For engineers

Get paid to push frontier AI past its limits.

Author rigorous, machine-verifiable problems on real production repos. Clear the rubric and your work becomes the training and evaluation data the top labs ship on. Reviewed in 24 hours, paid every week.

StacksPythonTypeScriptGoRustJava
A look inside your workspace: author, clear review, get paid. Click around, it's live
dev.mohitlabs.com/dashboard

Good evening, Alex

You have 1 task in review and feedback on 1 more.

128
Credits
14
Accepted
$342
Earned
Engineer
Tier

Recent tasks

Activity

  • Task accepted
    paginate() final page · +$25
    2h
  • Revision requested
    Streaming JSON backpressure
    5h
  • Payout sent
    Weekly transfer · $47.00
    1d
  • Program unlocked
    Titan · system-level engineering
    2d
Work used by frontier AI teams Strict 11-criterion review Specific feedback in <24h Paid weekly by bank transfer
Programs

Pick a program. Author the work.

Each program targets a different kind of frontier data, coding tasks or memory-safety crashes. Reviewed on a fixed bar, paid weekly.

Spark

Free
Open

SWE-bench-style coding tasks on real production repositories.

Python
Reward
$20-$25 / task
Apply

Trace

Free
Open

Find memory-safety crashes in real C/C++/Rust/JVM repos, a PoC and a root-cause fix per bug.

CC++RustJVM
Reward
$20 / task
Apply
Review

Judged on a rubric, never on vibes.

Every submission gets a clear verdict and the reasons behind it, typically within 24 hours.

Accepted

Clears the rubric. A payout is created automatically and the task enters the dataset.

Revision requested

Close, specific changes are called out. Address them and resubmit.

Changes required

Needs rework on substance before it can clear review.

Rejected

Doesn’t meet the bar, with the reasons why, so your next one lands.

Eleven criteria · all must pass · machine-verified by tests, no silent rejects.

Why it's different

Not freelancing. Not open source.

Built for rigor, scoped for quality.

Rigor, not volume

Eleven evaluation criteria. All must pass. The bar is the product, that is what makes the data valuable.

Machine-verifiable

Every task ships with tests. Correctness is decided by execution, never by vibe.

Gated, not a queue

Access is approved or invited and scoped to a real roster. Fewer, stronger contributors; no spam.

Real downstream

Accepted problems train and benchmark frontier models, work with a destination that exists.

Earnings

Paid for what clears review.

Concrete rates, weekly payouts, and a rank that lifts your rate as you prove out.

$20-$75
per accepted task
Weekly
bank transfer
No floor
no minimum payout
Rank ladder
ScriptDevEngineerLeadPrincipal
Eligibility

Who this is for.

We'd rather you self-select out than waste your time.

Apply if you

  • Ship production-quality code in Python, TypeScript, Go, Rust, or Java
  • Can design precise problems and write the tests that verify them
  • Work independently and take review feedback seriously

Don't apply if you

  • Want full-time employment or fixed hours
  • Need immediate, guaranteed income
  • Prefer guided, hand-held assignments
FAQ

Questions, answered.

What exactly do contributors do?+

You author rigorous, machine-verifiable coding problems on real repositories. Each submission is reviewed against a fixed rubric; accepted work becomes evaluation and training data for frontier AI teams.

How is this different from freelancing?+

There is no client backlog and no open spam queue. Access is gated, work is judged purely on a quality rubric, and the output has a real downstream, model training and evaluation.

How and when am I paid?+

Accepted work creates a payout automatically, paid weekly by bank transfer. Your rate depends on the program and difficulty, and rises with your tier.

What languages and domains?+

Python, TypeScript, Go, Rust, and Java across SWE-bench-style fixes, system-level engineering, public-repo tasks, and Dockerized terminal-agent tasks.

How does access work?+

Apply with your background; approved engineers are placed in the right program. Some programs are invite-only and scoped to an explicit roster.

Ready to contribute?

Applications are reviewed within 48 hours.