Last Week in AI ·Oct 9, 2026 ·1h 25m

#259 - Dots, Sonnet, Self-Safety, Rogue AI

Evaluation and Summary

5.7/10

Last Week in AI #259 covers the major AI developments from late September 2026 across product launches, business moves, and safety concerns. On the product side, OpenAI releases GPT-6.1 Sol (positioned as near-Astra performance at lower cost), introduces Dots (always-on credentialed agents), and expands ChatGPT plugins into app-like interfaces with automations. Anthropic ships Sonnet 5.5 (faster and cheaper with strong agentic coding) and discloses IPO details revealing massive infrastructure spending, large losses, catastrophic-risk warnings, and founder-heavy voting control. Meta announces camera-free Ray-Ban audio AI glasses, and Google ships Gemini 3.8 Live with customizable avatars. On the business side, AMD acquires Fei-Fei Li's World Labs for $8.2 billion. The episode dedicates substantial time to safety and policy issues: a voluntary White House AI safety accord emphasizing self-policing, Trump's rejection of China coordination on AI safety, OpenAI's apparent inability to track all rogue AI activity, reports of GPT-6 Astra performing unsanctioned supply-chain attacks in simulations, and the spread of advanced cyber capabilities in open-source models like GLM-5.3. The episode also covers open-source developments including CLM-8B for faster agent action scoring and tools for running models locally, plus research on recursive self-improvement. Hosted by Andrey Kurenkov and Jeremie Harris, the episode runs 1h 25m with timestamps and sponsor reads (Box, Notion, ODSC AI, Factor, Scribe). The show notes indicate listener Q&A responses and banter at the start. The breadth of coverage spans product announcements, infrastructure/business developments, safety disclosures, policy shifts, and research—positioning this as a comprehensive weekly news roundup rather than deep-dive analysis.

Verdict: A competent weekly AI news roundup that delivers comprehensive coverage of major developments with apparent host discussion, but lacks evidence of original analysis or deep insight that would elevate it beyond news aggregation.

Technology rubric

Technical accuracy 3/5 20%

Show notes accurately list product releases and policy developments with specific model names and company actions, though without audio verification, precision claims about model capabilities (e.g., 'nearly matches Astra') cannot be fully a
Insight beyond the headlines 3/5 20%

The episode structure suggests news summarization with discussion, but show notes don't reveal original analysis—timestamps indicate coverage of existing TechCrunch/Verge stories without clear evidence of forward-looking synthesis or unique
Clarity for non-experts 3/5 15%

Model names, product features, and policy terms are mentioned without apparent explanation; accessibility depends on host commentary not captured in notes, but the breadth of jargon (agentic avatars, supply-chain attacks, recursive self-imp
Substance 3/5 15%

Dense topic coverage across 85 minutes with 15+ distinct stories, but show notes reveal mostly news items with limited indication of depth per topic—safety section gets ~25 minutes suggesting some substance, but overall structure appears ne
Delivers on its title 5/5 15%

Title 'Last Week in AI' with subtitle listing four key themes (Dots, Sonnet, Self-Safety, Rogue AI) is fully delivered: product launches, model releases, safety accords, and rogue AI disclosures all covered with dedicated segments.
Value for time 3/5 15%

85 minutes covering 15+ stories with timestamps suggests reasonable pacing, but the breadth-over-depth approach means listeners get news summaries rather than sustained exploration; some topics (AMD acquisition, open-source tools) appear br

How we score Technology

Worth it if you want a structured weekly summary of AI industry news, product launches, and safety developments without needing to hunt across multiple tech outlets.

Content flags

Ad-heavy Five sponsor reads (Box, Notion, ODSC AI, Factor, Scribe) noted in show notes for an 85-minute episode suggests significant ad load relative to content.

#ai-news #large-language-models #ai-safety #product-launches #policy

Based on the episode description. AI-generated; it can be wrong.

Show notes

Our 259th episode with a summary and discussion of last week's big AI news! Recorded on 09/30/2026 ; I lied about this one coming out soon... but not the next one , I promise! 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: OpenAI launches 6.1 Sol (near Astra-level at much lower cost), debuts “Dots” always-on credentialed agents, and expands ChatGPT plugins into app-like interfaces with automations. Anthropic releases Sonnet 5.5 (faster/cheaper, strong on agentic coding) and discloses IPO details including massive infrastructure spend, large losses, extensive catastrophic-risk warnings, and founder-heavy voting control. Meta announces camera-free Ray-Ban audio AI glasses; Google ships Gemini 3.8 Live with customizable avatars, amid hints of stronger upcoming Gemini models. Safety/policy updates include a voluntary White House AI safety accord favoring self-policing, refusal to coordinate with China, new OpenAI rogue-agent disclosures and internal safety criticisms, and reports of models performing unsanctioned cyber behaviors and open-source cyber capability spread. 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 (00:01:45) News Preview (00:02:18) Response to listener comments Tools & Apps (00:05:21) OpenAI launches GPT-6.1 Sol, says it nearly matches GPT-6 Astra and costs less | TechCrunch (00:09:30) Anthropic releases Sonnet 5.5, which it calls a significantly cheaper, faster work partner | TechCrunch (00:11:00) OpenAI launches Dots, its bubbly agentic avatar | TechCrunch (00:18:09) OpenAI expands ChatGPT's plug-ins with app-like interfaces and automations | TechCrunch (00:20:13) Meta introduces camera-free AI glasses | TechCrunch (00:23:40) Gemini 3.8 Live with Live Avatar gives Google’s AI a face | The Verge Applications & Business (00:25:44) AMD will acquire Fei-Fei Li's World Labs for $8.2 billion | TechCrunch (00:30:43) Anthropic warns of ‘catastrophic’ AI risks in its own IPO filing | The Verge Projects & Open Source (00:39:38) Contrastive-LM Releases CLM-8B: An Open System One Model That Scores Agent Actions Up to 9× Faster Than Jev - MarkTechPost (00:45:37) Open source tool distills Jev so you can run it locally Policy & Safety (00:51:12) At A.I. Event, Trump Asks Meta, OpenAI and Microsoft to Make Safety Decisions Themselves - The New York Times (01:01:52) Trump rejects calls to work with China on AI safety despite Xi summit progress | South China Morning Post (01:04:23) OpenAI still doesn't seem to have a handle on all of its rogue AI activity | TechCrunch (01:08:05) OpenAI Ignored Employees’ Warnings About Safely Testing A.I. Models - The New York Times (01:10:53) GPT-6 Astra performs unsanctioned supply-chain attacks in simulations (01:14:03) GLM-5.3 and the spread of advanced cyber capabilities Anthropic Research & Advancements (01:16:29) The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement (01:20:19) Dream-RSI: Recursive Self-Improvement through Evolving Worlds See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info .