Continually improving our agent harness · Cursor
Models need a harness to become fully useful, and improving that harness means continuously iterating on context, evaluation, and model-specific tuning.
Blog / research We approach building the Cursor agent harness the way we'd approach any ambitious software product. Much of the work is vision-driven, where we start with an opinion about what the ideal agent experience should look like. From there, we form hypotheses about how to get closer to that vision, run experiments to test them, and iterate using quantitative and qualitative signals from evals and real usage. That process depends on having the right online and offline instrumentation, so we can tell when a change actually makes the harness better. When we get early access to new models
saved by
related reading
- Continually improving our agent harnesscursor.com
- Best practices for coding with agents · Cursorcursor.com
- The third era of AI software developmentcursor.com
- What we’ve learned building cloud agents · Cursorcursor.com
- Cursor: AI coding agentcursor.com
- Towards self-driving codebases · Cursorcursor.com
- Context Engineering for AI Agents: Lessons from Building Manusmanus.im
- Dynamic context discovery · Cursorcursor.com
- How we built Linear Agentlinear.app
- Effective context engineering for AI agents \ Anthropicanthropic.com
- Context Engineering & Coding Agents with Cursoryoutube.com
- Scaling long-running autonomous coding · Cursorcursor.com