AI coding cost per output hour falls 15% in September after 4x climb
AccBalanced · x · 2026-10-17
Weave Index data shows an inflection in AI coding economics:
- Cost per expert hour of code output climbed from $2.80 in January to $11.00 in August — a 4x rise over 7 months, as token spend outpaced code
- September flipped the trend: cost fell to $9.39 (-15%), code per dollar rose 18%, the first month output grew faster than AI spend
- The author frames this as the start of an "efficiency era": engineers learning which tasks to hand off, how to prompt, and when not to use AI
- Caveat: one month isn't a trend
More from AGI Musings
- Podcast: quantifying trust as a metric misses that it is a dynamic human relationship — jathansadowski · 2026-10-17
- repligate: dunking on the alignment faking paper's framing isn't missing the point — repligate · 2026-10-17
- Garrison Lovely: "Companies are trying to replace us with machines" — no doom needed — GarrisonLovely · 2026-10-17
- Anthropic staff estimate AI already does half the work of building frontier models — felpix_ · 2026-10-17
- Analyst: the widening compute gap alone may end China's frontier-model catch-up — teortaxesTex · 2026-10-17
- AI fund satire: the capital class believes AI ends labor and class struggle because it pays — generativist · 2026-10-17