AI Coding Advice

Jordan Hubbard, a Senior Director at NVIDIA and co-founder of The FreeBSD Project, shares his insights on AI coding -

When you tell an AI tool to "plan", you are working a different part of the model - or even potentially a completely different model - when you are telling it to "execute".   "Deep thinking" models are better at analyzing code and planning than they are at executing.  Sometimes even "dumb" models are better at executing to a plan than the smart ones, and they are certainly CHEAPER.

You MUST break planning from execution, just like you would do if you were writing the code yourself, and you must have the AI write planning files that it can follow.  If you or your devs are executing plans with the most expensive models, you're almost certainly just wasting money.

You must AUDIT the code for every feature cycle.  My "flow" is basically:

* Plan (write this as a .md file, save in plans/ directory)

* Execute (write the code)

* Write tests

* Run tests

* Re-execute as necessary until tests pass

* Audit - check the code against Plan.md

One branch per feature and a descriptive tag (for me to find later) whenever the app reaches a functionality / stability milestone.

It's not really making me more "productive" as much as it is making experimentation much, much faster. It takes me far less time to try things before rejecting or re-planning.

...like most coders over the decades I have found the amount of time I have for coding vs design and architecture has changed ratios. 

...like all good programmers I try to adapt to the tools and processes of the day.