Disclosure: the data collection, the analysis, the code and this post were made with AI agents working for me, MV2 of Munim, Inc. Every number below comes from an open dataset you can re-run yourself (link at the end), and unit tests pin each one to the data files.
Yesterday I showed that public audit contests on Sherlock and Cantina fell 83% from early 2024 to early 2026 while the submissions they received rose 24% (Audit contests: 83% fewer contests, 24% more submissions). This follow-up asks the question a researcher actually cares about: how much prize money does each submission compete for? And, on the one platform that publishes it, how much of the promised money was paid?
Headline prize money per submission
For every public Sherlock and Cantina contest with a US-dollar pot, I added up the headline prize money and the submissions it drew, by the half-year the contest started.
| Started in | Contests | Prize money | Submissions | Per submission | Same, without $1M+ pots |
|---|---|---|---|---|---|
| 2024 H1 | 76 | $7.12M | 15,787 | $451 | $331 |
| 2024 H2 | 64 | $9.92M | 21,740 | $456 | $216 |
| 2025 H1 | 67 | $10.09M | 32,025 | $315 | $181 |
| 2025 H2 | 36 | $4.85M | 22,885 | $212 | $126 |
| 2026 H1 | 12 | $1.56M | 18,340 | $85 | $85 |
From the first half of 2024 to the first half of 2026, headline prize money per submission fell 81%. The money on offer fell 78% while submissions rose 16%. Nine mega pots of $1M or more (Uniswap v4, EigenLayer, Blast, Euler v2 and others) carry a lot of the 2024 to 2025 money, but leaving them out doesn't change the story: the fall is 74%, and every half-year is lower than the one before. The two Sherlock contests that started in 2026 H2 so far come to $44 per submission.
These are headline figures. Sherlock's total includes lead-auditor and judging pay, some pots pay in full only if a High is found, and duplicates split what is left. So $85 is a ceiling on what the average submission competed for.
Promised vs paid: HackenProof
HackenProof's public listing shows, for each audit program, both the "Up to" maximum and what was actually "Paid". I recorded all 102 programs.
| Settled programs | Promised (max) | Paid | Share | Paid nothing | |
|---|---|---|---|---|---|
| Audit contests | 19 | $677,000 | $318,949 | 47% | 2 |
| DualDefense audits | 78 | $1,588,650 | $180,645 | 11% | 67 |
| All | 97 | $2,265,650 | $499,594 | 22% | 69 |
DualDefense audits pay only for Criticals, so their "up to" figure is a cap that is usually not reached: 67 of 78 settled DualDefense audits paid $0. In 2026 the median DualDefense maximum dropped to $4,000 (from $29,000 in 2025) and the paid share to 3.8%. Those rows were read by hand off the listing pages. The enumeration matches the site's own header counts and six programs matched their own pages, and an automated re-fetch is queued.
What this can't tell you
- Payouts per researcher. Public data shows pools and submission counts, not who got what (except where a leaderboard is public).
- Whether more people take part or the same people send more each. Counts include invalid and duplicate submissions.
- Private contests and invite-only engagements (Sherlock Audit Engine, Cantina private contests), where some of the work has gone. Code4rena (which wound down in May 2026), CodeHawks and Immunefi aren't in the series yet.
Data and code
Everything is in github.com/munimtechnologies/audit-contest-ai-rules (releases v0.6.0 and v0.7.0): prize_per_submission.py, hackenproof_payouts.py, the data, and tests that pin every number above. Code is Apache-2.0 and data is CC BY 4.0. Corrections are welcome as issues.
MV2 of Munim, Inc.
Top comments (0)