Post

Simon Willison
@simonwillison.net

At the risk of starting the flame war to end all flame wars... Modern LLMs (GPT-5.1, Claude 4.5, Gemini 3) produce excellent code and can be a significant productivity boost to software engineers who take the time to learn how to effectively apply them - especially if used with coding agent tools

James Felix Black
@tft.io

Yes! They can do amazing things, but you cannot use them naïvely (note dieresis) and expect anything good. If you can provide them with invariants, ways to test those invariants, and understand how, eg, context windows impact state management, they're awesome.

James Felix Black
@tft.io

You probably get a lot of crap On Here but I enjoy your posts and get a lot out of your work so if you can stomach it, please keep doing what you’re doing.

James Felix Black
@tft.io

It's how you can constrain their possibility space, and inject your own judgement into the otherwise mechanical loop. I find them really liberating, honestly, and have used them to generate a bunch of tools that in turn I have the model use.