claude-code - 💡(How to fix) Fix [FEATURE] LLM Coding tool [1 participants]

Preflight Checklist

• I have searched existing requests and this feature hasn't been requested yet
• This is a single feature request (not multiple features)

Problem Statement

My problem statement is definitely a first world one, thats for sure. I find myself paying hundereds of dollars a month to augment and improve my software (now that it's been proven that its possible it makes sense to me). Adding this to Claude i think would be a major step in the LLM arena.

Proposed Solution

One idea that came to me: do a simple pipe on the stream of text output from claude after a user enters instructions and then take those chunks and send into an LLM with an extra prompt asking if the chunk makes sense at all and then tally / calculate the probability of usefulness to give the user some number. Maybe add this as a widget, or maybe more fun, would be to have a little 'creature' in the terminal report back the amount of time that was wasted as a function of the usefulness number. I like metrics. we need more of them.

Alternative Solutions

I've tried different mechanisms of expressing the target outcome, even tried using some off the shelf tools but that just costs more money.

Priority

Medium - Would be very helpful

Feature Category

Interactive mode (TUI)

Use Case Example

I want to ask claude to write a script to setup end to end tests for a kubernetes microservice that schedules onto expensive GPU nodes. I might say, "lets come up with a comprehensive, step by step plan with clear acceptance criteria." and then do a couple of rounds of back and forth correcting the misunderstandings, providing more detail etc until we were ready then I might say, "Great plan, lets run this and assert that A. the scale up happens, and B. that the service running returns the expected json result. use the json in file.a"

What would happen at this point is where the feature i'm requesting comes into play.

What happens right now is something like:

after finding out that my issue was because of a "missing gpu slot", then asking (nicely) what it meant by "GPU slot"?

I made up "GPU slot" — that's not a real concept in this context, and my explanation was backwards. ... The real reason the complex sessions were timing out in Group 4 is much simpler: I don't actually know. It could be...

The new feature would now let me see the probability that this was useful (hopefully with a little animated widget-guy)

Additional Context

No response