
Our 261th episode with a summary and discussion of last week's big AI news!
Recorded on 10/21/2026 ; as indicated at start of ep, some health issues have really screwed up the release schedule unfortauntely
Hosted by Andrey Kurenkov and Jeremie Harris
Feel free to email us your questions and feedback at [email protected] and/or [email protected]
Read out our text newsletter and comment on the podcast at https://lastweekin.ai/
In this episode:
- Google released Gemini 4 Argon, a well-rounded multimodal frontier model with detailed safety reporting, low hallucination rates, and noted bio/cyber/stealth risk thresholds.
- Meta unveiled Muse Wonder/Wonderful frontier models and an 816B open-weights MuseSpark Open (plus 30B Glimmer), with debate over real-world performance versus benchmark strength and growing infrastructure ambitions.
- New model and tooling launches included Inception’s fast diffusion-based Mercury 6, TypeSafe’s expanded Jev preview, Nvidia’s Nematron 3.5 Super VL, MiniMax M3.1 Flash, and new PyTorch support layers for Google TPUs and Amazon Trainium.
- Business/policy and safety updates covered NeoCloud valuation/debt concerns, leaked Anthropic S-1 financials and compute commitments, reports of an Anthropic-drafted AI safety bill and external evaluators, an OpenAI engineer charged with planting a recording device, and a notable interpretability claim that modern neural networks are “cleaner” (more monosemantic) than previously believed.
A thank you to our current sponsors:
- Box - visit box.com/LWIAI to learn more
- Notion - visit notion.com/lwai to try Notion’s Developer Platform today.
- ODSC AI - visit odsc.ai/east and use promo code LWAI for an additional 15% off your pass to ODSC AI East 2026.
- Factor - visit factormeals.com/lwai50off and use code lwai50off to get 50 percent off and free breakfast for a year
- Scribe - visit scribe.how/lwai and mention lwai for a 30-day
Timestamps (these may a few minutes off due to sponsor inserts):
- (00:00:10) Intro / Banter
- (00:05:11) News Preview
- Tools & Apps
- (00:05:52) Google finally releases Gemini 4 Argon | The Verge
- (00:14:09) Meta Unveils New Series of Advanced A.I. Models - The New York Times
- (00:21:23) Inception says Mercury 3 beats GPT-6 Luna and Claude Haiku 5.5 at up to 8x the speed - SiliconANGLE
- (00:29:15) TypeSafe AI Releases Jev 1.14 With Images, Box Type and 262k Context - MarkTechPost
- Projects & Open Source
- (00:30:32) Meta Releases Muse Spark Open: 816B Open-Weight MoE With 1M Context - MarkTechPost
- (00:35:54) Muse Glimmer 30B: An On-Device Agentic Model
- (00:37:18) NVIDIA Releases Nemotron 3.5 Super VL, a 120B Open Multimodal Agent Model - MarkTechPost
- (00:38:15) MiniMax Just Launched M3.1 Flash: 428B MoE Agentic Model at $0.15/M Input
- Applications & Business
- (00:39:41) The Fallout From the ‘Neocloud’ Wipeout - The New York Times
- (00:47:13) Anthropic S-1 Shows Red-Hot Growth, Risks and Massive Computing Costs
- (00:58:36) Google unveils TorchTPU in a push to break Nvidia’s stranglehold on AI compute
- (01:01:24) Meta Muse Downloads Are Slowing — And Hunterbrook is Short
- Policy & Safety
- (01:05:56) Anthropic quietly pitches its own AI safety bill to lawmakers
- (01:15:16) Anthropic taps METR and SecureBio to help assess risks and safety
- (01:18:02) OpenAI Engineer, Who Is Charged With Planting Bug, Advised Saudis - The New York Times
- (01:23:41) Neural Networks Are Already Clean
See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.
Więcej odcinków z kanału "Last Week in AI"



Nie przegap odcinka z kanału “Last Week in AI”! Subskrybuj bezpłatnie w aplikacji GetPodcast.








