Home › The gap, in 3D
756 occupations
The gap, in 3D
Left to right: how much of the job an LLM alone could do or halve the time of, per human raters. Up: how much of it shows measurable use in Claude conversations. Depth: the same “could” question answered by GPT-4. For 338 of the 756 jobs, observed use is under half of what the human raters said an LLM alone could do. That is the gap.
Drag to rotate, scroll or pinch to zoom, tap a point then “Open”. Dot size grows with the number of tasks. Axes run 0–80% (humans, observed) and 0–100% (GPT-4). Without WebGL you see the flat chart above, and every dot is a link.
Sources: task list from O*NET® Database (U.S. Department of Labor, CC BY 4.0; modified by noworkleft.com); “could” labels from Eloundou et al., GPTs are GPTs (OpenAI/University of Pennsylvania, MIT licence, labels from early-2023 models); “seen in use” from the Anthropic Economic Index (Claude usage only, CC BY). Dates, versions and limits.
O*NET 31.0, built 2026-11-06.