🤖AI Agents Now Finish Nearly Half of Research Tasks Solo, Study Finds
Your research team is about to look very different
TL;DR
Agents now finish nearly half of two-to-four-hour AI research tasks without help. Median researcher spend on coding agents hit $2,400 a day in October, up 62%. If you're running training or infra, your team's workflow is already shifting.
Here's the headline: AI agents now complete nearly half of two-to-four-hour research tasks on their own, and researchers are spending serious money to keep them running. The median researcher burned through about $2,400 a day on agentic tools in October, up 62% from September. The 90th percentile hit $23,000 a day, roughly double the prior month. If you work on training pipelines, inference stacks, or infrastructure, this maps directly onto your week. Agents hit 79% success monitoring training runs, 35% launching them, and 25% writing code for inference and serving. Monitoring is basically solved. Launching and writing serving code is where the human still matters, for now. The usage numbers are wild. Average agent workdays per researcher per day jumped from 0.5 to 5.7, an elevenfold increase. Over two-thirds of researchers ran at least four agents concurrently at peak, and more than 10% ran sixteen or more. Pretraining and posttraining teams led the climb, while Safety and Alignment saw the biggest absolute rise in human effort, with documented workweeks up 77% from January to September.

Key Points
Agents succeeded 79% of the time monitoring training runs, 35% when launching them, and 25% when writing code for inference and serving stacks.
Keep reading — it's free
Enter your email to keep reading — plus the best of AI & tech, daily. Free, forever.
Already a member? Sign in
Comments
Be the first to comment