How we think
The frameworks we use on real engagements - named, measured, and specific enough to argue with. Every piece states a position.
AIThe Three Disciplines of AI Engineering
AI is not one workflow - it's three, and the team that names which one they're running knows exactly what autonomy they're granting and where the review boundary sits.
AIAI Code Nines
An error-budget model for AI-written code: the more AI per cycle, the faster quality burns - and the human code budget is where the value is.
AIContext and Session Management
The context window is a capacity plan and the session is disposable - the model cannot compartmentalize, so you do it for the model. Capacity bands, context rot, and session abandonment as a discipline.
The Three Disciplines, discussed
The disciplines walked end to end: the autonomy spectrum, why each role enters AI at its own boundary, and the onboarding punchline. The commute version.
Read the piece →
AITemplate-Based AI Engineering
The model doesn't design your architecture - it instantiates one you've already verified. Why a curated template library outperforms model upgrades and fine-tunes.
ProductValue Stream Mapping for Product Definition
From a one-line pitch to a full board: experiences, activities, components - then personas, interaction moments, churn risks, and what you are NOT.
EngineeringThe Quality Drawdown
Measure code quality like a portfolio: burn and replenish cycles, moving windows, and the golden/death cross - data already in your git log. Not AI-specific; it works for any team.
AIThe Measurement Problem
Net LOC is misleading and "LOC is garbage" is also wrong: absolute, net, and ratio together show a constant engine in different gears.
AIAI Impact on Career Ladders
Career ladders give you a level; only staff level gets a shape. AI adds a new axis - distance to code - and ladders fork at intersections nobody has named yet.
AIVertical Scaling, or "I Stopped Coding Six Months Ago"
When a staff engineer earns that sentence, the org just scaled vertically for the first time. The gates that make it responsible - and the 3% error budget that never reaches zero.
AIThe Stabilization Pass
Every stretch of AI velocity accrues invisible debt; the cleanup cycle isn't overhead, it's where patterns get codified.
AIA Working AGENTS.md, Annotated
The onboarding file we run real engagements with - section by section, with the reasoning behind each rule.
AIContext Modeling
Documentation written for the model, not the human: capped, table-driven code-reading guides - tested by running real tasks with and without them.
AIThe Session Review
When a PR lands wrong, the diff tells you what happened. The session tells you why - and which part of the loop to improve: the ask, the autonomy, or the review.
EngineeringDelivery Maturity for AI Enablement
Delivery standards existed long before AI coding tools - they now determine how much autonomy your team can grant on day one.
AIMake the Change Easy
Kent Beck's rule at AI speed: prepare the codebase mechanically, then land small changes in a clean zone - on a twenty-minute cadence with review inside the loop.
Want these frameworks applied to your team?
Let's Talk