Ooak Data

We pay for an anonymized copy of how work actually happens inside your company.

Frontier labs have run out of public internet. What’s left is how real work gets done: the tickets, the deal reviews, the incident chatter. We license that — non-dilutive revenue for data already sitting in your tools.

View the bounty board How it works

Why now

The public internet is spent.

Every major lab has already trained on it. The next public web page adds almost nothing new.

Unlicensed data now has a cost.

Anthropic paid $1.5B to settle a case over an unlicensed training dataset. Not a licensing deal — a settlement over infringement. It pushed the industry toward paying up front.

Books and articles were the last era.

Publishers licensed their backlists. This era is business workflows: how companies actually operate, day to day.

We're building Alexandria.

The largest library of real-world workflow data, training AI agents that can do real work. Y Combinator S26.

What we’re looking for

A Slack export alone, or a repo alone, isn’t a dataset. We need enough connected history to reconstruct how a workflow actually ran.

014+ connected tools covering one workflow (CRM, support desk, chat, docs)
023+ years of continuous history across those tools
03Enough coverage to reconstruct the workflow end to end, not one step of it

How it works

1
Intro call
30 minutes. We explain what we need; you decide if it's worth a closer look.
2
NDA
Standard mutual NDA before we go deeper on data or systems.
3
Data quality assessment
We map your tools and history against the spec — usually the CEO, with an AI coding tool.
4
Pricing
A firm quote, based on coverage, history, and current cohort demand.
5
Extraction
We pull, anonymize, and send a sample. Nothing ships until you approve it.

Total cost: half a day to two days of CEO time, 2–5 touchpoints. Usually just the CEO and an AI coding tool. Sometimes a CTO for an hour. No IT project.