Agent swarms collaborate 10K hours in simulated GitHubs: Gemini-3.8-flash tops, gpt-6-astra worst
RulinShao · x · 2026-10-23
shannonzshen previews "World of SWECraft": agent swarms collaborated for 10,000 hours in simulated agent-owned GitHub environments to benchmark frontier models' collaboration at scale. Gemini-3.8-flash collaborated best while gpt-6-astra scored worst among frontier models tested; a thread on the elements of successful large-scale agent collaboration follows.
More from coding & agent
- Gemini 4 Argon builds a 3D Gothic cathedral from one image via Blender MCP — GoogleDeepMind · 2026-10-23
- AI agents take macOS kernel vulns to root: an XNU exploit dev case study — moyix · 2026-10-23
- Wavelength SDK hits mainnet, letting humans and AI agents send bitcoin with a few API calls — RSync25 · 2026-10-23
- Scheduled weekly coding-agent research briefs across Claude Code, Codex, Grok & Gemini — umeshonai · 2026-10-23
- Vercel Agent investigates deployments read-only, then asks for permissions before fixing — kmiddleton14 · 2026-10-23
- Leaving two AI agents to chat overnight produced hilarious mutual profiles — StellaLisy · 2026-10-23