Repository navigation
feat(post): Going All In on AI Agents for Five Weeks: What I Learned - #782
Conversation
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…arned Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
luizonthedrums
left a comment
There was a problem hiding this comment.
I added some comments. Somehow, I think it could emphasize more that one learning is that we still need software engineering expertise, even when AI is doing all the coding work
Thanks for your thoughts here @luizonthedrums - I actually thought that was the main point I wanted to show in this thread: writing more tech plans, choosing the right architecture and the appropriate design patterns is still something we need to do and it's made easier with prototyping and the new models |
araujobarret
left a comment
There was a problem hiding this comment.
Thanks for condensing all the information, I don't have any opinion on that but a question about the metrics, are we sure we can use these metrics considering a 40hrs week or did we end up extrapolating it?
I'm sure we extrapolated because I see commits that came in during weekend or outside EU working hours too 😅. So that affects the metrics a lot imo.
|
anandaroop
left a comment
There was a problem hiding this comment.
Thanks for writing this up. I think you've laid out well the benefits of working in this style. I'd be interested in hearing a little more about risks and pitfalls. But I understand if that is not the piece you wanted to write. We'll have more chances I'm sure 😄
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This post covers what I learned building City Guide in five weeks with AI coding agents: tech plans, prototypes, QA, MCP and owning code you didn't write.
FWIW, these are my own thoughts, not necessarily something I think others should agree with fundamentally. I am happy to make changes of course and add bullet points to things I might have missed and if it's something that can be published in the blog