That’s awesome to hear! How’s the experience been for you? I’m curious about 2 things
- how do you keep track of the project and it’s milestones. Do you have a sense of how close to the finish line you are? How do you know what needs to be done next?
- when you run into places where the agent can’t quite get it right, what do you do?
I just started iterating. Claude makes good stylistic suggestions in addition to the code. So I gave a basic description to ChatGPT which spit out an image which Claude was able to reverse engineer. it got to work, and I’ve been giving notes ever since. Claude tracks the code block as its own artifact and names it by version
The only thing so far it hasn’t been able to do after some effort is dim the page when a video plays. Everything else it got after a couple tries at most.
It would be unthinkable to build a site this way with an actual person, changing the code as I get new design ideas. 300 versions. But it works perfectly fine with AI.
I connect big time with the build and iterate mindset. I feel like it’s the act of building that teaches me where to go next.
In the times when you can’t get something quite right, check out Second Opinion.
It’s an MCP that you just drop in. Then, when your agent can’t quite get a design right, it can reach out to other models to get help so you can move forward. 2nd0p.com
