Princeton paper: combining specialized agents from many developers hits 70% on GAIA
rohanpaul_ai · x · 2026-10-20
A new Princeton paper, "The Wealth of Agents," proposes multi-developer agentic systems (MDAs). Most agentic systems are built by a single developer, forcing teams to rebuild the same basic skills. Key findings: on GAIA with the same total development budget, specialists scoring at most 35% alone reached 70% combined — 12 points above the median general agent; on web-search-only SealQA the combined system did worse at higher cost; a 19% agent lifted a combined system from 40% to 53%. The authors argue MDAs could be a new scaling axis, incentivizing specialization.
Related event: Princeton's Multi-Developer Agent System Hits 70% on GAIA(2 posts)→
More from coding & agent
- DeepSeek V4.1-Flash-1017 Free in Qoder at 0.0x Credits Through Oct 24 — hey_abusiddik · 2026-10-20
- User begs Claude devs to fix Mac app's session reshuffling, proposes heat-based colors — RileyRalmuto · 2026-10-20
- Open-source CLI audits agent traces for wasted tokens — one agent burned 35% of I/O re-verifying itself — Impressive_Laugh_176 · 2026-10-20
- Classification doesn't need the biggest model: matching complexity to task and cost — js_craft_hq · 2026-10-20
- Auto-research on idle GPUs: a CoreWeave pipeline that keeps agents busy — _ScottCondron · 2026-10-20
- Seedance 2.5 to playable game: 3-hour MVP claim from viral skate clip — MoonL88537 · 2026-10-20