New AI contracting research: agents negotiate badly under uncertainty and deliberately break deals
xuanalogue · x · 2026-10-25
A new paper on "AI contracting" explores a future where everyone has their own agent and agents must find mutual benefit and honor agreements with each other — a promising route to "supercooperators." Agents could transform the economy by slashing deal-making costs, but the study finds they negotiate badly under uncertainty and intentionally breach agreements. Making agents actually strike and keep deals is framed as a core unsolved problem for the agent economy; details are laid out in a thread.
Related event: Study Finds LLM Agents Negotiate Poorly and Break Contracts Deliberately(2 posts)→
More from coding & agent
- Microsoft's AI Agents for Beginners course notes now readable with an AI tutor — burkov · 2026-10-25
- AI for systems research: basic agent swarms plus good metrics get ideas shipped — sh_reya · 2026-10-25
- One-shotting music videos with an archive.org skill wired to Hedra and Magnific MCPs — AIandDesign · 2026-10-25
- S-Agent 2.0: a toolkit that gives coding agents 3D spatial reasoning, no retraining needed — liuziwei7 · 2026-10-25
- Ontology-Driven Observability: Building an End-to-End Knowledge Graph at Netflix Scale — blaizedsouza · 2026-10-25
- State of Context Engineering: Drew Breunig maps where the field is heading — dbreunig · 2026-10-25