I take apart software that does not want to be understood.

Malware, game engines, stripped binaries. I have done it in public since 2018.

Contact me GitHub

Photo: Yumeto, CC BY-SA 4.0, Wikimedia Commons

AI and agents

Agents do the typing. I decide what is true.

  1. Why it works

    Years of reading programs that lie taught me to check every claim. I check an agent the same way.

  2. I measure it

    In October 2026 I audited 270 of my own agent threads. 46% of the input tokens were old tool output being read again, so I rewrote the rules.

  3. I build with it

    Overcook puts an agent inside a LaTeX editor. Cosmo redeploys my projects on every commit. See the work

  4. I teach it

    In 2026 I started writing practical lessons on AI fundamentals, memory and agent systems.

Some people will hate this direction. I chose it and I own it.

  • I keep notes on every model I use: what it is good at and where it fails. Open the AI Lab

Journey

  1. My first program, in C.
  2. My first public malware write-ups. Haruko, my sample tracker, goes online.
  3. Haruko passes 2,600 samples.
  4. RM3, a report on banking malware, with Fox-IT and NCC Group.
  5. Tools for the Resident Evil games: randomizers and a speedrun tracker.
  6. My own AI revolution started.
  7. Creating practical lessons and courses on AI fundamentals and agent systems.

Skills

Reverse engineering

Finding out what a binary really does.

IDA, Ghidra, x64dbg, WinDbg, dnSpy, Process Hacker, objdump, x86 and x64 assembly

Shown in Anatomy of a packer

Malware and threat research

From the unpacked sample to a public write-up.

Unpacking, C2 protocols, config extraction, YARA rules, MITRE ATT&CK mapping, sandbox evasion, 14 public articles

Shown in the RM3 report

Software engineering

Tools I build, ship and keep running.

C#, Python, Lua, JavaScript, Hugo, Jekyll, tests

Shown in Work

AI and agent orchestration

Teaching AI fundamentals and building reliable agent workflows.

AI fundamentals courses and lessons, context and memory, embeddings, tool use, agent harnesses, model evaluation

Shown in the AI Lab

Game and file-format tooling

Reading and rewriting what a game keeps in its archives.

REFramework scripts, game archive formats, randomizer logic, speedrun tooling

Shown in the randomizer solver

Systems and infrastructure

How services are built, deployed and kept running.

Containers, reverse proxies and TLS, deployment automation, secure tunnelling, self-hosting. Tools: Docker, nginx, Caddy, Git, SSH, TeX Live

Shown in Cosmo

There will always be someone better. I am still going to be the best I can be at this.

Contact

A binary, a project or a question: write to me.

Contact GitHub