Automating orchestration without automating judgment
Why durable state and verification mattered more than simply adding more AI workers.
Read →Ideas tested in real work.
Short essays on systems, finance, product, strategy, operations, and quantitative modeling—each grounded in a project, decision, failure mode, or operating lesson I actually worked through.
Systems & operations
How work stays reliable as complexity grows.
Product semantics, workflow design, communication, AI-assisted execution, and failure-aware systems.
Historical truth is a product requirement
Why identity, sync, history, and analytics semantics have to agree before a product can be trusted.
Read →Digitizing a workflow is not the same as improving it
What changed when validation, correction paths, and information reuse became part of the process.
Read →Making the work visible is part of doing the work
Why communication became a trust and control mechanism, not just a status update.
Read →Finance & quantitative work
What the model is actually saying.
Valuation, uncertainty, model governance, dependence, downside, and capital-allocation rules.
A model can run and still be wrong
How I learned to distinguish bad data, bad assumptions, bad implementation, and stale documentation.
Read →Why dependence modeling became the hardest part
Marginals, rank dependence, PSD repair, temporal persistence, and why complexity has to be earned by evidence.
Read →When financing constraints change the strategy itself
Why the rollout policy should sometimes change the realized operating plan—not merely the discount rate.
Read →A positive NPV is not the end of the decision
How sensitivity and downside analysis turn a valuation result into operating controls.
Read →When valuation methods disagree, don’t average them
Why FCFF, DDM, peer P/E, and market price were telling different economic stories.
Read →Strategy & decision design
Structured judgment without false certainty.
Decision models, responsible AI, activity systems, and robust choices when information is incomplete.
A weighted score is not a strategy
Why a ranking is useful only when the tradeoffs and governance choices underneath it remain visible.
Read →AI confidence is not user trust
Why probabilistic outputs need explainability, hard constraints, and honest product boundaries.
Read →Strategy frameworks are better when they connect
How creator economics, discovery, distribution, and IP adaptation became an activity system instead of separate slides.
Read →Robust decisions beat perfect forecasts
How I separate researchable gaps from structural uncertainty and design around the range that matters.
Read →