Writing

Agile let the business see where work stood. An AI coding agent can hide it again in unreviewed work. How to keep predictability, transparency and quality.

Part two of the Argus story. How six years of option history were tested under a pre-registration the code enforces, why nine correct predictions still meant stopping, how a search for an edge closed 136 variants, and the fail-closed data discipline behind it. For quants and engineering leaders guiding AI research.

A trading-infrastructure veteran directed Claude Code to build, deploy and run a live options trading system under hard safety rails, then shut it down on the evidence. The constitution, the kill switch, the audit chain, the defects the process caught, and ten rules for CTOs letting AI write production code.

How the owner of a family restaurant computerized it, taught himself software, started building websites, and took his first programming job in 1998.

How to find and fix the hardest bugs: reproduce the problem, check configuration first, rule out every cause on the execution path, and ask another developer.

Principles for enterprise software architecture: collaborate, organize, form a vision, execute in short sprints, sustain, transform legacy systems and test.

Run your own work as a sprint: a personal backlog, one-week sprints, a daily review and a weekly retrospective, all kept in a simple spreadsheet.

Seven core principles and four key strengths for leading software developers, and how they play out in communication, integrity, mentoring and team building.

Successful software is no accident. A 2012 guide, revised, to the work beyond code: reviews, test plans, builds, releases, QA, estimates and support.