Security researcher and founder of Accretion, with four years auditing Solana programs. The security mind behind hackhack: what the engine looks for, and how it decides a bug is real.
on-demand solana securitySolana security reviews.
Solana security reviews.
On your time.
Run a security review today. Get verified findings in a few hours, without waiting weeks for an auditor. Built for Anchor, native Rust, and Pinocchio.
real findings
Bugs we’ve already found, on protocols you know.
Confirmed findings from Solana security reviews and bug bounty research. Every entry below was validated locally.
1,000+
findings to date
| Detected | ID | Severity | Source | Verified | Target | Finding |
|---|---|---|---|---|---|---|
| 2026-04-01 | HH-4195 | critical | customer | ✓ | aggregator | An adapter leaked the protocol signer to a chosen program |
| 2026-03-13 | HH-4013 | high | bounty | ✓ |  Jupiter | A denial of service locked users in borrow/lend positions |
| 2026-03-31 | HH-4161 | critical | customer | ✓ | aggregator | Fake deposits, real withdrawals |
| 2026-02-27 | HH-3948 | critical | bug hunt | ✓ |  Ellipsis (Phoenix) | An unauthenticated transfer let anyone move user collateral |
| 2026-03-31 | HH-4159 | critical | customer | ✓ | aggregator | Cancel your withdrawal, drain someone else’s escrow |
| 2026-04-15 | HH-4361 | high | customer | ✓ | vault | A liquidator could force a trade on a healthy account |
| 2026-03-28 | HH-4128 | high | customer | ✓ | gated token | A wrapper hid a transfer outside the signed envelope |
| 2026-03-28 | HH-4122 | high | customer | ✓ | gated token | One signature approved two different instruction streams |
| 2026-03-31 | HH-4155 | high | customer | ✓ | aggregator | A public swap spent ten times the approved amount |
| 2026-04-15 | HH-4368 | high | customer | ✓ | vault | A refund pool paid one user out of another’s deposit |
how it worksRun it today.
Run it today.
Get results in a few hours.
Every reported finding goes through our verification pipeline. We check whether the issue is real and what it means for your protocol, so you can spend your time fixing bugs with a clear understanding of why they matter.
what a finding looks like
The bug, the affected code, its impact, and the reasoning behind the finding. Enough context to understand the issue and work on the fix.
benchmark80% of what a human auditor finds.
80% of what a human auditor finds.
In hours, not weeks.
We measure recall against findings that senior auditors confirmed on the same code. hackhack is not a replacement for a human review. It is the layer that gets you most of the way there before one starts.
Human auditorsbaseline100%
hackhackautonomous review80%
Frontier model, guidedwith a security prompt65%
Frontier model, rawsingle-pass prompt50%
Static analysisoff-the-shelf tooling18%
Recall against a fixed corpus of previously disclosed Solana vulnerabilities. Human-auditor recall is the 100% baseline.
the pre-audit engineAccretion runs hackhack
before a human audit begins.
Every Accretion engagement starts with a hackhack pre-audit across the customer’s repositories. The machine clears what a machine can clear, so senior auditors spend their time where human judgement actually matters.
80%
fewer bugs left to find
The pre-audit clears the mechanical findings before a human auditor opens the repository, so the engagement starts further along.
what that buys you
Auditors go deeper
The engagement is spent on design, economics and intent, not on the bugs a machine had already found.
A cleaner final report
Routine issues are gone before the report is written, so what is left in it is the substance.
Less back and forth
The simple bugs are fixed before the audit starts, which cuts rounds off remediation at the end.
You get a better audit, not a thinner one.
01
Your repository
Anchor, native Rust, or Pinocchio. Whatever you are shipping.
02
hackhack pre-audit
An autonomous review with verified findings your team can fix before the audit.
03
Accretion audit
Senior auditors start on the hard part: design, economics, intent.
standalone
Your repo. Your schedule.
Start today, without an audit engagement, an introduction, or a sales call. Submit your repository and get verified findings in a few hours.
Run a security reviewwith an audit
Or as part of an Accretion engagement.
Included as the pre-audit layer in every Accretion audit. Submit a review here first and book an audit later, and what you spent here is credited against that engagement, up to the cap.
leadershipAn auditor and an AI researcher.
An auditor and an AI researcher.
Leading the build.
hackhack is built by people who find bugs for a living, with a wider team of engineers and researchers behind them. The engine encodes how we audit Solana programs by hand, which is why it looks for the things that actually break protocols.
AI researcher with a PhD in deep learning, and an engineer first. Has built at Backpack, Anchor, MetaDAO and Marginfi. The brain behind the models and the evaluation engine.
pricing
One token. One review.
Buy tokens and start a review whenever you need one. Before a release, after a fix, or while you build. Every run includes the same verification pipeline, with results in a few hours.
1 review token$2,500$2,500 per review
One program, one review. How most teams start.
3 review tokens$6,000$2,000 per reviewsave 20%
An initial run, a re-run on your fixes, and a final review.
10 review tokens$15,000$1,500 per reviewsave 40%
Continuous coverage through an active dev cycle. Security you can plan around.
- Full repository review, all severities
- Verification pipeline for every reported finding
- Root-cause analysis on every finding
- Findings export as JSON or Markdown
- Wire the output into your own tooling
- A re-run after fixes costs one token
- Credited against a later Accretion audit (capped)
Your code is ready.
Your review can start today.
No waiting weeks for an auditor’s calendar to clear. Run hackhack on your time and get real, verified findings in a few hours.
Run a review today