The Infrastructure Mindset¶
About¶
This site documents what happens when an infrastructure engineer with forty years of operational experience starts collaborating with AI — and refuses to trust it.
The articles here are not AI-generated content. They are the product of a structured collaboration between a human author and AI research tools, where every factual claim is independently verified, every source is scored for reliability and bias, and the complete evidence trail is published alongside the article. The methodology is open source. The evidence is public. Check my work.
The subjects range from AI trust and sycophancy to research methodology, DevOps history, and infrastructure engineering. The common thread is operational thinking: how systems fail, why assumptions compound, and what it takes to build things that work at scale — whether those things are distributed file systems, engineering teams, or AI workflows.
About the Author¶
W. Phillip Moore is an infrastructure engineer with roughly forty years in the field. His career spans Morgan Stanley (where he was lead architect of one of the largest AFS installations on the planet and self-described "all-time global outage leader" — framed as a learning credential, not a brag), through decades of work in distributed systems, middleware, and operational tooling.
He has a consistent 25-year pattern: abstract a vendor's painful API into something maintainable by others. The MQSeries Perl module (CPAN, 1999 — the first open-source release from Morgan Stanley), AFS::Command, NetApp filer management tools, and pymqrest (PyPI, 2026) all follow the same template: wrap complexity in something a human can reason about.
He co-authored the USENIX LISA '95 Aurora paper on Morgan Stanley's next-generation global production Unix environment, and has presented at Decorum '97 and the AFS Best Practices Workshop at Stanford (2005). His 48-day empirical study of AI-assisted polyglot development — 8 repositories, 5 languages, 648 pull requests — produced the taxonomy of competence vs. discipline failures that runs through much of the writing on this site.
He is currently building the AI Research Methodology, an open-source tool for making AI produce defensible, auditable, evidence-based research rather than confident-sounding summaries.