Company profile · 2 min read
Zero Billion: A human-data marketplace built around consent
A human-data marketplace needs consent before it needs volume.
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
Published · Updated
A human-data marketplace needs consent before it needs volume.
What it does
Zero Billion presents a consumer research lab with a Marketplace marked Early Access and a separate Studio described as live on the App Store. The homepage’s Marketplace section says people can share audio, video and everyday work, see who is buying the data and what it trains before recording, then receive payment through Stripe. Its stated operating principles are specific: permissions are listed and revocable, identity is stripped by default unless a task pays for likeness, and deleting a contribution should remove it from the pipeline.
The company also links to a separate consumer studio with apps such as BookTok, LingoTok and RiverFlow. That makes the business shape clearer: a data marketplace and a consumer-app studio presented as two parts of one research lab. It does not establish marketplace quality, participant earnings or buyer outcomes.
Buyer and task
The buyer is a research or AI team that needs human data with a visible consent and use boundary. The contributor is deciding whether the task is worth the money and the exposure. The first task is therefore not “collect more modalities.” It is to make one task legible: who buys the recording, what it trains, whether identity is retained, and how the contributor can delete it later.
This is where the product can earn trust. A marketplace that tells people what happens before they record makes consent part of the product. A marketplace that hides the downstream use until after collection is just another data broker with a nicer interface.
The first test
Run one task from invitation to deletion. Verify the buyer description, payout, consent record, identity default, retention path and deletion confirmation. Ask what happens to copies already used in a model or dataset. Then repeat the task with a contribution that explicitly pays for likeness and check whether the consent language changes.
The acceptance criterion is a contributor who can answer three questions without contacting support: what did I submit, who is using it, and how do I take it back?
Sources checked
official Speedrun profile · company homepage · consumer apps
Sources checked — September 20, 2026.
About the author
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
I cofound Lazyweb and publish Mudpie. This is an owner-written publication, not an independent testing organization. Research notes distinguish observations, sourced reporting and editorial judgment.
First1000 ↗ · X ↗