Anthropic upgrades Claude Fable and Mythos to 5.2

#1 📝 Anthropic News

Introducing Claude Fable 5.2 and Claude Mythos 5.2 - Anthropic News: On November 9, 2026, Anthropic announced upgrades to Claude Fable 5.2 and Claude Mythos 5.2, with improvements in coding, reasoning, and vision. In prerelease testing, the models helped Anthropic engineers complete otherwise-infeasible software migrations.

Also covered by: @Mike Krieger, @Claude, @Cursor, @How I AI Podcast, @Peter Yang, @Guillermo Rauch, @Cognition, @Aravind Srinivas, @Cursor, @Mike Krieger, @Peter Yang, @Anthropic, @v0, @There's An AI For That, @There's An AI For That, @Dan Shipper

#2 𝕏

Mistral AI announced Apps and Managed Agents in public preview and Workflows in general availability on Mistral Cloud. All three are planned for private-cloud and on-premises deployments at the end of November, though the year was not specified.

#3 𝕏

AI at Meta announced that Muse Ember is being brought to Google Cloud alongside Muse Wonder and Muse Spark 1.3, quoting the original Muse Ember launch post.

#4 📝 OpenAI News

ChatGPT Enterprise now available in AWS Marketplace - OpenAI News reports that ChatGPT Enterprise became available through AWS Marketplace on November 9, 2026, giving organizations a simpler way to discover, buy, and manage it.

#5 𝕏

Thinking Machines announced that Inkling 2 Small is available on OpenCode for free for a limited time. It was trained for agentic coding and knowledge work.

Also covered by: @Mira Murati

#6 𝕏

Mistral AI demonstrated Voxtral Transcribe 3 at AI Engineer Paris, transcribing every main-stage talk in real time. It boosted accuracy through context biasing with speaker names and technical terms from the conference agenda.

#7 📝 OpenAI News

Control Room: A new way to track and stop AI workloads - OpenAI News: OpenAI created Control Room, a central system for tracking model workloads across its research infrastructure. It is intended to detect potential issues and quickly stop models when needed.

Also covered by: @OpenAI

#8 𝕏

AI at Meta reposted Meta for Developers’ guide to Muse Ember, which covers the model, access, and recipes.

#9 𝕏

OpenAI shared that users can now connect multiple local computers to the same dot and choose which connected computer it uses for a task.

#10 𝕏

Santiago shared how connecting Codex or Claude Code to Swamp can turn assigned tasks into deterministic TypeScript workflows, reusable skills, and reusable behaviors for each step. The partnered post includes an installation link and attached video.

#11 ▶️

AI Agents are too good at marketing now

Greg Isenberg

#12 𝕏

claire vo 🖤 shared that she spends 90% of her time on @chatprd dealing with increasingly creative international token fraud operations. She also discussed fraud handling, product trade-offs, free trials, an enterprise pivot, and challenges in building AI products.

#13 in

Claire Vo recapped how William U. Imoh, despite a 1,000-day Duolingo streak, built a native iOS and Android voice tutor that refuses to speak English. The episode covers its differentiating prompt, real-time voice costs, agent-written code, and the agent team—including Gandalf as chief of staff—that Imoh uses to run his 25-person company.

#14 in

Udi Menkes recapped an implementation introduced “this month” for agents answering in chats such as Slack and Teams, with a video showing a Q3 revenue response rendered as charts, a customer table, and export and sharing buttons. He recommends evaluating both component choice and content accuracy and using a flexible harness; Vercel’s json-render and Chat SDK let agents build interfaces from preapproved components.

#15 𝕏

Dharmesh Shah shared that YouSpot supports MCP in both directions, connecting its second brain and tools to AI harnesses such as Claude or ChatGPT while also allowing users to bring an MCP server into YouSpot.

#16 𝕏

Claude recapped results on MIPLIB, a library of optimization problems including airline crew and delivery truck scheduling: of 203 open problems, Claude beat the listed mathematical solution in 72 cases and proved a solution was optimal in 45.

#17 𝕏

LlamaIndex 🦙 shared five LlamaParse deployment models spanning SaaS, customer cloud or data center, external inference providers, and customer-owned GPUs, with the choice driven by document residency and infrastructure management. LlamaIndex says its inference does not retain customer data or train on it.

#18 𝕏

Bolt shared instructions for building a customer inactivity warning dashboard in Bolt to spot early warning signs, identify customers who have gone quiet, and support efforts to win them back.

#19 𝕏

Harrison Chase shared a post describing eval-driven development (EDD) as TDD for AI agents and characterizing it as an article from LangChain.

#20 📝 Surge AI Blog

Training on ComplexConstraints: +10.1 on MultiChallenge, +8.5 on AdvancedIF - Surge AI Blog demonstrated that an unnamed 4B model trained on 1,000 expert-written ComplexConstraints rubrics reached parity with a 60x larger model, with gains transferring to unseen external benchmarks. The title reports +10.1 on MultiChallenge and +8.5 on AdvancedIF.