Writing Your Own Code Considered Harmful

[mayas_]

In modern, well-structured codebases with decent linting and tests, human-written code is often the weakest link.

Tools like Cursor, properly configured, consistently produce higher-quality code than humans—cleaner, faster, and bug-free. We’ve moved past the age of hero coders and into the era of high-leverage guidance.

The best engineers don’t write code—they steer it.

They shape context, define intent, and curate constraints.

Coding has become a meta-skill: less typing, more thinking.

Your job is to describe problems clearly, not solve them line by line.

Cursor doesn’t replace you—it promotes you.

Welcome to the age of the micro product manager.

[taylodl]

> In modern, well-structured codebases with decent linting and tests, human-written code is often the weakest link.

This is a statement written by someone having little work experience. It seems trivially obvious until you get to work and realize what the real problem is: incomplete and/or inconsistent requirements. Close on those heels are poorly documented APIs. Poor architecture and design round at your problems.

Writing the code? That's the easy part. Maybe that's what we should be emphasizing to people wanting to be professional software developers: the code itself has never been the hard part. That's table stakes.

Am I supposed to be impressed that AI has taken the easiest part of code development and has made it a little bit easier? Maybe? Don't forget I still have to create tests because I need evidence the code actually does what it's claimed to do. Which is ironic, because test creation and management is an area software developers really struggle with and now it's more important than ever!

[addoo]

> Humans

Please describe in more specific terms. Are we talking non-technical, intern, junior, or senior experienced humans?

Literally just yesterday I was diverted to help someone who is a senior developer, but a novice in Python itself, figure out why code that AI helped them write was completely busted. They were extra perplexed because most of the code was identical to a block of logic we use today (and works), but in this new context didn’t work at all. Turns out whatever the AI did, it didn’t have a concept of method overloads, so the types being passed around were suddenly wrong.

AI works well for people who know nothing (it can do things for them that work well enough), or people who know ‘everything’ (it can get them 95% of the way, they can use their experience to find and fill the remaining 5%). It’s absolutely terrible for people with middling experience.

[mayas_]

interesting

i've been writing code for my employers for the last 10years and i feel liberated

now i almost cheer when the product guys report bugs, mainly pre-llm legacy codebase i slice through like butter

well except on a few rare instances to be fair

i assume it can be counterproductive in the hands of an inexperienced dev?

[Juliate]

Sounds like very generated LinkedIn-ish: no substance, miraculous-PR-like, signs of lack of experience of the promoted solution's limitations in the discourse, and no call to action or discussion.

[ferguess_k]

> Tools like Cursor, properly configured, consistently produce higher-quality code than humans—cleaner, faster, and bug-free.

Just checked my phone. We are still in 2025, not 2035 or further.

[JohnFen]

> Tools like Cursor, properly configured, consistently produce higher-quality code than humans

This seems very doubtful to me. Do you have evidence?

[mayas_]

only anecdotal evidence

with gemini 2.5 pro combined with a good https://docs.cursor.com/context/rules directory i've had consistently impressive quality code,

i spent quite some time crafting the .cursor directory though

i tried to be as exhaustive as possible in capturing the subtleties of our codebase and what our team considers good taste

[thesuperbigfrog]

>> In modern, well-structured codebases with decent linting and tests, human-written code is often the weakest link.

I call BS.

If a human has never written the code, how will AI generate it?

If a human does not know how to verify that code works correctly, how can the code (regardless of who writes it) be verified?

Do you trust AI to write the code that controls the airplane you fly in?

Would you trust your life (literally) to AI-generated code?