I build software at whatever layer the problem lives at. Most of my recent work is in LLM agent infrastructure. The repo list also spans game AI, security tooling, Unity, firmware forks, and cloud utilities.
An event-sourced Claude agent runtime I built that ships PRs to codebases across GitHub. Three processes share one SQLite file on a Hetzner VPS. It runs when I start it (not 24/7, API cost is a real constraint) and works through an operator-approved queue. Every interaction feeds a closed-loop learning system: successful patterns become teachings, failures become antibodies, hostile behavior tightens the fear-map.
Status: currently resting. Request a run.
| Repo | Merged | Latest |
|---|---|---|
Evaliphy/evaliphy |
3 | #32 feat(ci): add vitest code coverage analysis |
lingdojo/kana-dojo |
3 | #13003 content: add new trivia question |
AlphaSudo/zerodast |
2 | #86 docs: add CONTRIBUTING.md with full contributor workflow |
leticiv/tarot-reddit-compiler |
1 | #8 feat: add personal notes field to each interpretation |
adandeigor/AlgoLab |
1 | #9 feat: add --version flag to CLI |
chatvector-ai/chatvector-ai |
1 | #236 fix(logging): route test logs to separate files when APP_EN... |
Last 11 merges (most recent first):
2026-04-20leticiv/tarot-reddit-compiler#8feat: add personal notes field to each interpretation2026-04-20adandeigor/AlgoLab#9feat: add --version flag to CLI2026-04-20Evaliphy/evaliphy#32feat(ci): add vitest code coverage analysis2026-04-20Evaliphy/evaliphy#40feat: add vitest code coverage with GitHub Actions reporting2026-04-19lingdojo/kana-dojo#13003content: add new trivia question2026-04-19chatvector-ai/chatvector-ai#236fix(logging): route test logs to separate files when APP_ENV=test2026-04-19Evaliphy/evaliphy#31fix(reporters): improve HTML report assertions section UX2026-04-13lingdojo/kana-dojo#13015content: add new trivia question2026-04-13AlphaSudo/zerodast#86docs: add CONTRIBUTING.md with full contributor workflow2026-04-13AlphaSudo/zerodast#87docs: improve README badge and image alt text for accessibility (#83)2026-04-13lingdojo/kana-dojo#12944feat(theme): add Lucky Bamboo theme
CI / lint unblock: 2 merged
Evaliphy/evaliphy#32feat(ci): add vitest code coverage analysisEvaliphy/evaliphy#40feat: add vitest code coverage with GitHub Actions reporting
Bug fix: 2 merged
chatvector-ai/chatvector-ai#236fix(logging): route test logs to separate files when APP_ENV=testEvaliphy/evaliphy#31fix(reporters): improve HTML report assertions section UX
- mcplint: Rust CLI for testing, fuzzing, and security-scanning MCP servers.
- analytical-mcp and unified-thinking: two different cuts at giving models structured reasoning primitives (TypeScript, Go).
- bruno: neural behavior engineering framework, named after Giordano Bruno.
- zeek-yara-integration: network-security learning platform built on Zeek and YARA.
- trellis-runpod-worker: serverless worker for TRELLIS text-to-3D mesh generation.
- mesh-firmware: personal fork of Meshtastic firmware.
- battlecode2026 and javabot: MIT Battlecode competition bots.
- ec2sensor: Go CLI for live EC2 sensor metrics.
Roughly 15 more MCP servers across the ecosystem (Obsidian, Grafana, Bear, Roblox, macOS shell, GitHub, Exa, Langbase, and others). Open an issue if one looks useful and isn't obvious.
Whatever the job wants. Most often TypeScript, Python, Rust, Go. Occasional C#, Java, C++. SQLite as default storage.