Cursor

We trained Composer to self-summarize through RL instead of a prompt. This reduces the error from compaction by 50% and allows Composer to succeed on challenging coding tasks requiring hundreds of actions.

214K

Cursor

@cursor_ai

Mar 18

Training Composer for longer horizons · Cursor

We built a fleet of security agents to run continuously on our codebase. We're sharing new automation templates for you to do the same.

Learn more about building custom security agents:

Securing our codebase with autonomous agents · Cursor

We're sharing a new method for scoring models on agentic coding tasks. Here's how models in Cursor compare on intelligence and efficiency:

We use a combination of offline benchmarks and online evals to measure model quality. This makes results more useful, especially as public benchmarks are increasingly saturated.