Blog
Posts on AI memory and agents, A/B testing, analytics practice, entrepreneurship, and selected industry topics.
A technical argument that abstention - explicitly refusing to answer when evidence is insufficient - is the most important property of trustworthy AI memory, why the field's incentives trained systems to guess instead, and what good abstention engineering looks like.
Issue #1 of the AI Model Radar digest: verdicts on Moonshot Kimi K3, Thinking Machines Inkling, Tencent Hy3, NVIDIA Nemotron-Labs-Audex-2B, and Ternary Bonsai 27B - what is independently verified, what is still vendor claim, and which model is worth an evening of your time.
A manifesto for AI memory portability: why single-vendor open formats fall short, what a neutral interchange standard requires, the GDPR and EU AI Act case, and the one question enterprise buyers should ask every memory vendor.
An honest technical explanation of why current AI agents have no persistent memory, what that costs in production, and what better agent memory architectures would actually need to solve.
How subscription metrics - ARR, ARPPU, retention, LTV, CAC, and unit economics - connect into one system, with formulas, worked examples, and the decisions each metric informs.
A practitioner's guide to the 10 machine learning algorithms that handle the majority of real-world tabular, text, and classification problems, with decision criteria and production tradeoffs.
Data-backed analysis of startup success and failure: modern risk frameworks, market validation, the first-mover myth, Lean Startup evolution, and how AI changed the math.
How confidence intervals, Z-scores, and statistical significance actually work in A/B testing, with worked examples and Python code.
A plain-language explanation of A/B testing fundamentals — control groups, test groups, confidence intervals, statistical significance, and common pitfalls.
Five practical tips on projects, resumes, and recruiter conversations for data science roles.