Nov 11 5 min
Glue work is still harmful in the age of AI
Nov 08 5 min
LLMs are terrible at designing AI systems
Nov 06 5 min
A database is a store for global variables
Nov 03 5 min
LLMs are not superhuman at reading and writing neural network weights
Nov 02 5 min
Reliably aligning LLMs is impossible in principle
Nov 01 5 min
Fast and slow ways to improve your reputation
Oct 30 5 min
Elegance and intelligence are anticorrelated
Oct 29 5 min
You can just ask AI to do things
Oct 28 5 min
Getting good at AI takes time and practice
Oct 25 5 min
Don't anthropomorphize big tech companies
Oct 23 5 min
Software is slow because of bad design, not slow languages
Oct 20 5 min
Laypeople support technical fields for technology, not art
Oct 19 5 min
What requirements don't you have?
Oct 18 5 min
Writing is no longer proof of work
Oct 16 5 min
AI is driving a boom in modestly successful software projects
Oct 12 5 min
Software engineers don't get to be Luddites about AI
Oct 11 5 min
Don't diagnose your opponents
Oct 10 5 min
Software's centaur age may last decades
Oct 03 5 min
Superpersuasion will look like bribery
Oct 03 5 min
Shipping is the foundation
Oct 02 5 min
Do not build the LLM torture factory
Sep 27 5 min
Human-AI partnerships are for alignment, not capability
Sep 26 5 min
Advice to a beginning software engineer
Sep 25 5 min
You should all be asking way more questions
Sep 20 5 min
System One models like Jev can train their own replacements
Sep 20 5 min
Grit your teeth and ship it
Sep 18 5 min
Two techniques for working with System One models