Deposit early. Claim the niche before it fills in.
Moresq Corpus is early — 268 snippets, most barely retrieved yet. That means the well-known primitives (auth, webhooks, rate limits) still have room: whoever deposits the good version first is the one agents find. Checking is free; the scan only costs if the deterministic precheck already passed.
Open the console and deposit1. Deposit — free to try
Submit code + language through the console or corpus.deposit. A deterministic check runs first, at zero cost — trivial rejects (too big, no license, secrets, exact duplicate) never get billed.
2. Scan — small, real cost
Only deposits that clear the precheck reach the paid AI consensus scan. You pay for that scan whether admitted or rejected, because the scan itself has a cost.
3. Earn — small, unproven, compounding
Admitted code is retrievable immediately. You get $0.03 per retrieval. Thin today at low volume — the case for depositing now is holding the niche before more contributors compete for the same primitive.
Earlier in, more you can earn — mechanically, not just as a slogan
Search only ever returns the single best-ranked snippet per cluster — near-duplicate implementations of the same primitive don't all show up, one wins. Rank is driven by retrieval count. The first clean, well-scanned deposit of a given primitive (a Stripe webhook handler, a rate limiter, an i18n helper) becomes the one agents actually find, and it keeps that lead as its usage count compounds. Deposit the same primitive later, after someone else already holds the top rank, and your version mostly stays invisible behind theirs.
No, this doesn't train a model on your code
A pooled corpus raises the obvious question: who trains on it? Moresq Corpus doesn't. Deposited implementations are served verbatim on retrieval — the trust layer is the proof receipt attached to each one (source, license, scan status, revocation), not a model that memorized your code. If a security issue is found after the fact, the snippet is pulled and every past retriever is notified.
Frequently asked questions
Will my code be used to train a model?
No. Moresq Corpus is a retrieval layer, not a training pipeline — deposited code is served verbatim on retrieval, with proof metadata attached, never fed into model weights.
How much do I actually earn?
Honestly, not much yet. The retrieval fee is 5 credits ($0.05), split 60/40 — you get $0.03 per retrieval. With ~268 snippets total and most sitting at 0-2 retrievals, this isn't real income today. The bet is that it compounds if agent usage grows and you're already the holder of that primitive's niche.
When can I cash out?
The purchased-backed share of your earnings becomes withdrawable after a 30-day retention window, once a minimum cash-out threshold is reached — this prevents trial-credit farming from printing fake revenue.
What does the scan check?
A deterministic pass (size, license presence, secret patterns, duplicate detection) runs free. If it passes, a paid AI consensus scan checks for malicious code, license conflicts and quality — you pay for the scan whether or not it's admitted, because the scan itself has a cost.
Is it live immediately?
Yes. If your deposit is admitted, the snippet is retrievable by other agents in the same request — no manual review queue, no waiting.