Research log
Page 2 of insights on application security, threat modeling, and building AI-powered applications.
The four controls I set up before my AI agents touch real work: least privilege connectors, tool output as data, network isolation with Tailnet Lock, and human approval gates.
The first build-in-public scoreboard for Agent Wins: what I shipped solo with Cursor and Grok, and the real numbers four days after launch, including 0 sales.
Agent Wins collects AI agent use cases that actually worked, each with a verified public source and a copy-ready recipe for the agent you already use.
From OWASP Top 10 updates to MCP vulnerabilities and agentic AI risks—what security teams need to know and do now.