Unpopular opinion: AI coding tools are great at writing code and bad at leaving the rest alone.

Mine rewrote my config, renamed a folder, and changed how the app starts. The login page worked. The app didn t.

I ve been relying on AI coding tools a lot more over the last few weeks to speed up my workflow.

At first, it feels like a total cheat code. You hit autocomplete, or drop a quick prompt, and boom syntax errors vanish and boilerplate logic is done in seconds

But lately, I ve hit a wall.

The moment a bug involves a deep state-management issue, multi-file dependencies, or some weird framework edge case, things go sideways fast. The AI starts giving confident answers that turn out to be completely wrong. Before you know it, you ve spent 45 minutes tweaking prompts and reading broken generated code something that a standard console.log or breakpoint would ve caught in 5 minutes

Paid someone to do a code review last week, first actual engineer to ever see what I'd built. Watched her scroll for a solid two minutes without saying anything. Then: "okay, so, this works, but why does this file talk to three different places to do one thing?"

I didn't have an answer. I'd asked the AI for a feature, it worked, I moved on, asked for the next feature, it worked, moved on again. Never once looked at whether feature three quietly duplicated logic from feature one. The app functioned perfectly the whole time. It just also had, apparently, three slightly different versions of the same idea living in different corners of the codebase.

What got me wasn't that it was bad, she was actually kind about it, said this is completely normal for anything built fast. What got me was realizing I'd been measuring progress purely by "does it work when I click it," and that measurement had been quietly lying to me the entire time.

I don't think the answer is doing it slower from scratch. I think it's that at some point, someone who actually reads code needs to look at the whole thing, not feature by feature.

Every time I vibe code there is always this huge lift that I constantly have to go through. Authentication, billing, password resets, emails, signup, waitlist, landing page and when it s all said done and the app is ready then comes the marketing, the blogging, the social media automation, the product hunt launch etc etc etc . So much repetitive crap that I have to do just to get a simple app up and running. How do you guys handle all this?

Since I am a coder and a hammer sees everything as a nail, I decided to create all this code as a template so I can jump into building an app right away. There is actually a lot more than what I mentioned above e.g customer support, chat, roadmap for building in public, email flows and more coming.

I noticed something annoying with AI-assisted frontend work.

The agent could implement the requested feature correctly...

Most vibe coding advice says to have the agent write the tests alongside the code. I think that's the most expensive habit in the whole loop.

The code and the test come out of the same reading of what you asked for. If that reading is wrong, both are wrong in the same direction and the suite goes green, which is worse than having no suite at all, because now you've got a receipt.

Issue: The agent hits an error and hides it so it can deliver the output the prompt asked for.

Example: An agent swallowed a failing database write and returned a 200 success payload. The UI told users their setup was complete while the database sat empty.

Since I started running Claude Code and Codex in parallel worktrees, my disk fills up much faster. Every worktree gets its own node_modules and build output, session logs pile up in ~/.claude and ~/.codex, and Xcode DerivedData does the rest.

I've seen people find 100GB+ of this stuff. How do you deal with it? Do you clean worktrees manually, script it, or just buy a bigger SSD?

It took me a good while to figure this out. When my programming agent makes an error and I point out the wrong part, the correction doesn't vanish inside the chat log. It's saved as a file. So the next session never trips over the same issue twice.

So far the agent has written 100 incident notes. It also keeps memory notes in the project, one fact for each file. Each time a session starts, a one-line index of those notes gets loaded. If the same mistake happens again, it becomes more than a mere memory. It becomes a hook. And that hook runs on every tool call and every reply, regardless of whether the agent remembers it. I currently have 63 hooks.

A few of those hooks came from pretty ordinary situations. After long sessions in another language, the agent started replying in English. According to the note, the more English documents it had read recently, the higher the chance it would drift. So now, whenever a reply comes out in English, one of the hooks forces a rewrite.

Another time it claimed something was fixed without ever testing it. That led to a rule: it must write down exactly how it checked. Dash usage also got restricted.

The tools keep getting better at the first week. Describe an app, get an app. Impressive and the demos aren't lying.

What I keep hearing about (and lived myself) is month two. You ask for one change and five files move. Styling drifts. The AI forgets decisions from last week because the only place your app's structure exists is the chat history and the chat history runs out. Every builder community has the same complaints in almost the same words.

So, honest questions for people actually shipping with these tools: