João Gonçalves
I build and run agentic engineering systems in production. Fifteen years scaling DevOps and platforms, now a 14-agent orchestration that ships full-stack features end to end.
Everyone's optimizing for speed. The more interesting question is what gets lost when you do.
I write about the engineering judgment AI can't replace, and the parts it already has.
Two Features, One Contract
BridgePort 3.0 shipped a Terraform provider and an MCP server. Most of the work went into the API contract they both depend on.
The Moat That Walks Out the Door
Part three. If the moat is the judgment in your team's heads, then the moat can quit. What it takes to keep that judgment from leaving with the people who hold it.
Rent the Loop, Build the Moat
Part two. If the harness is the moat, here's where the line sits between what you rent, what you build, and the discipline that keeps it yours.
The Harness Is the Moat
The model is the part everyone benchmarks. The reliability layer around it is the part that actually ships.
The Real AI Skill Isn't Prompting. It's System Design.
Most AI training teaches the wrong half of the job.
