tinamnstr

5 ms·

1. 1.▲
   by tinamnstr 7mo ago

   Thanks, I’ll definitely look into this! I’m planning some benchmark experiments with different prompting techniques since even small model updates (like the 1M token Claude change) seem to change how it uses the tool. If you try it out with
2. 2.▲
   Give your AI a codebase map or it will build one the expensive way(codeir.dev)

   2 points by tinamnstr 7mo ago|2 comments