AI Daily · Nov 10, 2026 · Tuesday
BAD INFLUENCE
Anthropic makes Claude Fable 5.2 available on all paid plans
Anthropic says Claude Fable 5.2 is available today on all paid plans. The post also says organizations pursuing beneficial applications in cybersecurity or biomedical research can apply for access to Claude Mythos 5.2.
Model releases & updates
Artificial Analysis@ArtificialAnlysOfficialClaude Fable 5.2 tops the Artificial Analysis Intelligence Index at 65
Artificial Analysis says Anthropic's Claude Fable 5.2 ranks first on its Intelligence Index v5 with a score of 65, five points ahead of Gemini 4 Carbon at high effort. The source says the gains come mainly from scientific reasoning, knowledge and FrontierCode 2.0, while agentic and terminal coding tasks roughly match Claude Opus 5.5. At max effort the average cost per task is $26.5, about 2.6x Opus 5.5.
AI at Meta@AIatMetaOfficial11 more sourcesMeta introduces Muse Ember, its smallest and most efficient model
Meta introduces Muse Ember, its smallest and most efficient model, built for complex reasoning execution and everyday agent work. The model can carry workflows across apps, call tools in sequence, run tests and make targeted code fixes. It has a 1M-token context window and natively understands text, images, audio and video.
- Artificial Analysis finds Meta's Muse Ember cheapest at Intelligence Index score 36Artificial Analysis Articles
AWS Machine Learning BlogOfficialAnthropic makes Claude Fable 5.2 available on Amazon Bedrock and Claude Platform on AWS
Anthropic has made Claude Fable 5.2 and Claude Mythos 5.2 available on Amazon Bedrock and Claude Platform on AWS. Anthropic calls Fable 5.2 its most capable generally available model for technical work in coding and the sciences. Mythos 5.2 is limited to authorized cyber security and life sciences work through Anthropic's verification programs.
Product launches & updates
Google@GoogleOfficial6 more sourcesGoogle expands eighth-generation TPU lineup with TPU 8i Max and TPU 8tx
Google says it is expanding its eighth-generation TPU portfolio with two new systems, TPU 8i Max and TPU 8tx, bringing the total to four systems launched in one year. TPU 8i Max is built for massive context windows and the highest compute density per rack, while TPU 8tx is optimized for secure and confidential inference. Google says the four systems give developers a purpose-built accelerator for each stage of the agentic lifecycle.
Industry news
Andrew Curran@AndrewCurran_XAnthropic plans for its shares to start trading next week
Andrew Curran says Anthropic plans for its shares to start trading on Thursday or Friday next week. The image is a Wall Street Journal article headlined that Anthropic is pushing ahead with its IPO despite market jitters, saying the company has investor meetings this week, according to people familiar with the matter.
Papers & research
Claude BlogOfficial2 more sourcesAnthropic reports a 1,000-agent team reproduces the Fermat's Last Theorem proof
Anthropic says a team of 1,000 agents running an early version of Claude Fable 5.2 reproduced its formal proof of Fermat's Last Theorem in 28.7 hours. The agents used one-third as many output tokens and wrote a proof one-seventh the length of the earlier Prove2Me-assisted formalization, with no researcher steering. Each agent ran in a sandbox without internet access, under a real-time classifier and an offline monitor, and none of the more than one million tool calls were blocked.
Rohan Paul@rohanpaul_aiXClaude Fable 5.2 completes a 23-year-old black-hole calculation
Anthropic's Claude Fable 5.2 completed a calculation on black hole formation in small universes that physicists had left unfinished since 2003. According to the post, Claude did most of the work largely on its own and found that black holes also form abruptly in those universes. A physicist has reportedly checked the work.
Anthropic ResearchOfficial1 more sourceAnthropic tests Claude on designing and prototyping real electronics
Anthropic and Diode Computers tested Claude Fable 5.2 on designing three electronic devices: a speakerphone, a voice-assistant hub, and a handheld virtual pet. Claude laid out the circuit boards, wrote firmware, and prepared parts lists, and Diode built working prototypes without circuit fixes. On the PCB Gym benchmark, Fable 5.2 scored 69% versus 25% for Fable 5. Anthropic says the prototypes have not been tested for certification or mass production.
Baidu's ERNIE Evolve automated research system reports gains on NanoChat and NanoGPT benchmarks
Baidu has released ERNIE Evolve, an automated research system that picks research directions, edits code, and runs experiments, keeping only changes that pass a hidden evaluator and a separate reviewer. Baidu reports it beat the best public result on three of four public metrics, including 38.9 s on NanoGPT Speedrun Track 1, and cleared production checks on ERNIE-5.5-Flash, with +33.7% deterministic throughput. The system is not released, and the results are self-reported.
A New AI Tool From Google DeepMind Proved the Signal Protocol Is Secure
Google DeepMind researchers used agentic AI and a new framework called AlphaProof Code to formally verify several key cryptographic guarantees of the Signal protocol in four weeks. The analysis found no new vulnerabilities in the parts examined, and it correctly flagged known, long-fixed flaws in older Signal versions. The researchers say the work covers only part of the protocol and does not verify the entire Signal mobile app, and they are releasing the specification and proof.
Tips & takes
One Useful Thing (Ethan Mollick)BlogEthan Mollick says Claude Fable 5.2 is almost AGI but still falls short
Ethan Mollick tested Claude Fable 5.2 on hardware hacking, business planning, academic research, creative projects, and a 3D world project. He concludes it is "almost" AGI under his limited definition, a view he calls Jagged AGI. He says the model failed to notice when its own work was joyless, overlong, or bureaucratic, which fell to him to catch.