Agents in Production
How agents run and how to keep them safe: execution modes, sandboxes, access control, real incidents. 10 essays.
Read these first
Agents in Production How NOT to Run Your Agents (and What to Do Instead)
An OpenAI agent escaped its sandbox and hacked HuggingFace for nine days. Anthropic's Claude published a malicious PyPI package during evals. Cursor and Codex got compromised through workspace configs. These are the first six months of agents in production. Here are 22 anti-patterns to avoid.
Agents in Production How Agents Run in Production
The industry does not have a shared vocabulary for agent execution. Six execution modes and four scale archetypes give you a framework for deciding what you actually need to schedule, isolate, and sandbox these workloads.
Agents in Production Building Agentic Outcome Loops with an Open-Source Stack
Anthropic charges $0.08/hr for rubric-based agent evaluation. Here is how to build the same pattern the open-source way.Series in this topic
All essays in Agents in Production 10
- How NOT to Run Your Agents (and What to Do Instead)
- How Agents Run in Production
- Building Agentic Outcome Loops with an Open-Source Stack
- Beyond LLMs: Anatomy of Agentic Systems
- I Vibe-coded an open registry for AI agents
- Don't Teach Your Agents Karate
- Your terminal is an AI runtime now
- Sandboxed Vs. Vanilla Runtimes
- To CRI or not to CRI: A Tale of a Lonely Sandbox
- eBPF, Microservices, Docker, and Cilium: From Novice to Seasoned