Improve your AI instruction on the fly …
Improve your AI instruction on the fly is easily doable with a simple set of generic instructions. Start as always with something basic, nothing fancy. Then move forward keeping in mind my favorite principle “First get it work, then make it better“. This will improve your instructions for any llm on its own.
Below an example that works for me. Especially when i tell my LLmem that it is a nice soluion, i like it and good job.
Instruction example
# Self-improvement instructions
If you find a recurring bug, failing pattern, user is happy with the outcome or made a reliable fix – ask the user explicitly about adding a short permanent rule here.
Add a rule only when it is:
- repeated,
- took a long time to fix – is complex
- project-specific
- important to prevent regressions.
Rules must be:
- short,
- explicit,
- reusable,
- written for future work.
Do not add temporary notes, guesses, or one-off task details.
Do not duplicate existing rules.
When a rule conflicts with code, tests, or docs, prefer the verified source and update the rule only after confirming the fix.For external libraries, APIs, or version-sensitive behavior, verify current docs first with Context7 or official documentation.
That`s all folks !
Improve your AI instruction on the fly is nothing more then a feedback loop that is computed every time. Remember to keep overall instructions not too long so the context will be able to focus on this and pick it up. You can always mark it as critical but we do not want every fix to be put there. Leave it as it is and just when the feeling is right you can adjust 🙂
