research
The engineering journal.
Practical notes from building AI systems, automation, and software in production. Thought leadership through evidence, not marketing content.
Agents that fail loudly: designing AI workflows for the bad day
Autonomous systems earn trust by how they behave when something goes wrong. Notes on escalation design, confidence thresholds, and why silent failure is the only unacceptable outcome.
MCP in production: wiring business systems to language models
Model Context Protocol moved integration work from custom glue code to a standard interface. Field notes from connecting real business tools to AI systems.
The solo stack: shipping production software as a team of one
AI-accelerated development collapsed the team size a real product requires. What that changes about architecture choices, and what it doesn't change at all.