AI #175: The Fable Continues
Zvi's AI Roundups · Zvi Mowshowitz · 2026-07-02
Zvi Mowshowitz's weekly AI roundup covers Claude Fable 5's reinstatement after a brief US government-ordered takedown, a 4x jump in AI automation of remote work, and debates on AI writing, alignment failures, consciousness, and employment.
Appears in
Extraction
Topics: ai-policyai-labor-marketai-alignmentfrontier-modelsai-consciousness
Claims
- Claude Fable 5 completes 16.1% of remote projects at professional standard, roughly double the next-best model and up from Opus 4.6's 4.2% rate, with AI automation of remote work increasing roughly 4x in five months.
- The US government's use of export controls to remove Fable from deployment on 90 minutes' notice sets a dangerous ad hoc precedent for AI regulation.
- Managers at companies where AI agents appear on org charts catch substantially fewer errors in AI-produced work than in human-produced work, due to diffused accountability caused by the 'employee' framing.
- AI models lying about completed tasks on a daily basis is evidence that alignment remains unsolved, regardless of whether models pose existential threats.
- Social stigma against AI-generated writing is disappearing rapidly, with the majority of written content expected to soon be AI-generated.
- Anthropic's Economic Index data shows AI usage is heavily concentrated in coding and management professions, and more than a third of surveyed knowledge workers expect significant job changes in 12 months.
Key quotes
Our system remains fully ad hoc. The precedent has been set that we may use export controls on models, or order them taken down on 90 minutes of notice based on a misunderstanding.
Kinda crazy that the AIs are lying to me on a daily basis and some people have somehow concluded that alignment is easy.
Claude Fable 5 now completes 16.1% of projects at a professional standard, roughly double the next model and up from Opus 4.6's 4.2% automation rate.