MIT Technology Review questions token billing as agentic coding costs climb

MIT Technology Review looked at measuring and billing AI usage by token and argued the focus should be on how much software gets delivered. Agentic coding tasks can use 10 to 50 times the tokens of ordinary chat, so under token billing the highest-value workflows become the most expensive to run. The report also says Uber reportedly used up its entire 2026 AI budget in April. The discussion centers on what agentic coding costs and on the gap between token counts and shipped software.

Written by AI from the coverage · updated 2 hours ago

Not enough continuous observations to draw a trend yet.