How to Give AI Agents Real Access Without Risking Your Data
AI agents can delete databases. Here's the governance architecture that lets you give agents real access to your business and sleep at night.
Archive · page 2 of 16
Everything I've written on Little Might, going back to 2013. The first page is always the homepage; this is the long tail.
AI agents can delete databases. Here's the governance architecture that lets you give agents real access to your business and sleep at night.
How I rebuilt my AI agent team after Anthropic cut off OpenClaw. OpenClaw as the runtime, Codex as the coding brain, Paperclip as the work layer, and local Claw adapter agents to keep Claude in the loop.
The Codex skill routes Claude Code plans to OpenAI's Codex CLI for automatic review before you approve anything. Here's why I built it and how it works.
How to connect Claude Code to your iOS simulator using the ios-simulator MCP server. Set up in 15 minutes and let Claude run your bug and UX checks before you ever open the simulator yourself.
g-brain (gbrain) is Garry Tan's open-source knowledge layer for AI agents. A founder's plain-English breakdown of how it works with OpenClaw and gstack.
Garry Tan open-sourced gstack — now a fast-moving workflow stack that spans Claude Code, OpenClaw, and other agents. Here's what it is in April 2026.
Every Codex CLI command, safety setting, model pick, and prompting pattern I actually use now that GPT-5.4 is here.
Generic AI automation lists are useless. These are 9 specific business cases with the exact tool, setup, time saved, and dollar impact, from someone who's actually done them.
After running OpenClaw since January 2026 in a real business, here's an honest review, what it does well, what's frustrating, and who it's actually for.
Not theoretical. These are 12 things I actually use AI agents for in my business, what they do, how they're configured, and what they save.