VIGILX / LABINDEPENDENT
ADVERSARIAL
RESEARCHSINGAPORE
2026
SECURITY RESEARCH FOR CRITICAL SYSTEMS

Find the break.
Prove the impact.

For validator, consensus, and cryptographic teams: vigilX searches beyond a single AI pass, tests whether an attacker can reach the path, and delivers exploits your engineers can replay.

Explore the Firedancer case Meet the founding engineerJeremyah Joel
Discuss your system
From parallel leads to verified resultFour independent research paths converge at a verification gate, with one path emerging to a reproducible result.PARALLEL LEADSREPLAYED RESULT
SEARCHVERIFYREPLAY
Proven on:
  • Sei
  • Babylon
  • The Graph
  • ENS
01 / CASEFIREDANCER
V1REDACTED
RECORD
01 / FIREDANCER V1 · IMMUNEFI AUDIT COMPETITION

Firedancer V1.

An independent Solana validator client, researched under a competition that required runnable proofs for reported issues.

REDACTED LAB EXCERPT4

Confirmed labels visible

Three High, one Low. The hidden account and report names prevent independent attribution.

TARGETIndependent validator

Solana client written in C.

SCOPE636,000 lines

Competition brief's stated codebase size.

REWARD POOLUp to $1M

Listed across awards; payout depended on findings.

EVIDENCE BARRunnable PoC

Required for every severity.

LAB RECORD / REDACTED EXCERPT
Four Immunefi reports labeled Confirmed; three High and one Low
Four confirmation labels are visible. Report names and capture date are not shown.
EVIDENCE, WITH ITS LIMITS

What this record shows.

Immunefi independently documents the competition scope and proof requirement. The four-row result comes from vigilX’s redacted activity image; its hidden account and report names prevent independent attribution. The image also omits attack paths, impact, and fixes. We keep those details out of the public claim.

COMPETITION SCOPE ↗FULL LEADERBOARD ↗VIEW LAB EXCERPT ↗
FOR YOUR SYSTEM

We apply the same evidence standard to a scoped engagement: agree on attacker capabilities and a research window, then deliver a runnable reproducer, observed impact, and mitigation notes for confirmed findings.

DISCUSS A SCOPE ↗
02 / METHODRESEARCH
PROTOCOL01 — 04
02 / RESEARCH METHOD

From lead to finding.

Autonomous search widens the search space. A verification gate narrows it to attacker-reachable paths with replayed impact.

01

Scope the code

Build import and call maps, identify attacker inputs, then give each investigation only the relevant code.

FOCUSED CONTEXT
02

Search in parallel

Run independent passes over the same targets, deduplicate repeated ideas, and rank the leads.

RANKED LEADS
03

Challenge the path

Check the cited code, data flow, guards, and external attacker reachability before investing in a proof.

TRIAGED PATH
04

Prove and replay

Run a reproducer, repeat the result, and document observed impact for the engineering team.

VERIFIED FINDING
THE EVIDENCE BOUNDARY

A model output is a lead, not a finding. We require a runnable reproducer and replayed impact before handoff.

PUBLISHED FIELD NOTE

Jeremyah’s published article explains scoped code context, parallel investigations, retries, and verification gates. Its benchmark examples are author-reported.

READ JEREMYAH’S ARTICLE ↗
03 / RECORDADDITIONAL
ACTIVITYIMMUNEFI
03 / ADDITIONAL ACTIVITY

Research beyond one case.

A separate redacted Immunefi image shows ten Paid entries across bug bounties and audit competitions. Its capture date is not shown.

IMMUNEFI / REDACTED ACTIVITYCOMPLETE 10-ROW IMAGE
Full redacted Immunefi activity image containing ten Paid rows across bug bounties and audit competitions
Complete redacted capture; date not shown. OPEN ORIGINAL ↗
IMAGE TRANSCRIPT / ALL TEN VISIBLE ROWS

What the image lists.

The image labels all ten rows Paid. The list transcribes their visible severity, venue, and outcome labels; it does not imply ten unique confirmed bugs.

  1. 01MediumBug bounty
  2. 02HighBug bounty
  3. 03HighAudit competition · Duplicate of valid report
  4. 04InsightAudit competition · Chief Finding
  5. 05HighAudit competition · Duplicate of valid report
  6. 06HighAudit competition · Duplicate of valid report
  7. 07LowBug bounty · Mediation
  8. 08LowBug bounty
  9. 09MediumBug bounty
  10. 10LowBug bounty
04 / INQUIRYPRIVATE
CHANNEL OPEN
04 / DIRECT CONTACT

What are you about to trust?

Start with three things: the system or repository, release milestone, and constraints. We return a written scope with the target, attacker model, research window, and evidence standard.

A short technical note is enough to begin. Include any trust boundary you especially want challenged.

REQUEST A SCOPEx@vigilx.io