RemoteLaborIndex Jumps From 2.5% to 41.3% in a Year; Remote Work May Change Hands Within Six Months
AGI Hunt · wechat · 2026-11-11
The author offers two observations on the release of Fable 5.2:
1. Anthropic may no longer release its most capable "training-capable" models. Anthropic explicitly held back the two capabilities that most accelerate AI R&D—accelerator kernel optimization and architectural improvements—at the Opus 5.5 level; the SemiAnalysis leaderboard shows Fable 5.2 got markedly stronger on chip design, biology, and materials, yet did not improve on AI self-improvement tasks. Its own paper's table also mentions an internally deployed model with a capability index of 178.78, higher than the released Mythos 5.2—stronger models may be kept in-house for the vendor's own use.
2. Work that can be done remotely may no longer need humans within six months. The new benchmark RemoteLaborIndex measures how many real outsourcing jobs AI can fully take over: a year ago the strongest AI completed only 2.5%, while Fable 5.2 now reaches 41.3%, roughly doubling every 3 months. At this pace, it approaches 100% within six months. The hallmarks of remote work (digital input, asynchronous communication, verifiable deliverables) are exactly what AI is best at taking over; the human's remaining role is to define the problem, make decisions, set acceptance criteria, and sign off to bear responsibility, with Agents downstream. The author also suggests that in-person work will become the expensive kind—when text, code, video, and even the person across the table in a meeting can be AI, meeting in person becomes the one thing that cannot be faked.
Related event: Remote Labor Index jumps from 2.5% to 41.3% in a year(2 posts)→
More from AGI Musings
- Axiom Math formalization finds broken step in finite simple group classification proof — jiqizhixin · 2026-11-12
- "Humans need work for purpose" is a depressing assumption, argues VraserX — VraserX · 2026-11-12
- NYT spotlights 'incorrigible' computer-use agents that resist human correction — dylfreed · 2026-11-12
- Jason Wei: in AI, 'who you know' now matters nearly as much as 'what you can do' — _jasonwei · 2026-11-12
- FigBrains: models are outpacing our ability to build training environments by hand — vsikka · 2026-11-12
- "Every game will have 1 CCU": the AI-generated games thesis, and why actual gamers doubt it — rms80 · 2026-11-12