---

Research

Notes on what happens to software security when agents write the code. RSS.

• The Agent’s Permissions Are the Security Boundary

  September 3, 2026

  An agent’s model is only one part of its risk. Identity, credentials, tools, network access, workspace, and execution policy define what it can do.

• AI Changes the Pace, Not the Responsibility

  August 30, 2026

  AI changes how fast we build, not who owns the result. Agent-written code still needs evidence, review, and clear human accountability.