The bottleneck is not coding anymore, it's reviewing and testing
And arguably it's harder to automate this with AI, because most of the time you need to replicate a very specific state in the app
I really like Claude Code, but I wish it had the checkpoint feature like Cursor does
It’s really the only thing that is missing for me
Yes you can use Git, but still you can’t commit on every iteration, it’s becoming a mess (I tried)
cc Anthropic
Really like the Bun full-stack dev server thingy with React
But here is the problem
I want to use Hono in Bun, but I can’t use it with Bun.serve which renders React directly
Yes ofc I can run a separate server for React, but I really don’t want to
Any ideas?
If you are using Claude Code, really recommend you to setup its Cursor integration
And then Claude to run IDE diagnostics after changes, catches quite a few errors
I like easiness of PaaS e.g Vercel / Railway
But in January this year I wanted to try self-hosting, and got myself a VPS
It's such a cool and refreshing experience, makes you 10x more skilled as a dev
With Cursor and Claude Code it's now very accessible, so worth trying
Lately I've been thinking the buzz around LLMs and that AI coding fun is way too much. Yeah, you can make code really quick and it feels great at first. But as you have more AI code, the bugs and edge cases just keep growing like crazy
You start trusting the AI more, checking