“Code was never the hard part” is an insult to all programmers

LLMs are the invention of the loom. Code is being industrialized.

Yes, there always will be artisanal weavers and a smaller number of them get paid a lot more money to do this by people who can afford it. Everyone else either started operating a loom or did something else.

Automated looms are now making mass amounts of textiles but the artisans are no longer doing the physical act of weaving. The artisans come up with cool designs, get feedback from customers, solve people's problems and outsource the rest (physical labor).

We are in the loom moment. Are you designing things people want? Are you doing the weaving? These two paths can coexist but they are diverging disciplines with diverging difficulties and diverging value.

Typing code into a computer is rapidly becoming physical labor now. The layer of creativity and problem solving is quickly rising to a level above the code since the code is a fluid now that comes and goes easily.

Coding is easy if you enjoy doing it, met plenty of professional programmers that hated it and never felt the desire to improve. To them it was a way to get a paycheck that paid reasonably well.

I agree, the whole 'developers don't code' AI defense was pretty ridiculous. I also concur that development is going to get a lot harder. Anyone that has successfully used AI coding tools knows that you can get massive productivity increases but now I have to figure out how to get a bunch of hyper active child like coding entities with memory deficiencies to build stuff without going off the rail and introducing massive security problems or building something completely mismatched to the requirements. If you can do it, yeah you can get 10x results. But now I am engineering harnesses, architecture specifications, agent structures, statistically sampling, and formal verification systems to guide the code instead of writing the code.

Code generation was never the hard part. Knowing wtf you’re supposed to actually write was always much harder and this hasn’t changed. It’s not hard to write a loop or an if-statement. It’s hard to actually solve a problem with them at production scale. Carmack’s Fast Inverse Squareroot is not hard to write - it’s hard to figure out how to even do one in the first place. Writing it out is the easy part.

Claude writes 90% of my code but the bottlenecks always were and still are:

* getting clear requirements from product

* getting the damn code reviewed so I can merge it

Neither of these are fixed. Frankly, overuse of AI has made both of these worse. Claude brained product owners going hog wild with Claude Design are a nightmare to deal with, and the volume of absolute trash quality code being submitted for review is soul crushing.

No, using AI to automate code reviews is not acceptable. Code Review isn’t about a systematic checklist (though they can help) it’s about making sure people understand the actual changes being made to the system because it’s people who are accountable for what happens in production. LLMs can be part of the process of reviewing code but they suffer from the same issues as any other chatbot based tools (hallucinations, context confusion or not enough context, getting bogged down in impossible code paths or other minutia, etc…) so you have to review that review carefully too. Human judgment is still king.

These days my job is primarily reviewing offshore slop and making sure it’s in a good enough state to merge. I’m doing merge and release management way more than actually coding (and it sucks btw because I actually enjoy coding with or without agents). If the quality of the code turned in for me to review and merge is any indication, engineers/system architects are going to have their hands full.

Yes, writing code that (appears to) work was not the hard part. Even before LLMs you could find very cheap contractors on Upwork or through offshoring companies to write "code".

Well-engineered code is hard. It still is. Code that is reliable, extensible, maintainable, scalable, legible, understandable is hard. Code where the specs and the "why" behind it were pressure-tested via thought and good intuition.