@petrillic I am not defending the overhype nor am I excusing the externalities
Bugs are not unique constructs. They follow common patterns (eg failure to sanitize input). With a large enough corpus, building statistical models of the common failure patterns can lead to identifying other things that have those patterns.
All bugs… no.
30%-50% of common bugs patterns identified… that’s reasonable
@petrillic Lolol. Really? This was a statement? Omg. No way.
I watched hundreds if not thousands of bugs timeout and get archived after years. 1. No one is going back for those. 2. They are and always have come in faster than the engineering systems can support. LLM or not, there are limits.
Regardless, humans will still be the limitation.
3. Fixing bugs always adds new ones.
@petrillic earlier this year I asked a stakeholder three questions. The answer to number 2 was fine, but the answers to 1 and 3 were self contradictory
It took quite a bit of follow up to fix the self contradictory nature of the spec I was given
With that in mind, what is a bug?