Anthropic hands the public Mythos-class AI
Anthropic released Claude Fable 5, a restricted public version of its Mythos-class AI that tops nearly all major benchmarks, with access limits and pricing changes coming June 22. The newsletter also covers a Perplexity/Harvard study on AI agents shifting knowledge work patterns, and profiles a self-taught Japanese farmer using AI to build his own farm automation systems.
Summary
The newsletter's lead story covers Anthropic's release of Claude Fable 5, the first public access to its top-tier Mythos AI class. The original Mythos Preview was limited to 150+ vetted partners via Project Glasswing due to serious flaws discovered across major OS and browsers. Fable is a more restricted version, routing sensitive queries on topics like cybersecurity, biology, and chemistry to the older Opus 4.8 model. Despite restrictions, Fable achieves state-of-the-art scores across major benchmarks, outperforming both Opus 4.8 and GPT 5.5. Access is free across all Claude tiers until June 22, after which separate usage credits will be required at $10/M input and $50/M output tokens. Mythos 5 remains available to Glasswing partners at reduced costs with fewer restrictions.
A study co-published by Perplexity and Harvard Business School analyzed 10,000 identical queries sent to both Perplexity Search and its Computer agent platform. While Search responds in 33 seconds versus Computer's 26-minute average, the estimated end-to-end workflow time favors Computer (36 minutes vs. 269 minutes when accounting for user follow-through). The study also found that agent users requested more cognitively complex, cross-disciplinary, and creative tasks — suggesting AI agents may be expanding user ambition rather than just improving efficiency.
The newsletter profiles Hiroki Tomiyasu, a self-taught broccoli farmer in Hokkaido, Japan, who uses ChatGPT and OpenAI's Codex to build his own greenhouse automation, satellite crop tracking, and farm management software. Without a technical background or engineering team, he built a greenhouse vent control system, a farm group chat bot, and an Airtable-based operations hub — framing AI as an always-available engineer that lowers the automation barrier for small operators.
Additional news items include Google's launch of Gemini 3.5 Live Translate for 70+ languages, China's $295B data center expansion plan favoring domestic tech, Microsoft AI CEO Mustafa Suleyman criticizing Anthropic for discussing Claude's potential consciousness publicly, New York's new law requiring disclosure of AI-generated actors in ads, and OpenAI's release of interactive charts in ChatGPT.
About this episode
PLUS: Automate financial research with Dexter
Key Insights
- The newsletter argues that Fable 5 is notable precisely because, unlike typical frontier releases, the broader AI world actually seemed to agree with Anthropic's 'best model in the world' claim — making benchmark controversy rare this cycle.
- The Perplexity/Harvard study found that users shifted to more ambitious, cross-disciplinary, and creative requests when using an AI agent versus search, suggesting agentic AI changes user behavior and not just task speed.
- Anthropic structured Fable as a deliberately restricted version of Mythos, routing sensitive topics to an older model — indicating the company views public access and safety capability as a tradeoff requiring active architectural management.
- Microsoft AI CEO Mustafa Suleyman publicly called it 'really, really dangerous' for Anthropic to discuss Claude's potential consciousness in its documentation, signaling a notable inter-lab dispute over AI anthropomorphization.
- The Tomiyasu farming profile is framed as evidence of the 'selfware era,' where individuals with no engineering background can now build custom automation tools themselves using AI, bypassing the need to wait for purpose-built industry software.
Topics
Transcript
Good morning, {{ first_name | AI enthusiasts }}. Frontier releases usually trigger a week of benchmark arguments. Anthropic's new model just launched into a class of its own. After months of drama over a Mythos too powerful for the public, Fable arrives as the compromise — with scores that, for once, make "best model in the world" uncontroversial… Even if the guardrails and future access are. P.S. The Rundown just took its first strategic investment from Electrify to expand our newsletter content and beyond. Read about the deal here and read Rowan’s founder reflections here . Anthropic hands the public Mythos-class AI Perplexity data maps the agent work shift Automate financial research with Dexter Codex helps automate a Japanese broccoli…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The Rundown AI
Economists, researchers put AI’s job shock on the clock
Over 200 AI researchers and Nobel laureates signed a Stanford statement warning that AI could displace jobs at historic scale within the next decade, requiring immediate government action on safety nets and labor policy. Meanwhile, the AI industry continues to evolve with new tools, research findings on AI personality variations, and ongoing feuds between major figures like Musk and Altman.
Apple takes OpenAI to court
Apple has filed a lawsuit against OpenAI alleging the company poached over 400 Apple employees and used them to steal confidential hardware secrets for an unreleased device designed by Jony Ive. The newsletter also covers new AI tools, training resources, and research on long-term AI scenarios.
OpenAI sends GPT-5.6 to Work
OpenAI launched GPT-5.6 with three tiers (Sol, Terra, Luna), featuring near-Fable performance at lower costs and introducing ChatGPT Work with desktop integration. The newsletter also covers Meta's Muse Spark 1.1 release and argues that AI will concentrate demand among top 1-5% professionals while displacing mid-tier service providers.
SpaceXAI, Cursor release the strongest Grok yet
SpaceXAI and Cursor released Grok 4.5, a powerful new AI model that combines strong performance with faster speeds and lower costs than competitors like Claude and GPT. The newsletter also covers OpenAI's new GPT-Live voice model, ByteDance's Seedream 5.0 Pro image generation tool, and several other AI developments in the industry.
Meta climbs the AI image leaderboard
Meta released Muse Image, an in-house AI image model ranking No. 2 on leaderboards behind only OpenAI's GPT Image 2, with a teased video model also performing strongly. Meanwhile, Beijing is considering restrictions on Chinese AI model exports, creating potential geopolitical reciprocity risks similar to U.S. export controls, while companies like DoorDash demonstrate that pairing multiple AI models significantly improves code review accuracy and cost-effectiveness.