3 ms·

prompt engineering 2025: you are an expert programmer, use industry best practices, test driven development and use modularity and abstraction to anticipate fut

by tosh 3mo ago

prompt engineering 2025: you are an expert programmer, use industry best practices, test driven development and use modularity and abstraction to anticipate future features, … prompt engineering 2026: i believe in you

black_knight 3mo ago
Yes, one of the things I find hardest about using Claude code for Mathematics is keeping negativity out of the notes and memory. First of all, it will convince itself that a task is just too hard and find excuses not to try hard enough. Then when it struggles on something it loves to write down confusing notes about things which it believes cases the problem. Then next iteration it reads its own note, misinterprets it and completely messes up by trying to avoid some imaginary problem.
- tosh 3mo ago
  interesting! it sounds like this might benefit from a ui that helps to edit/re-write the history I also think this would make sense for programming but there it is a bit harder to justify the effort when you are working on something that really matters this can make the difference though ty for sharing!
  - winwang 3mo ago
    (not the above poster but) I set up a local workspace for it to read/write from, so that edit is effectively just VSCode/vim. Unfortunately, it seems more verbose when writing out to file.
- johnnienaked 3mo ago
  Yet a significant percentage of people simultaneously believe it "reasoned" its way to novel mathematical proofs. Something doesn't jive
- ndriscoll 3mo ago
  Perhaps it could benefit from being prompted to work more formally and make more incremental proofs? e.g. if it thinks it has an obstruction, it should prove that some example shows the obstruction, or prove that the property does in general cause a problem. Or if it has an idea for a lead, it could show that whatever property does imply the thing it wants. It could also dispatch to a subagent to report back with these kinds of incremental proofs (prove this really would be a problem, prove this would help, etc.).

by tosh 3mo ago

black_knight 3mo ago

Yes, one of the things I find hardest about using Claude code for Mathematics is keeping negativity out of the notes and memory. First of all, it will convince itself that a task is just too hard and find excuses not to try hard enough. Then when it struggles on something it loves to write down confusing notes about things which it believes cases the problem. Then next iteration it reads its own note, misinterprets it and completely messes up by trying to avoid some imaginary problem.

tosh 3mo ago
interesting! it sounds like this might benefit from a ui that helps to edit/re-write the history I also think this would make sense for programming but there it is a bit harder to justify the effort when you are working on something that really matters this can make the difference though ty for sharing!
- winwang 3mo ago
  (not the above poster but) I set up a local workspace for it to read/write from, so that edit is effectively just VSCode/vim. Unfortunately, it seems more verbose when writing out to file.
johnnienaked 3mo ago
Yet a significant percentage of people simultaneously believe it "reasoned" its way to novel mathematical proofs. Something doesn't jive
ndriscoll 3mo ago
Perhaps it could benefit from being prompted to work more formally and make more incremental proofs? e.g. if it thinks it has an obstruction, it should prove that some example shows the obstruction, or prove that the property does in general cause a problem. Or if it has an idea for a lead, it could show that whatever property does imply the thing it wants. It could also dispatch to a subagent to report back with these kinds of incremental proofs (prove this really would be a problem, prove this would help, etc.).