Bigtech agents/funding: Anthropic Fable 5 pulled by government; OpenAI GPT-5.6 Sol; Meta Watermelon; Vercel CEO decoupling; Trump restrictions; California SB 315 audits
Key Questions
What happened to Anthropic's Fable 5 model?
Anthropic's Fable 5 was pulled by government order, ending the 'move fast' era for that release. It was previously free until July 7 but is now unavailable.
How did OpenAI's GPT-5.6 Sol perform on benchmarks?
GPT-5.6 Sol crushed Opus 4.8 on TerminalBench with an 88.8% score. Questions remain about the reliability of such benchmarks.
What is Meta's Watermelon model achievement?
Meta claims its Watermelon model matches GPT-5.5 performance. It is part of broader efforts including a $6.5B Samsung chip deal.
What did Vercel CEO say about models and agents?
Vercel CEO Guillermo Rauch confirmed the decoupling of models from agents as a key industry battle. This validates Manus's multi-model strategy.
What regulatory changes are affecting frontier AI companies?
California SB 315 requires annual audits for frontier AI companies starting in 2028. Trump restrictions on private models are also pushing attention toward open source.
What does the first autonomous ransomware attack indicate?
The Jadepuffer attack underscores growing security risks for AI agents. It highlights the need for stronger governance in autonomous systems.
What market forecast supports the agentic AI thesis?
Gartner forecasts up to $234B in SaaS spending disruption from agentic AI. This supports continued investment and development in the space.
What funding news emerged for AI legal startups?
Norm, an AI legal startup, raised $120M at a $1.2B valuation. This signals strong market interest in AI agents for legal applications.
Anthropic's Fable 5 has been pulled by government order, ending the 'move fast' era in AI; previously free until July 7, now moot. OpenAI GPT-5.6 Sol crushes Opus 4.8 on TerminalBench but benchmark reliability questioned. Meta Watermelon model matches GPT-5.5. Vercel CEO confirms decoupling of models from agents as key industry battle, validating Manus multi-model strategy. Trump restrictions on private AI models turn attention to open source. Meta's $6.5B Samsung chip deal. First autonomous AI ransomware attack underscores security risks. Claude pricing remains opaque but users hooked. Gartner $234B disruption forecast supports agentic AI thesis. Greg Isenberg's thread on agent startup opportunities further validates the ecosystem. California SB 315 requires annual audits for frontier AI companies starting 2028, adding US regulatory pressure. Norm, an AI legal startup, raises $120M to $1.2B valuation, signaling strong market for AI agents in legal. New tool: Mixture of Agents Smart Router Edition routes coding tasks to best model per prompt, signaling trend toward model routing/ensembling. Rumored July timetable for GPT-5.6, Grok 4.5, Gemini 3.5, DeepSeek v4, Opus 5. Claude Code Loops deep dive on loop engineering provides competitive intelligence on Anthropic's agentic orchestration.