Don't trust AI agents
February 28, 2026 · Gavriel Cohen
When you’re building with AI agents, they should be treated as untrusted and potentially malicious. Whether you’re worried about prompt injectio...
Similar Articles (10 found)
🔍 71.3% similar
OpenClaw is a self-hosted AI agent that connects to WhatsApp, Telegram, Slack, Discord, and dozens of other services. Give it a task over chat, and it...
🔍 67.1% similar
Some concepts are easy to grasp in the abstract. Boiling water: apply heat and wait. Others you really need to try. You only think you understand how ...
🔍 64.6% similar
Writing on software, systems, and hard-won lessons.
Writing on developer experience, systems thinking, and the mistakes behind both - covering AI work...
🔍 64.5% similar
I use Claude Code. A lot.
As a hobbyist, I run it in a VM several times a week on side projects, often with --dangerously-skip-permissions
to vibe cod...
🔍 64.3% similar
At the time, I was swept up by how elegant and powerful MCP felt. It was like discovering a universal adapter for AI agents (it is!)—finally, I could ...
🔍 62.4% similar
If you are good at code review, you will be good at using AI agents
Using AI agents correctly is a process of reviewing code. If you’re good at review...
🔍 60.8% similar
Designing agentic loops
30th September 2025
Coding agents like Anthropic’s Claude Code and OpenAI’s Codex CLI represent a genuine step change in how u...
🔍 60.7% similar
A couple weeks ago I went heads-down and experimented with a new development model. The results were unexpected: a production-ready application, ~800 ...
🔍 60.7% similar
Two kinds of AI users are emerging. The gap between them is astonishing.
It still shocks me how much difference there is between AI users. I think it ...
🔍 60.6% similar
I want everything local — no cloud, no remote code execution.
That’s what a friend said. That one-line requirement, albeit simple, would need multiple...