Daniel JalkutWe will know that AI has really succeeded when every braindead-obvious bug in every piece of software disappears.
dxzdb@danielpunkass Hmmm ...You mean without us learning how it got there? Sounds like a recipe for disaster. Especially because the magic that does that is likely to introduce some really hard to diagnose NOT obvious bugs.
@dxzdb mostly just alluding to how much faster it is to find and fix bugs. In my opinion LLMs are even better at this than writing fresh code.
Markus Fritze@danielpunkass I let AI check every unexplained Crashlog for our apps I could my hands on. 80%-90% are now considered fixed by engineering. The rest is either still in review or an engineer is working on a fix, after AI provided some new ideas. Only one crashlog was a dud. Many could have been fixed by an engineer, but 25% were considered “no clue what is going on”.