Seungju Chae

Manifesto

There is an idea that keeps coming up at model labs: if the outcome of a task can be verified, AI can replace the human doing it.

To say a task is verifiable simply means there are criteria for judging whether it was done well or poorly. This is what we mean by a verifiable task.

Can the work you do, or the work your business does, be verified?

Verification may require taste or judgment, and its difficulty will vary. But if a task can be verified even in theory, I believe there is a high chance that models will take it over within a few years.

I’m working to grow the Task Economy which is an economy built around turning human work into these verifiable tasks AI can learn from.

I believe the economy of direct human labor will shrink, while the Task Economy will grow. As models take over human work, turning that work into problems AI can learn from will become increasingly important.

Economic scale
Task Economy and direct human labor A conceptual view of an economy where direct human labor declines and the Task Economy grows over time.

Companies creating these tasks and training data are already growing rapidly.

Annualized revenue of AI data companies
Mercor1 Gross revenue
  • 2025.09 $500M
  • Early 2026 $1B
  • 2026.06 $2B+
Handshake2 AI business, gross revenue
  • 2026.01 $550M
  • 2026.04 Nearly $1B
Snorkel AI3 Revenue
  • ~1 year earlier ~$20M
  • 2026.09 $350M+

Reported revenue run rates, annualized at each date.

This is revenue, not valuation. Mercor surpassed $2 billion in annualized gross revenue in June 2026, doubling in just four months.1

There is already plenty of good data for teaching AI. But much of it is sitting idle because people don’t yet know about the Task Economy. Turned into tasks, that data could be worth a lot of money.

I believe AI can take over any task whose outcome is verifiable and drive the shift shown in the graphs above.

We are at a turning point in history. The Industrial Revolution reshaped work and education. As AI takes on more work, we need education that prepares people to teach AI.

That’s why I want to make this market big. More people should have the knowledge and skills to take part in it. That possibility excites me.

Today, , I declare:

I will teach people to turn their expertise into tasks AI can learn from. That is where I will start. I will grow this industry, and I will lead this transition.

If you’d like to build this future with me, get in touch.

Sources

  1. TechCrunch, These AI startups are growing revenue at faster and faster rates (2026-07-08). ↩
  2. The Information, Handshake, Mercor Revenue Surges on Demand for Human Contractors to Train AI (2026-04-13). ↩
  3. Reuters, Snorkel AI valued at $3.5 billion in $350 million funding round (2026-09-22). ↩

These are annualized revenue run rates, not actual revenue earned over a full year. Mercor and Handshake report gross revenue before contractor payments; Handshake’s figures cover its AI business. Reporting dates are shown for each company.