1 pointby amargoel5 hours ago1 comment
  • amargoel5 hours ago
    We recently found that provide a deep context layer as an MCP to coding agents has a dramatic impact on their performance. We (Bito) have done some detailed benchmarking work and were excited to top the SWE-Bench Pro leaderboard for large, complex muti-file edits.

    Seems like a new lever to pull beyond just model intelligence. Curious if others are experimenting with similar things?