Uncommon human data for Frontier AI
Data / Code
Engineers work inside real repositories, pinned to a specific commit. Every task ships with a reference solution and is graded against rubrics and model judges.
See Code →Data / Multimodal
Experts and everyday users complete real tasks on screen or in the real world while we record every step, thought, interaction and word. Sessions run from minutes to days, with prompts, assets and human references attached.
See Multimodal →Data / Audio & speech
Unscripted speech recorded solo, in pairs and in groups across up to 8 languages. Each session is segmented and transcribed, with separate tracks per speaker.
See Audio & speech →Data / Trajectories
Professionals work through real tasks with an AI assistant. We capture every turn, tool call and decision through to the final output, with structured ratings on quality and effort.
See Trajectories →Data / Custom
If you can describe it, we can capture it. We built our entire end-to-end tool stack, so we can capture anything in high fidelity.
How custom capture works →Samples
Start with samples.
Around 10 real tasks, custom or off the shelf, so you can check format, quality and fit before a pilot.
From the Data Review Studio
Human eval · Text-to-speech
VOICE-H
Real people judged AI voices against the original human recording, on real conversation, and gave the reason behind every preference.