WST: AI-Powered Knowledge Extraction Pipeline for Technical Books
How I built Wan Shi Tong — a pipeline that reads technical books and extracts structured, actionable knowledge into a FAISS vector store for semantic search.
Project write-ups, security research, development logs, and technical deep dives.
How I built Wan Shi Tong — a pipeline that reads technical books and extracts structured, actionable knowledge into a FAISS vector store for semantic search.
Track Claude, Codex, and Gemini usage in a single desktop dashboard — with multi-account support, progress bars, and a paranoid ephemeral password helper.
A per-user communication calibration system that builds a vocabulary profile from your message history and routes each Claude response to match what you actually know.
A policy and mission-control layer for terminal agents: classify commands, enforce hard denies, track mission drift, and keep long-running autonomous sessions bounded and auditable.
A Chrome extension that turns GitHub's 10+ click fine-grained token creation into a single click. Pick a repo, set an expiry, get a disposable credential that dies in days.
How I stopped 15+ daily hard reboots caused by dual RTX 3090 power spikes by building an async reverse proxy that enforces GPU power budgets, request queuing, and model affinity.
An open-source honeypot that plants fake AI tools, APIs, and config files to detect attackers probing for LLM infrastructure on compromised systems.
Building a Chrome extension that uses machine learning and LLMs to intelligently skip intros, find topics, and chat with video content.
Setting up a comprehensive security lab for practicing offensive techniques and building defensive monitoring - from vulnerable VMs to SIEM integration.
Lessons learned building autonomous AI agents - from prompt engineering patterns to tool orchestration and real-world deployment challenges.