Independent · Named · Fail-closed
Before your capital touches a venue,
someone should try to kill the idea.
I do adversarial, reproducible capital-safety diligence on perp-DEXs, vaults and DeFi venues — for funds, treasuries and builders who need a straight answer on where capital can be lost or trapped. I am paid to say KILL when the math says kill.
1
published validation-pass KILL — on my own trading idea
5
venues normalized, fail-closed
0.979
observed price correlation in the published KILL case
UNKNOWN
is disclosed, never silently treated as safe
Live · multi-venue
Not a description of the work. The instrument, running.
Funding for the same asset across venues, read live and normalized to a comparable annualized rate — Hyperliquid settles hourly, most CEXes every eight hours, so the raw numbers aren’t comparable until you fix that. The gap between them is where positions quietly bleed. If a feed is unreachable it reads UNKNOWN; it is never filled with a stale or invented number.
Published verdicts
The work is public, and it says no.
Most paid research says yes. These show the evidence actually collected, disclose what remains UNKNOWN, and attach the working needed to re-run the method.
Capital-Safety Scan: Rysk RFQ option markdowns
A validation pass that reaches KILL on my own trading idea before a full measurement run. It fails on capturability (the maker earns the spread; a read-only operator collects none of it) and on edge size — corr = 0.979, while the hedge costs more than any plausible markdown. Custody, withdrawal and settlement checks were outside this sample and remain UNKNOWN.
Read the scan →SurveyPerp-DEX capital safety: where your money is actually at risk
Six failure modes that move money out of allocator accounts — funding mechanics, vault drawdown, fabricated volume, illiquid rewards, venue shutdown, and unhedgeable funding — each priced against verifiable evidence.
Read the survey →Method
Why the verdicts hold up.
Kill-criteria first. Independent recompute. Fail closed. A person signs. The full operating method is its own page — not a footnote.
01
Payments-grade reconciliation
A background in telecom inter-carrier billing, where a silent mischarge is found by recomputing the other side’s numbers rather than trusting them.
02
DeFi domain depth
Settlement, funding and liquidation mechanics; custody and withdrawal paths; pause and upgrade keys; sequencer and oracle dependencies — read from the contracts, not the marketing page.
03
Research that argues with itself
A research pass, then a disconfirmation pass that tries to break it, then an independent recompute. Every claim ships with a source you can check.
Engagements
Three ways to engage. Fixed prices, fixed scopes.
Entry
Capital-Safety Scan
$1,500
5 business days · one venue
- A focused go/no-go on whether capital can be lost or trapped — not a feature review.
- Settlement, funding and liquidation mechanics: the docs versus what the contracts and API actually do.
- Custody and withdrawal path: can you always get out?
- 3–5 named kill-criteria checked adversarially, each with a reproducible artifact.
Core
Diligence Report
$6,000–9,000
2 weeks · full pre-deployment
- Everything in the Scan, plus the economic model under stress: depeg, funding-regime flip, thin-liquidity unwind.
- Counterparty and maker-capture analysis: who actually earns the edge.
- Oracle and index reconstruction; admin-key and upgrade-path review.
- 15–25 page report, executive verdict, reproduction appendix, and a readout call.
Recurring
Standing Retainer
$3,500/mo
3-month minimum
- Two Capital-Safety Scans per month.
- Always-on monitoring of approved positions: config, oracle and admin-key drift.
- An async judgment line — "should we touch X?" answered in writing the same week.
- A monthly one-page integrity digest across everything watched.
Book
Want this on your venue?
One venue. Five business days. A clean GO / GO-WITH-CONDITIONS / KILL with the reproducible work behind it — cheaper than being wrong once.
Prefer a full page? Open the booking desk →