🌐 Personal site: aznikline.github.io

I build systems at the boundary of AI agents, quantitative research, and low-level engineering — and I learn by writing source-first implementations I can actually read end to end.

• 🤖 nanoagent — a source-first study of how agent capabilities accumulate: tool calling → memory → subagents → teams → context compaction → safety, one chapter at a time in ~100 lines of Python each.
• 📊 alpha-mining-system — a modular Alpha factor mining platform (genetic programming + DeepAlpha + WorldQuant-style expression engine) across A-share and US markets.
• 🛡️ OSINT & AI security tooling — actively contributing to maigret (3000+ site username dossier) and following shannon (white-box AI pentester).
• 🧱 oslab — an aarch64 microkernel with capability-based IPC, scheduler, and memory design (private, in progress).

My work has moved through a few areas, but the connective tissue is the same: take something usually hidden behind a framework, and rebuild it small enough to understand completely — visual trackers, RL agents, factor engines, and now LLM agents and kernels.

• 🔭 Experimenting with agent safety boundaries (Voight-Kampff empathy tests for AI agents) and context handoff patterns.
• 🌱 Sharpening systems programming on a from-scratch microkernel.
• 📫 Reach me via aznikline.github.io, GitHub issues, or discussions on any repo above.

_{More old work in sc2018, CC150, SSSRNet, YouTubeBB — the place where I started.}