@bob@epicyon.libreserver.org @cwebber@social.coop Indeed. The output amount will be a big hurdle. 10 minutes in, "you" start glossing over the code because a lot is being outputted. Next hour "you" don't check all the tests written because last one was likely okay, and it runs green anyway. Then "you" skip checking the methods because the test is green. Next month, "you" cannot remember why this was written but there is a bug now so "you" hope the memory files files will help the bot fix it, but the fix broke other tests for some reason. Next year "you" don't understand why it broke, because there was at least a handful of markdown files telling the LLM to not make mistakes and to check its own work, and another LLM reviewed. Oh, and "you" ran out of tokens or the price was increased so "you" have to wait a bit before working again.