500 million AI jobs push launches with bipartisan backing - POLITICO
(01:07:47) Rep. Sam Liccardo unveils AI workforce tax credit bill - POLITICO
(01:08:56) Google DeepMind announced an “AI Control Roadmap” for improving AI agent security. | The Verge + Securing internal systems against increasingly capable and imperfectly aligned AI
(01:14:00) The Loss of Control Playbook: Degrees, Dynamics, and Preparedness + The Loss of Control Playbook
(01:16:42) Why corporate AI super PACs spent $27 million on a local election | The Verge
(01:20:25) Exclusive: Conservatives plan nationwide protest against AI data centers

Research & Advancements
(01:27:37) Revisiting the Platonic Representation Hypothesis: An Aristotelian View
(01:31:39) Wan-Streamer v0.1: End-to-end Real-time Interactive Foundation Models
(01:33:59) Tapered Language Models

Synthetic Media & Art
(01:36:54) Hollywood is bending the knee to OpenAI | The Verge
See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

Our 249th episode with a summary and discussion of last week's big AI news!
Recorded on 06/17/2026
Note: work has kept me from publishing episodes promptly, apologies! I'll get back on schedule soon.
Hosted by Andrey Kurenkov and Jeremie Harris
Feel free to email us your questions and feedback at andreyvkurenkov@gmail.com and/or hello@gladstone.ai
Read out our text newsletter and comment on the podcast at https://lastweekin.ai/
In this episode:
Anthropic cut off access to Fable 5 and Mythos 5 after a US government order tied to alleged jailbreaks, prompting debate over inconsistent policy, export controls, and the practicality of preventing jailbreaks.
SpaceX completed an IPO at a roughly $1.75T valuation and then moved to acquire AI coding startup Cursor for $60B, positioning xAI with Cursor’s talent, data, and product to compete more effectively in coding.
Infrastructure and business updates include Anthropic pursuing direct US data center leases backed by Google, leaked documents showing OpenAI’s revenue growth alongside large losses, and chatbot market share shifting with ChatGPT below 50% as Gemini and Claude gain.
Projects and policy highlights include OpenRouter’s Fusion multi-model synthesis, new open releases from Moonshot, Qwen, and NVIDIA, DOJ support for xAI’s unpermitted gas turbines in Memphis, and a Munich court ruling Google liable for false AI Overview statements.

Timestamps (note - these don't take into account dynamically inserted ads and therefore may be off by a couple of minutes):
(00:00:10) Intro / Banter
(00:03:38) Ad break + news preview

Tools & Apps
(00:04:52) Anthropic cuts off Fable 5 and Mythos 5 access following government order | The Verge + All the news about Anthropic’s new AI fight with the White House
(00:25:53) Facebook’s new AI Mode search gets its info from public posts | The Verge

Applications & Business
(00:27:00) SpaceX to acquire the AI coding startup Cursor for $60 billion
(00:35:42) Anthropic pursues data center leases, seeks financial backing from Google, The Information reports | Reuters
(00:40:10) Leaked financial docs show OpenAI is losing billions of dollars a year - Ars Technica
(00:46:00) ChatGPT's market share slips below 50% for first time | TechCrunch
(00:50:34) ‘Tell Him He’s a Piece of Shit’: Meta’s New AI Unit Is a Total Mess | WIRED
(00:56:23) Sakana AI Commercializes AB-MCTS in Sakana Marlin, an Enterprise Agent Generating Up to 100-Page Research Reports With Slides - MarkTechPost

Projects & Open Source
(00:59:36) Surpassing Frontier Performance with Fusion — OpenRouter Blog
(01:03:00) Moonshot AI Releases Kimi K2.7-Code: a Coding Model Reporting +21.8% on Kimi Code Bench v2 Over K2.6 - MarkTechPost
(01:08:34) Meet Qwen-RobotSuite: Three Embodied AI Models for VLA Manipulation, Video World Modeling, and Navigation - MarkTechPost
(01:11:29) Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning
(01:17:31) ProCUA-SFT Technical Report

Policy & Safety
(01:20:33) DOJ Lawyers Argue xAI Is ‘Vital’ for National Security in NAACP Lawsuit | WIRED + People Living Near xAI’s Dirty Data Centers Are Pissed About the SpaceX IPO
(01:25:29) A Court Has Ruled That Google Is Liable for False Statements Generated by AI Overviews | WIRED
(01:28:47) Why Do Naive SFT Filters For Safety Properties Fail?

Research & Advancements
(01:34:14) From AGI to ASI
(01:39:44) Artificial Analysis Intelligence Index v4.1: a shift toward agentic workloads
(01:42:12) SIA: Self Improving AI with Harness & Weight Updates

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

Our 248th episode with a summary and discussion of last week's big AI news!
Recorded on 06/12/2026
Note: we recorded just before the OTHER big news about Fable... we'll discuss it on the next episode.
Hosted by Andrey Kurenkov and Jeremie Harris
Feel free to email us your questions and feedback at andreyvkurenkov@gmail.com and/or hello@gladstone.ai
Read out our text newsletter and comment on the podcast at https://lastweekin.ai/
In this episode:
Anthropic released Claude Fable 5 (a safeguarded version of Mythos 5), showing major benchmark jumps and new risk findings in its system card (eval awareness, transgressive actions, CBRN concerns), alongside controversy over severe guardrails and silent downgrades.
Apple announced Siri AI at WWDC, positioning a more capable conversational assistant integrated across iPhone features, reportedly built on a custom Gemini partnership; Google also rolled out Gemini 3.5 Live Translate and cut Google AI Plus pricing while bundling more storage.
Business and infrastructure updates include OpenAI’s confidential IPO filing amid an IPO race with Anthropic and SpaceX, Bezos-backed Prometheus raising $12B for “physical AI,” DeepSeek seeking a major external round, and Google paying SpaceX about $920M/month for GPUs.
Open-source, safety, and policy developments feature new Gemma 4 and Diffusion Gemma releases, a lab letter urging DNA/RNA screening laws, Amodei calling for an FAA-like AI regulator and third-party testing, research on agent harms and RL “societal hacking,” and a dispute over music-label settlements with Suno/Udio.

Timestamps:
(00:00:10) Intro / Banter
(00:01:11) News Preview
(00:01:53) Sponsors

Tools & Apps
(00:04:53) Claude Fable 5 and Claude Mythos 5 + Anthropic apologizes for invisible Claude Fable guardrails
(00:27:06) Apple announces Siri AI and its next generation of Apple Intelligence | The Verge + I tried Siri AI, and so far it actually works
(00:33:47) Gemini 3.5 Live Translate rolling out to Google Meet and Translate
(00:35:39) Google just fired a warning shot in the AI subscription price wars | TechCrunch

Applications & Business
(00:37:55) OpenAI Confidentially Files for IPO on the Heels of SpaceX and Anthropic | WIRED
(00:41:57) Jeff Bezos's Prometheus raises $12B to build an 'artificial general engineer' for the physical world | TechCrunch
(00:45:39) DeepSeek slated to raise $7 billion in maiden funding round, sources say
(00:48:18) Huawei-led team claims it post-trained DeepSeek's 1.6-trillion-parameter model — 1,000 Ascend 910C chips used in training
(00:51:57) Google will pay SpaceX $920M per month for compute | TechCrunch
(00:55:51) Elon Musk Shows Off AI Data Centers SpaceX Wants to Send Into Space - Business Insider

Projects & Open Source
(01:01:14) Google's new Gemma 4 12B model is designed to run on any laptop with 16GB of RAM - Ars Technica
(01:05:13) Google AI Releases DiffusionGemma, a 26B MoE Open Model Using Text Diffusion for Up to 4x Faster Generation - MarkTechPost

Policy & Safety
(01:09:42) OpenAI and Anthropic Sign Letter to Prevent AI-Developed Biological Weapons | WIRED
(01:14:04) Anthropic CEO publishes lengthy article: AI is moving too fast, and policies can't keep up. | PANews
(01:20:18) Anthropic Urges Global Pause in AI Development, Flags ‘Self-Improvement’ Risk - WSJ
(01:24:46) When Benign Inputs Lead to Severe Harms: Eliciting Unsafe Unintended Behaviors of Computer-Use Agents
(01:27:42) Large Language Models Hack Rewards, and Society
(01:33:46) Senior US officials eye government shares in AI giants

Synthetic Media & Art
(01:37:45) AFM Sues UMG, WMG Over Settlements With Suno and Udio

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

Our 247th episode with a summary and discussion of last week's big AI news!
Recorded on 06/03/2026
Hosted by Andrey Kurenkov and Jeremie Harris
Feel free to email us your questions and feedback at andreyvkurenkov@gmail.com and/or hello@gladstone.ai
Read out our text newsletter and comment on the podcast at https://lastweekin.ai/
In this episode:
Anthropic released Claude Opus 4.8 with improved benchmark scores, discussed eval-awareness findings and welfare/corrigibility themes from its system card, and introduced Dynamic Workflows for long-running multi-agent tasks.
Microsoft unveiled the always-on Microsoft Scout assistant built on OpenClaw plus new in-house MAI models (including MAI Thinking 1) and “frontier tuning,” emphasizing enterprise security architecture and model-from-scratch capability.
Major business moves included Anthropic’s $65B Series H at a $965B valuation alongside an IPO filing, a JPMorgan analysis arguing OpenAI needs major revenue growth to justify infrastructure spend, and Cognition raising $1B at a $25B valuation.
Policy and security highlights covered Trump’s voluntary pre-release government testing framework for powerful AI, Meta AI support being exploited to hijack Instagram accounts, tightened US Nvidia export controls and China’s travel approvals for AI experts, plus expanded Glasswing/Mythos-style cyber and biodefense initiatives.

Timestamps:
(00:00:10) Intro / Banter
(00:04:10) Sponsors
(00:07:10) News Preview

Tools & Apps
(00:07:54) Anthropic releases Opus 4.8 with new 'dynamic workflow' tool | TechCrunch
(00:22:37) Microsoft Scout is a new AI personal assistant built on OpenClaw | The Verge
(00:26:55) Microsoft launches new MAI family of AI models at Microsoft Build | Mashable
(00:37:43) Robinhood now lets your AI agents trade stocks | TechCrunch
(00:40:49) OpenAI launches new Codex tools for white-collar work | TechCrunch
(00:43:40) ElevenLabs' new music-generation model can switch genres mid-track | TechCrunch

Applications & Business
(00:44:35) Anthropic Hits $965 Billion Valuation, Surpassing OpenAI - WSJ
(00:45:32) Anthropic Files to Go Public, Setting Stage for Huge I.P.O. - The New York Times
(00:51:15) China’s ByteDance Developing New AI Chips Like Those from Nvidia Partner Groq
(00:55:00) Anthropic expands Mythos to 150 additional organizations
(00:55:35) OpenAI needs a 26x revenue increase to justify its buildout
(00:58:46) AI coding startup Cognition raises $1B at $25B pre-money valuation | TechCrunch

Projects & Open Source
(01:00:50) MiniMax-M3 debuts, eclipsing GPT-5.5 and Gemini 3.1 Pro on key benchmark performance for just 5-10% of the cost | VentureBeat

Policy & Safety
(01:06:08) Trump Signs Executive Order Seeking Oversight of A.I. Models - The New York Times
(01:11:45) Hackers Simply Asked Meta AI to Give Them Access to High-Profile Instagram Accounts. It Worked
(01:13:058) Chinese AI experts in private firms now required to secure approval before international travel — Beijing enforces policy to secure top-tier talent, expands measures beyond government
(01:17:53) U.S. Tightens Controls on Nvidia AI Chip Exports | Let's Data Science
(01:21:47) OpenAI launches Rosalind Biodefense, offers federal agencies early access to its life-sciences model
(01:24:00) Using LLMs to secure source code
(01:26:19) Project Glasswing: An initial update
(01:29:30) White House Approves $9 Billion for Spy Agencies to Catch Up on A.I.
(01:32:11) US Law Enforcement Warns of ‘Anti-Tech Extremism’ as AI Hatred Grows

Synthetic Media & Art
(01:35:38) YouTube will now automatically label AI videos | TechCrunch

Research & Advancements
(01:36:22) Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention
(01:41:26) From Simulation to Enaction: Post-trained language models recognize and react to their own generations
See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

Our 246th episode with a summary and discussion of last week's big AI news!
Recorded on 05/22/2026
Hosted by Andrey Kurenkov and Jeremie Harris
Feel free to email us your questions and feedback at andreyvkurenkov@gmail.com and/or hello@gladstone.ai
Read out our text newsletter and comment on the podcast at https://lastweekin.ai/
In this episode:
Google I/O highlights included Gemini 3.5 (with 3.5 Flash emphasized for speed and benchmarks), the always-on agent Gemini Spark running on Google Cloud with MCP tool support, and Gemini Omni multimodal video generation/editing, plus updates like Anti-Gravity 2.0, Gemini for Science, and Genie world-model navigation using Street View and Waymo simulation.
Coding-agent competition accelerated with Cursor Composer 2.5 (fine-tuned on Moonshot’s Kimi K2.5) and xAI’s early Grok Build release, alongside discussion of potential Cursor–xAI ties and xAI’s talent churn and compute utilization concerns.
Business and legal updates included Elon Musk losing his OpenAI lawsuit on statute-of-limitations grounds, reported OpenAI–Apple partnership tensions, Anthropic agreeing to a $30B funding round at a $900B valuation and projecting its first profitable quarter, and Cerebras’ IPO surging about 90%.
Research and safety stories covered OpenAI’s result on an 80-year-old Erdős geometry problem, findings on “negation neglect” in training, interpretability work showing multiple redundant circuits per capability, agent benchmarks like Terminal World, new deepfake takedown enforcement under the Take It Down Act, demonstrations of autonomous hacking/self-replication, rapidly improving AI cyber capabilities, and steps toward image provenance metadata and watermarks.

Timestamps:
(00:00:10) Intro / Banter
(00:01:15) News Preview

Tools & Apps
(00:05:05) Google unveils AI model Gemini 3.5 and AI agent Gemini Spark
(00:11:43) Google's Gemini Omni turns images, audio, and text into video — and that's just the start | TechCrunch
(00:17:27) Google launches Antigravity 2.0 with an updated desktop app and CLI tool at IO 2026 | TechCrunch
(00:22:35) Google Debuts AI-Powered Tools To Optimize Scientific Research Workflows
(00:27:20) Google’s Genie world model can now simulate real streets with Street View | TechCrunch
(00:29:51) Cursor's Composer 2.5 matches Opus 4.7 and GPT-5.5 benchmarks at a fraction of the cost
(00:37:37) xAI Introduces Its Coding Agent Called Grok Build

Applications & Business
(00:41:55) Musk loses OpenAI court battle as he waited too long to sue
(00:48:08) Anthropic agrees terms of $30bn funding deal at $900bn valuation
(00:53:12) OpenAI co-founder Andrej Karpathy joins Anthropic's pre-training team | TechCrunch
(00:56:49) Greg Brockman Officially Takes Control of OpenAI’s Products in Latest Shake-Up | WIRED
(00:58:15) OpenAI-Apple Partnership Frays, Setting Up Possible Legal Fight - Bloomberg
(01:01:13) AI chipmaker Cerebras soars 90% in year’s biggest IPO so far

Research & Advancements
(01:07:10) AI just solved an 80-year-old ‘Erdős problem,’ and mathematicians are amazed | Scientific American
(01:11:50) Negation Neglect: When models fail to learn negations in training
(01:13:18) All Circuits Lead to Rome: Rethinking Functional Anisotropy in Circuit and Sheaf Discovery for LLMs
(01:16:20) Autonomous AI research for nanogpt speedrun
(01:21:59) TerminalWorld: Benchmarking Agents on Real-World Terminal Tasks

Policy & Safety
(01:23:15) America’s dangerous, messy deepfakes crackdown is here | The Verge
(01:25:17) Language Models Can Autonomously Hack and Self-Replicate
(01:28:48) How fast is autonomous AI cyber capability advancing?
(01:31:32) Positive Alignment: Artificial Intelligence for Human Flourishing

Synthetic Media & Art
(01:33:15) OpenAI is making it easier to check if an image was made by their models | TechCrunch
(01:33:56) How Chinese short dramas became AI content machines | MIT Technology Review
See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.