AI agents go from 2.5% to 41.3% on real freelance benchmark in 13 months
kimmonismus · x · 2026-11-15
Citing Scale AI's Remote Labor Index, kimmonismus notes the best AI agent jumped from completing 2.5% to 41.3% of real paid freelance projects in 13 months, with Claude Fable 5.2 leading. These were paid projects averaging 29 hours of human professional work, with briefs provided and client calls excluded. The leaderboard shows Fable 5.2 at 41.25, Opus 5.5 at 21.25, GPT 6 Astra at 20.83, and DeepSeek V4.1 Flash at 9.58. The author argues white-collar work faces a wake-up in 2027.
More from AGI Musings
- Stanford posts full CS329A course: 9 lectures on self-improving AI agents — FinanceYF5 · 2026-11-15
- 'Human extinction' never appeared in official Anthropic statements, but it did in OpenAI's — erikphoel · 2026-11-15
- Roko: Anthropic may be the greatest threat to the human species, the ultimate irony of EA — basedjensen · 2026-11-15
- Unverified claim: Anthropic will have superintelligence internally by year-end — iruletheworldmo · 2026-11-15
- AI may finally unlock billions of human potentials that poverty and disease have wasted — VraserX · 2026-11-15
- Mirendil raises $1B for Lor platform where hundreds of agents run multi-day AI research — mark_k · 2026-11-15