OpenAI's coding agents have significantly improved their efficiency, now completing 46% of research tasks that previously took humans two to four hours, a substantial increase from 23% in March. This advancement has led to an elevenfold rise in agent time per researcher. However, the success of these agents is varied, with high performance in monitoring training runs (79%) but lower rates in writing inference code (25%). The daily API credit consumption per worker has also surged to $2,400. AI
IMPACT Increases efficiency for AI-assisted research tasks, but highlights significant operational costs.
RANK_REASON This is a product update/improvement for an existing AI tool, not a frontier release or significant industry event.
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →