Debugging impossible problems with LLMs

I’ve been prototyping an app which integrates LLMs with code debuggers (like pdb or lldb). These are tools developers use to step through complicated programs line-by-line. The integration makes interactive debugging with an agent in the loop possible.