Put the Guardrail in the Code, Not the Prompt
Hard limits for AI agents belong in code, not the prompt: the model reads prompt rules and breaks them. Plus the deny-by-default trap that does nothing.
I lead engineering teams.
Senior Engineering Manager at Staffbase — communities, mobile, and a plugin platform built for the agent era.
I ship with agents.
Specs, plans, reviews, phase gates — I manage AI agents the way I manage engineers.
I write about what that changes.
Agentic workflows, engineering leadership, and the craft in between.
Latest writing
All posts →-
-
AI Finds the Issue. Then It Throws the Hot Potato.
AI made finding code issues free, but owning the fix still needs a human. Why finding throughput is a vanity metric and the hand-back is the real work.
-
Simplicity Is Invisible by Design. That Makes It a Management Problem.
Engineering teams reward complexity by accident because simplicity leaves no artifact. Four ways a manager can make avoided complexity visible.
-
Meet APE — Agentic Product Engineering
We shipped a greenfield project in three weeks by building the system around the AI models, not the prompts. I call it APE: Agentic Product Engineering.
-
AI Engineering Workflows Are Built for One Orchestrator. Here's What Breaks at Team Scale.
AI engineering workflows assume one expert orchestrator. Here's what breaks when five engineers run agent sessions on one codebase — and our fix.