Why I Wrote AI Coding Guidelines and You Should Too

Source: Dev.to

Hiring Experience

We recently hired someone we already knew well.
We’d worked together before, trusted him, and liked how he approached problems. He has that entrepreneurial mindset—he sees something that needs doing and just does it. That’s the main reason why we hired him.

He isn’t ultra‑experienced and uses Cursor a lot. That’s not a problem, because we do too. Cursor is a great tool and it clearly accelerates development. Speed is the holy grail of every startup.

What we started noticing, though, wasn’t about speed or output. Things worked. PRs went through. But explanations sometimes got thinner. Changes were faster, but when there were questions, the reasoning and explanations were not explicit.

Not just for him – I felt that applied to all of us.

AI makes it very easy to move forward without fully thinking through what’s happening in the code. It’s a double‑edged sword and you shouldn’t use it carelessly.

The Problem Isn’t AI Usage — It’s AI Dependency

There are many horror stories of corporations outright banning AI use in their teams. This isn’t about limiting anyone or telling people to “use AI less.” That misses the point.

Of course AI can sometimes spit out bad code – so‑called AI slop – but that’s not where the risk lies.

The risk is slowly losing the habit of thinking, reasoning, and being able to explain your own decisions.

If you can’t explain what a piece of code does and why it exists, it doesn’t matter how fast it was generated.

Why I Wrote the Guidelines

I wrote the AI guidelines mostly for myself — and for us as a team.

• Not to slow anyone down.
• Not to restrict tools.
• To make sure we don’t outsource thinking.

The core idea is simple:

You’re always responsible for the code you ship, even if AI wrote it.

AI is great for:

• Accelerating implementation
• Reducing boilerplate
• Helping explore solutions

It should not replace understanding.

What I Want to Preserve

We hire people because they think, reason, and take ownership. AI should boost and amplify that, rather than slowly erode it.

The guidelines exist to keep us intentional and capable of explaining our systems six months from now, not just shipping something that “works today.”

That’s it.

Almost forgot, here is the link to read the actual guidelines: