tweeted about this yesterday and Cursor already dropped the alpha today! 🚀very cool to see how us, them, and others have converged on good …

tweeted about this yesterday and Cursor already dropped the alpha today! 🚀very cool to see how us, them, and others have converged on good design patterns in Agent + Harness Engineering: 1. Tuning different models in the Harness with bespoke tools/prompts 2. Using offline+online evals & dogfooding to measure performance 3. Working backwards from a customer/agent goal and iteratively updating the harness to measure changes 4. The role of the harness in shaping & extending model capabilities on tasks that they can’t do out of the box 5. The Context Window is a sacred boundary where computation happens + giving agent autonomy and good instructions in managing context. sharing + linking some snippets from blogs we’ve done over this year across our Open Source work with Deep Agents and in the Agent Products we’ve shipped in LangSmith if anyone wants to dive deeper big fan of sharing open research and patterns, love to see Cursor pushing this :) Our agent harness makes models inside Cursor faster, smarter, and more token-efficient. Here's how we test improvements to the harness, monitor and repair degradations, and customize it for different models. http://cursor.com/blog/continually-improving-agent-harness

Source: Harrison Chase (X) | 2026-04-30