A staff engineer's journey with Claude Code

Sometimes I spend half an hour writing a prompt and realize that I’ve basically rubber-ducked the problem to the point where I know exactly what I want, so I just write the code myself.

I have been doing my best to give these tools a fair shake, because I want to have an informed opinion (and certainly some fear of being left behind). I find that their utility in a given area is inversely proportional to my skill level. I have rewritten or fixed most of the backend business logic that AI spits out. Even if it’s mostly ok on a first pass, I’ve been doing this gig for decades now and I am pretty good at spotting future technical debt.

On the other hand, I’m consistently impressed by its ability to save me time with UI code. Or maybe it’s not that it saves me time, but it gets me to do more ambitious things. I’d typically just throw stuff on the page with the excuse that I’m not a designer, and hope that eventually I can bring in someone else to make it look better. Now I can tell the robot I want to have drag and drop here and autocomplete there, and a share to flooberflop button, and it’ll do enough of the implementation that even if I have to fix it up, I’m not as intimidated to start.

You’re coming at this from a highly biased and even angry position, which means I don’t think you’ll be satisfied with anything people can show you.

Which isn’t entirely unreasonable; AI is not really there yet. If you took this moment and said AI will never get better, and tools and processes will never improve to better accommodate AI, and the only fair comparison is a top-tier developer, and the only legitimate scenario is high quality human-maintainable code at scale… then yes, AI coding is a lot of hype with little value.

But that’s not what’s going on, is it? The trajectory here is breathtaking. A year ago you could have set a much lower bar and AI still would have failed. And the tooling to automate PRs and documentation was rough.

AI is already providing massive leverage to both amateur and professional developers. They use the tools differently (in my world the serious developers mostly use it for boilerplate and tests).

I don’t think you’ll be convinced if the value until the revolution is in the past. Which is fine! For many of us (me being in the amateur but lifelong programmer camp) it’s already delivering value that makes its imperfections worthwhile.

Is the code I’m generating world class, ready to be handed over to humans at enterprise sclae? No, definitely not. But it exists, and the scale of my amateur projects has gone through the roof, while quality is also up because tests take near zero effort.

I know it won’t convince you, and you have every right to be skeptical and dismiss the whole thing as marketing. But IMO rejecting this new tech in the short term means you’re in for a pretty rough time when the evidence is so insurmountable. Which might be a year or two. Or even three!

That seems to be a great example of precisely the sort of program an AI would be good at. A small focused product that only does one thing. Mainly gluing together other people's code. It's a polished greenfield project that does one tiny bit of focused functionality.

Interestingly, this guy has been making pretty much the same app as you, and live-streamed making it on youtube:

https://www.youtube.com/@RayFernando1337

Looks like he's now pivoted to selling access to his discord server for vibe coding tips as I can't find a link to his product.

But if we're honest here, it's not going to take a ton of code to make that. All the functionality to do it is well documented.

Many people here could make a competitor in a week, without agentic AI, just using AI as a super-charged SO. The limiter pre-AI (aside from AI transcribing it) would have been reading and implementing/debugging all the documentation of the libraries you're using, which AI is great at circumventing.

Your product looks really good, and is an excellent example of what vibe coded AI is great at. I hope you're getting good traction.

I wouldn't dive for these. Vibe coding is a slot machine - sometimes you get wonderful results on the first prompt, more than often - not. So, a cherry-picked example is not a proof it works.

If you want me to show an example of vibe coding, I bet I can migrate someone's blog to Astro with Claude Code faster than a frontend engineer.

> It should not be on a greenfield project, because nearly all coding is not.

Well, Claude Code does not work the best for existing projects. (With some exceptions.)

I picked these specific projects because they are a) mature, b) complex, and as a result c) unlikely to have development needs for lots of new boilerplate code.

I would estimate the majority of developers spend most of their time on problems encompassing all three of these, even if their software is not as meaningful/widely used as the previous examples. Everyone knows that LLMs are fantastic at generating greenfield boilerplate very quickly. They are an invaluable rapid prototyping/MVP generation tool, and that in itself is hugely useful.

But that's not where developers spend most of their time. They spend it maintaining complicated, mature codebases, and the utility of LLMs is much less proven for that use case. This utility would be most easily measured in contributions to open-source projects, since all commits are public and maintainers have no monetary incentive to misrepresent the impact of AI [0, 1, 2, ...].

[0] https://www.businessinsider.com/anthropic-ceo-ai-90-percent-...

[1] https://www.cnbc.com/2025/06/26/ai-salesforce-benioff.html

[2] https://www.cnbc.com/2025/04/29/satya-nadella-says-as-much-a...