What it feels like to work with Mythos
One Useful Thing · Ethan Mollick · 2026-06-09
Ethan Mollick's early-access evaluation of Claude 5 Fable finds the Mythos-class model executes multi-hour autonomous research and coding workflows far beyond prior models, concluding it fundamentally shifts the human role from active collaborator to passive commissioner.
Appears in
Extraction
Topics: claude-5agentic-aiai-capabilitieshuman-ai-collaborationmulti-agent-systems
Claims
- Claude 5 Fable outperforms all previously available public AI models by a considerable margin across research, coding, creative, and data analysis tasks.
- Fable autonomously orchestrates multi-agent workflows, running sub-agents for research and verification for up to twelve hours on complex multi-page specifications.
- Fable shifts the human role from active collaborator steering a process to a patron who briefs, funds, and judges finished work without controlling intermediate decisions.
- Fable is twice as expensive as Claude Opus and burns tokens at high rates, though delegation to cheaper sub-models may reduce the real cost considerably.
- Fable's cybersecurity guardrails are overly sensitive, defaulting to the less capable Claude 4.8 Opus at the faintest hint of a security-adjacent task.
Key quotes
I am no longer sure I am the wizard. I am closer to a patron. I describe what I want, I pay for it, and I judge the result. The conjuring happens somewhere I cannot watch, in hundreds of small choices I never get a vote on.
The work has shifted from process to outcome. I no longer steer; I commission.
A patron commissions a single artist. Fable is closer to a whole studio, where I am the client who signs off on the final work without ever setting foot on the floor.