Thanks for sharing. It sounds like a reasonable approach. I would try these technics for sure.
Sounds like a lot of ground work 🤔
With the pace of AI dev tools evolution I wonder if this ground work will pay off.
And I am really curious if this approach works with other AI tools. An entry ticket price for Claude Code is high. I would expect quite many folks would be more keen to experiment with free options like Gemini.
Yeah, you're totally right about how fast these tools are evolving! The good news is this approach pretty much works everywhere - I've been testing Claude Code, Gemini CLI, Warp 2.0, and the others, and they all basically want the same thing: plan first, write some docs, then code.
Fair point on whether all this prep work is worth it when everything changes so fast. I think it is though - we're basically learning how to talk to AI assistants effectively, and that skill carries over no matter which tool you use.
About the cost - yeah, Claude Code isn't cheap. It's definitely the best right now (especially for refactoring stuff), so maybe see if your company will pay for it? Otherwise, check out Amazon's new thing called Kiro (https://kiro.dev/) - it's free right now and has Claude Sonnet 4. Gemini's CLI(https://github.com/google-gemini/gemini-cli) worth a shot too, though honestly it's not quite there yet in my opinion.
In any case, the real skill is learning how to work with AI, not just learning a specific tool. That's gonna be valuable whether we're using Claude Code, Kiro, or whatever shiny new thing drops next month 😄
Thanks for sharing. It sounds like a reasonable approach. I would try these technics for sure.
Sounds like a lot of ground work 🤔
With the pace of AI dev tools evolution I wonder if this ground work will pay off.
And I am really curious if this approach works with other AI tools. An entry ticket price for Claude Code is high. I would expect quite many folks would be more keen to experiment with free options like Gemini.
Yeah, you're totally right about how fast these tools are evolving! The good news is this approach pretty much works everywhere - I've been testing Claude Code, Gemini CLI, Warp 2.0, and the others, and they all basically want the same thing: plan first, write some docs, then code.
Fair point on whether all this prep work is worth it when everything changes so fast. I think it is though - we're basically learning how to talk to AI assistants effectively, and that skill carries over no matter which tool you use.
About the cost - yeah, Claude Code isn't cheap. It's definitely the best right now (especially for refactoring stuff), so maybe see if your company will pay for it? Otherwise, check out Amazon's new thing called Kiro (https://kiro.dev/) - it's free right now and has Claude Sonnet 4. Gemini's CLI(https://github.com/google-gemini/gemini-cli) worth a shot too, though honestly it's not quite there yet in my opinion.
In any case, the real skill is learning how to work with AI, not just learning a specific tool. That's gonna be valuable whether we're using Claude Code, Kiro, or whatever shiny new thing drops next month 😄