Independent. Public since October 2026. Prior history: not for a README.

This page carries what I've decided to release. The rest — the queue, the scoring model, the undisclosed entries in the index — remains off the record.

Rule: no claim without an artifact. Everything stated reproduces.

 fleet        12 libFuzzer harnesses · AFL++ CmpLog · guided MSan scope
 lineages     release · git-master (daily rebuild) · AOSP vendor forks
 sanitizers   ASan+UBSan everywhere · -fsanitize=memory on a full
              instrumented dependency tree
 scheduling   systemd scopes, cgroup CPU quotas 140/40/20/20, two cores
 refresh      timer-driven git pull → incremental build → relaunch
 triage       normalized stack signatures → dedup buckets →
              append-only index · timeouts replayed as a separate class
 telemetry    per-target 8-bit counters, 60s cadence
 uptime       self-healing; survives SSH drops, fuzzer exits, scope death

→ fuzzing-infra — source, harnesses, methodology.

• AOSP standalone builds — external/ forks compiled off-tree via a stub layer for the Android build environment (log/log.h, cutils, media interfaces, errorWriteLog); source lists and cflags lifted from Android.bp. The exact bytes devices ship — under sanitizers.
• Divergence measurement — shipped-vs-upstream diffs across external/: libexif pinned at 0.6.21 (2015 baseline), FLAC 1.4.3 vs 1.5.0 parser-hardening delta. Measured on more libraries than listed here.
• Write-path fuzzing — HEIF decode → x265 re-encode → box serialization → reparse → decode. The least-exercised machinery in mature libraries.
• Corpus engineering — conformance vectors, upstream regression inputs, a structure-aware HEIF box mutator, frame-splice seeds that force allocate_output_ realloc transitions in the FLAC decoder.
• Signature-dedup triage — 778 artifacts → 97 signatures; findings classified by reachability and re-confirmed under MSan where it matters.

Public count: one. Index count: higher. Timing is a choice.

• Android's external/ libraries: hunting n-days in code that upstream already fixed — dev.to · Medium
• The Illusion of Progress: how a typo in my custom mutator sent my fuzzer into the wrong parser — dev.to · Medium · Хабр (RU) — postmortem: data[0] % 6, three weeks, all metrics green
• Interview Q&A · RU

Vulnerability research / product security — remote. Serious inquiries get personal answers: [email protected].