-
Read, write, critical: permission tiers for a local AI agent
Every tool call my local AI agent makes is rated read, write or critical. Writes need my approval, critical actions need a one-time code bound to that exact operation, and unattended runs only propose.
-
Why local LLMs could not drive an autonomous pentest agent
I tuned an autonomous security-testing agent until the mechanics were clean, then watched my local 27B model read 1.44 million tokens and find none of three planted bugs. Here is how to separate fixable plumbing from a model that cannot do the job.
-
A chat-ops bot is a doorway, not a convenience
My infrastructure chat bot went silent for six hours because the router had blocked the messaging API's one IP. That and a self-heal webhook wired to nothing taught me how to build and test a privileged chat-ops bot.
-
Typed decisions from logits: LLM judgements that cannot fabricate
Reading a 9B model's option probabilities instead of its text took a disambiguation task from 22.7% to 86.7%, added a usable confidence, and removed fabricated reasons. Most of the gain was wording.
-
An MCP server for a private library, gated per user
Building an MCP connector so AI assistants can search and read my library: FastMCP's 421 Host trap, per-user tokens and role-filtered tools, reading long books, and gating restricted collections without regex bypasses.
-
An AI studio that makes films and 3D prints in Blender
My AI agents build printable models and short films in a sandboxed Blender over a tool protocol. The hard part is that a text-only model cannot see its own output, so it ships broken geometry as 'watertight' unless you give it sight.