Software laws: a leadership reading path
A ranked starting point for teams, delivery, and modernization.
STARTING PATHS · FIELD NOTES · REFERENCE
Each note: the idea, how to apply it, and where to read more.
Searches titles, descriptions, and topics.
22 entries
STARTING PATHS
A ranked starting point for teams, delivery, and modernization.
A Kanban-based lifecycle for agent-assisted engineering, with a named owner at every phase.
Give each team a clear purpose, an explicit way of working with other teams, and work it can hold in its head.
Record the context, choice and consequences close to the code, including the shortcuts, and draw the system at the level its reader needs.
Stay close to the code, own the outcome, remove friction, and share what works.
FIELD NOTES
My own work: case studies, essays, and the APEL model.
Acceptance criteria are where testing starts. A feature is done when it works for the person who needs it.
Most work runs on patterns that worked before. For fundamental change, rebuild the problem from what is actually true.
I built an internal platform that joined delivery and quality signals, then used what it showed to raise the bar on how work reaches QA.
Two employers, one pattern: decompose the monolith behind stable service boundaries, and write down the existing system's rules before changing it.
Buying coding agents is easy. Getting engineers to use them well took hands-on leadership, shared context and verification that never depends on trust.
Hold a position firmly enough to act and loosely enough to change. Aim for the better answer, even when it isn't yours.
REFERENCE BY TOPIC
Add autonomy only where a simpler path can't do the job, and settle its limits, access and owner while the design is cheap to change.
A spec says what to change, context says how the system works, and a harness catches what goes wrong. Grow all three from real failures.
Use a model where a wrong answer is visible and recoverable, then divide what you spend by the answers people actually use.
Capture behavior, change a bounded slice, translate at its edges, and verify before expanding.
Use 1:1s for growth, and design the work so understanding still forms when an agent writes the code.
Connect telemetry to follow a failing request and test an explanation.
Strategy chooses the direction, the roadmap sequences the bets, and execution delivers. Take it upward by leading with the decision and its stakes.
When agents write more code than people can read, automate the checks, test at the lowest useful level, and save human review for the risky changes.
Speed and stability from the delivery pipeline, and feedback, focus and cognitive load from the people doing the work. Read them together.
Usage is easy to count and easy to misread. Measure outcomes against a baseline, with quality beside speed.
Make processing, queues, and handoffs visible from commit to verified production.
No entries match. Try a different search or topic.