
Using AI for Coding Workflows

Open-source, complexity & AI coding 🔧 — with Salvatore "Antirez" Sanfilippo
Topic 6 of 7
Published Apr 3, 2025 · Clip 34:13–50:15 (16:02)
Salvatore describes his massive daily use of AI for coding: he writes dense C code himself without completions, using Claude's web interface as a design colleague and for rigorous code reviews with multiple restated prompts. For Python tests he specifies behavior precisely and provides representative examples plus full source code as context rather than docs, and uses AI for documentation templating. He estimates he is roughly five times faster, explains that prompting skill and early adoption explain the gap with people reporting only 10-20% gains, and says people underreport AI help out of fear and ego. He gives examples like ground-truth vector set tests, threaded vector sets design validation, and explains he writes tests after implementation because he then better understands failure modes.