AI is not replacing programmers yet

Using AI makes me faster (in some areas). Let’s talk about writing code.

These days I use ChatGPT and Claude. I’ll explain how in a bit, but let’s first address GitHub Copilot.

I’ve found it to be nearly useless. The autocomplete feature gets things right just occasionally and is mostly a distraction. The ability to talk to it and ask it for code doesn’t work as well as asking ChatGPT 4-o directly.

The way I use AI now is that I give it a few files, craft a good prompt, and get a decent answer. I then tweak that back and forth between the editor and the LLM interface.

The process is a bit clunky, but the results are good.

For me, it works best when describing how I want the UI to look and behave and generating HTML + CSS (+ JS). That allows me to rapidly prototype.

It can also do API requests, boilerplate, and GitHub Actions decently.

There’s still a lot of work to make it production-ready.

Am I faster? Yes.

Can you go from a prompt to a production-ready app without knowing how to code in 2h? Hell no.

The main thing LLMs lack is deep reasoning abilities and a big enough context window.

The AI companies are working on both, but progress is slow. ChatGPT o1-preview is a bit better at reasoning and coding, but not magical.

Everyone is waiting for what the next generation will bring.

So no, AI is not replacing developer jobs yet, but it’s making us faster (at some things).

That’s why solution-based pricing is an even bigger advantage than before.

If AI makes you faster and you bill by the hour you’ll be losing money.

The way to earn money is to solve big problems, faster and price the value of the solution, not the work that went into producing it.

Yours,

Taj