dashboard/app: where can LLMs help us?

[tarasmadan]

I think we have a quite good feeling of what problems can LLMs help us to solve.
Let's enumerate them:

1. Bugs de-duplication. Current context looks big enough to load all the bugs.
2. Bugs moderation. The the silent rules we follow to check for example KASAN bugs. Writing the rules in text may be the way to at least semi-automate this analysis.
3. What else?

[dvyukov]

Some use-cases that come up before:

• bug prioritization
• help debugging/root cause bugs (point to potential root cause)
• fix generation
• reproducer generation
• test-case generation
• syzlang descriptions generation
• syzlang descriptions type refinement for clang-tool-generated desriptoins
• reproducer simplification
• incoming email classification (question, complaint, praise, feature request, ...)
• FAQ answering (for syzbot and syzkaller mailing lists)

Hint: you can ask LLM how it can be used for syzbot as well ;)

[dvyukov]

• cause bisection (pick commit from recent commits that touch guilty files)
• patch reviews (review-prompts)