Why do most tech subs seem to hate Claude so much?
I started using Claude early in university, long before I became a founder.
At first, I was skeptical. Like a lot of developers, I thought: there’s no way an AI can actually understand a real codebase.
Then I started using it seriously.
Today, as a founder, I can work across pretty much any language or stack. And honestly, Claude has completely changed the way I build software.
The key isn’t just asking AI to “write code.” You have to give it the full context, let it understand the architecture, test what it produces, and give it immediate feedback.
And when something breaks? You can give Claude the error, the context, the logs, and ask it to investigate and debug the issue.
Yet whenever I browse software engineering communities, I constantly see:
“Claude broke our entire system.”
“It’s impossible to debug code you didn’t write.”
“Claude is way too expensive.”
“It writes the worst code I’ve ever seen.”
And I genuinely don’t understand some of this criticism.
If an AI-generated change breaks production, isn’t the developer still responsible for testing it?
If you don’t understand the code Claude generated, maybe the problem isn’t that Claude wrote it, maybe it’s
how you’re using the tool.
And the cost? I’ve shipped an insane amount of code with Claude, and the total cost has been around $1,000. Compare that to the cost of even a junior developer.
Can Claude write bad code? Obviously.
Can it write better code when you give it the right context, constraints, and feedback? Absolutely.
The biggest change isn’t that AI can code.
It’s that the role of the developer is changing.
We’re moving from:
“I write every line of code.”
to:
“I define what needs to be built, guide the AI, review the result, test it, and iterate.”
That’s a completely different way of engineering software.
And yes, I’m absolutely aware that AI could eventually replace parts of what I do.
But right now, it also makes me dramatically more productive and lets me build things that would have required a much larger team a few years ago.
Sometimes I wonder if the community is just a massive doom-and-gloom bubble.
Because the difference I see isn’t really between developers who use AI and developers who don’t.
It’s between people who are willing to adapt to a completely new way of building software — and people who keep trying to work exactly like they did before.
How has AI actually changed the way you work as a developer?
Replies
The $1,000 is doing a lot of work in that argument. What people complain about isn't the tokens, it's the two weeks before anyone notices the thing that looked fine and wasn't. We run four providers behind one chat, the adapters were agent written, and the tests were written in the same session. They passed for weeks while asserting the shape our own code produced instead of what the provider actually sends. Nothing broke loudly, which is the whole problem, and more context in the prompt doesn't fix a test that was never pointed upstream in the first place.