Work & writing.
A selection; the portfolio is ongoing. Several analyses were produced as public demonstrations of AI-accelerated work, and some client engagements are necessarily absent.
Analyses, start to finish.
Business and policy work samples — several written openly to show what AI-accelerated analysis can do.
Rethinking California housing from the premises up
A multi-part structural analysis that goes after the root incentives behind the crisis — land speculation, Prop 13, the tax treatment of housing, the cultural default of the suburban house, and procedural obstruction — rather than layering on more subsidies or mandates. Forthcoming.
Migrating a hospital’s BI platform to Azure
A community hospital weighing a $2.7–4.1M, 12–18-month move: costs, benefits, vendor lock-in and staffing risk, and competing stakeholder views, laid out before leadership commits.
Vaping regulation
A risk-proportionate case for U.S. nicotine-vape rules with strict youth enforcement — drafted in roughly 16 hours with AI assistance, as a demonstration of the method.
Manufactured food impacts
An analysis of the health and systemic effects of ultra-processed, manufactured food. Available on request.
How the work goes well.
Short essays on getting real value from an analytical engagement.
Executive sponsor engagement makes all the difference
Assign a director- or C-suite sponsor who meets weekly — to test ideas, keep the analysis on live problems, settle stakeholder disputes, and carry the knowledge forward.
Help me understand the organizational landscape
Map who truly holds decision power — formal and informal — and set clear routes for escalating uncomfortable findings, so good analysis shapes outcomes instead of stalling.
Weekly check-ins beat the big reveal
Replace one polished final presentation with regular reviews, so recommendations stay relevant and leaders help build them.
Ask for different stakeholder analyses
Commission separate analyses tuned to each key stakeholder’s priorities, rather than one generic report, so decision meetings reach consensus faster.
On AI, reasoning, and knowledge.
The research behind the practice. Read on the publication.
The End of Thought
How AI is teaching us not to think — and why a fluent performance of thought is not the same as reasoning.
Why RLHF can’t teach professional reasoning
What artifact-based evaluation conceals, and an apprenticeship approach built on expert inquiry and reasoning reconstruction.
Doxatropy: when markers displace quality
A behavioural class for systems that converge on the markers of quality instead of quality itself.
A high-inference training workflow
Aligning projects, tasks, trainers, and reviewers, with feedback checks, performance instrumentation, and incentive design.
Dex: a governed-corpus framework for domain-expert AI
Treating an AI’s knowledge corpus as a governed data asset — curated, classified, and admitted on authority before similarity.
Measured, not asserted.
Each major platform’s structured assessment of my AI-collaboration practice.
Ultra-expert tier
An assessment placing the collaboration practice in the ultra-expert tier, ~99.9th percentile.
~99.8th percentile
A percentile assessment, with the strongest gains credited to multi-agent orchestration.
97–99th percentile
Noted for system design, behavioural specification, and a multi-perspective analysis framework.