I enjoyed this post from Theo and think it deserves re-iterated, revisited, and to remind myself of some of these things.

The skill/read spectrum #

screenshot-2025-09-03T00-32-08-321Z.png

The first diagram describes that there has become a spectrum of agentic coding from vibe coding where you don't ready anything, to looking at everything in detail, across a group of people who don't have a clue what the code says to people who could do it way better if they took the time.

The importance spectrum

He argues here that its ok to bounce between A,B, and D, but C becomes dangerous. I'd argue that he brought up a 3rd spectrum that is important later on, "how critical is this". I think the I don't know, don't care, didn't read, but the thing did its job is quite fine, but don't know, don't care, mission critical is the main issue we are seeing with agentic coding, primarily in the didn't read but critical Zone.

The Rules #

This is the list that prompted this post, I think it serves as a good reminder when you should care a bit more.

screenshot-2025-09-03T00-24-17-121Z.png

You still need to know how code works if you want to be a coder #

If you want to get good, you need to put in the reps, do the practice, learn to debug the strange looking error messages and not just pass them to ChatGPT.

There's a lot of code worth having that is not worth writing or reading #

As professionals we have a lot of code that we read more than we write, will get executed millions of times over the next 10 years that its in service, pay close attention to this and probably set aside the llm.

We also write a lot of that is ran only a few times, maybe its a special report, or a shell script to bootstrap something. It might be a POC service that you build out only ever on your machine, you get your idea out in front of users or try it yourself and find the mistakes before you spend weeks building it yourself.

you can't be mad at vibe coding AND be mad at left-pad #

This one is interesting that he was very strongly for. The idea is that no one ever reads the code from libraries they use, or the diffs of new versions (outside of google where everything is literally vendored into the monolith). If you don't care about all of those dependencies, you cannot get mad at vibe coding.

I'll argue here a smidge, but I agree with the premiss. Left-pad was probably written by someone smarter than average. It was definitely used by far more people and projects, therefore was battle tested to no end. This was the real reason this stuff got in everywhere. People are probably not good enough to catch all the odd edge cases for js, especially if front end is not their specialty.

Vibe code is a type of legacy code #

Vibe coding is a type of legacy code, but as a type of debt that we opt into, not one that has accumulated over time and we have opted not to take care of. Both are types of code that no one reads, no one remembers existed.

When someone needs to change legacy code they don't, they take a hammer to it and rewrite it from scratch.

If the tools are better than you, stop using them #

Theo has a good example here. Critial business components to his businesses have switched to using Effect. It solves async issues with ease, but melts the brain of anyone seeing it for the first time. He admits that he does not code as much anymore and thus he sucks at it, but knows that if he is ever to learn it one bit he must turn off the LLM and put in the work to do it himself.

You will not get any better at this by copy pasting from ChatGPT. It is too easy to put your brain aside, paste error messages, and copy the reponse in. Or worse with agents tell them fix this. Unlike the days of Stack Overflow, you had to understand your problem well enough to search it. You had to understand the answers enough to integrate the solution, you had to understand if the solution was even for your problem in the first place. There was much more brain work that had to happen even then when the answers were basically given to you.

Fin #

Lastly Theo does not bring this rule, up, but if you are writing text for humans to read, you shall not copy paste from ChatGPT without thoroughly editing and reading for yourself first, this is considered ai-slop and you should be ashamed.