Yeah! I feel that the current generation LLM (basically the second generation of models trained to execute tools rather than trying to be a smarty pants on their own) is getting pretty good at search, compare, incorporate patterns.
The other day my codex'd GLM was without me asking explicitly mapping known vulns to code when reviewing a massive release milestone on a library. I merely instructed "all details must be referenced" and "leave no stone unturned" among acceptance criteria, which apparently can lead to employee-of-the-month behavior. So I guess GLM is going to be just that, lol.
Yeah! I feel that the current generation LLM (basically the second generation of models trained to execute tools rather than trying to be a smarty pants on their own) is getting pretty good at
search, compare, incorporatepatterns.The other day my codex'd GLM was without me asking explicitly mapping known vulns to code when reviewing a massive release milestone on a library. I merely instructed "all details must be referenced" and "leave no stone unturned" among acceptance criteria, which apparently can lead to employee-of-the-month behavior. So I guess GLM is going to be just that, lol.