I'm noticing for some nontrivial work I really have to be on top of things. Also noticing that there's a lot more up front work of testing assumptions and writing out a detailed plan, otherwise it gets lost in the sauce. But it's an impressive leap forward. https://bsky.app/profile/phillipcarter.dev/post/3liz2h4no5s2q
At the moment there is always an aspect of you having to learn the limitations and working within its constraints. I learned over the months what the model is good at and what not. So yeah, for me that are certain types of tasks, and I know it will not do well on others
Yeah, this one's a little different too. I managed to get Windsurf + Claude 3.5 + NextJS apps into a great development loop, but now distributed tracing proxies in Go + Claude Code is just different!
Oh yeah, most definitely. Should really wax philosophical about this but IMO the future looks a lot like clearly defining problems to be solved and criteria for what solving it effectively actually looks like. And then kicking off N experiments via agents (or something?) and ranking results
Comments
Also, as a former member of the .NET team, I appreciate that y'all are adopting our versioning schemes :D
Wild time to be alive