Write tests and actually do the practice of testing every single line with no regression testing. That will solve the AI problem. Since if it does work, there is no problem. If doest work just remove it. If it does work, who cares if it’s AI or not?
Maybe big companies and projects with lots of people do white tests, but indie devs don’t. FOSS is not the well organized machine you may have it confused for.
Humans in the loop? Code would have to be read and understood by the human, in relation to the problem being resolved, and the rest of the code base it may or may not interact with… And that’s difficult.
No, it’s shitty because it’s complicated, looks correct, but actually filled with errors and hallucinations. Including libraries that may not exist, and that leaves code open to injection attacks (where people make fake libraries/dependies based on commonly hallucinated code, then package them with genuine FOSS software which is now playing cover for malware/viruses).
Really awful website with more ads and shit than content.
Nevertheless worrysome and indeed food for tough. Ai is here to stay, so we all need to find ways to deal with it, that we like it or not.
Maybe specialized humans in detect ai slop? Because using ai to detect ai seems kind of hironic.
There is the new vouch system too
Use ai to make code to detect and block AI. Actually though wish we could force an AI tag tat could just be blocked with UBO
But what polices that? And what’s to stop humans pasting in the AI code, which may look like human code, just behave poorly or contain hallucinations.
Write tests and actually do the practice of testing every single line with no regression testing. That will solve the AI problem. Since if it does work, there is no problem. If doest work just remove it. If it does work, who cares if it’s AI or not?
Maybe big companies and projects with lots of people do white tests, but indie devs don’t. FOSS is not the well organized machine you may have it confused for.
Humans in the loop? Code would have to be read and understood by the human, in relation to the problem being resolved, and the rest of the code base it may or may not interact with… And that’s difficult.
If ai slip vibe code is so shitty, then it should be pretty easy for humans to detect it and shut down without wasting too much time right?
No, it’s shitty because it’s complicated, looks correct, but actually filled with errors and hallucinations. Including libraries that may not exist, and that leaves code open to injection attacks (where people make fake libraries/dependies based on commonly hallucinated code, then package them with genuine FOSS software which is now playing cover for malware/viruses).