On the
record.
How we think about agentic AI, software delivery, and building products that actually ship. No filler. Published when it's worth reading.
All articles
3 publishedWhy agentic AI changes what's possible in a 30-day build
Traditional agencies pad timelines against unknown unknowns. Agentic systems compress the routine work and surface edge cases early — making fixed-scope delivery actually reliable, not just theoretically possible.
Inside the ReAct loop: how our agents think before they act
A walkthrough of the reason–act–observe loop that powers every Northbound build — with real examples from production code.
PromptingPrompts you can steal: our fill-in-the-blank meta-prompts
The reusable prompt templates we use to scope features, design APIs, and catch edge cases before they become bugs.
Case StudyWhat building a mobile app with AI agents actually looks like
Day-by-day account of a 30-day React Native build: what the agent ran, what the human overrode, and one security issue caught before it shipped.
ProcessHow we scope a project in 20 minutes
The three questions that determine whether a project is fixed-scope-able — and what the manifest looks like when they're answered.
ProcessThe checkpoint model: where agents stop and humans decide
The three categories of decision that always pause the agent loop — and why reading the agent's reasoning beats reviewing its code.