Reading Millions of Conversations
How questions, clustering, and privacy gates turn AI conversations into evidence—and how to tell what the resulting map can actually show.
Oct 23, 2026
Writing
Essays on machine learning, evaluation, data infrastructure, and distributed systems.
Essays
Start with The Evaluation Machine, or browse all selected essays below, newest first.
Better Taste for a Tiny Camera
A $30 keychain hit, taken apart one question at a time: what it is, how to get inside, how it makes a photo, what could change, how to compile better taste into cameras people already own, and what happens when looks are cheap.
Oct 23, 2026
AllReduce, All the Way Down
Follow AllReduce from a simple operation through collective algorithms, network topology, and GPU execution.
Sep 8, 2026
The Evaluation Machine
How contracts, evidence, environments, and release gates help teams decide whether learned behavior is ready to ship.
Sep 2, 2026
Rocketships for the Mind
A system I built to recover project state from 6,604 AI conversations, reconcile evidence, and construct task-specific memory.
Jul 31, 2026
From Payments to Claims: Where Finance Infrastructure Goes Next
Financial instruments are architectures for economic workloads. As business flows become machine-legible, observability may change not just underwriting, but the unit of financing itself.
Apr 8, 2026
How to Design Metrics That Change Decisions
A practical framework for choosing metrics, setting action thresholds, and retiring the ones that no longer help
Dec 12, 2025
Golden Datasets: Launch Criteria for LLM Systems
How to decide what must pass before a new model, instruction, or workflow ships
Aug 25, 2025
LLMs for Backend Engineers: From First Principles
A systems view of LLM latency, cost, reliability, and evaluation
Jun 17, 2025
No matching items
Interactive research
Interactive essays and experiments from my independent research site, Novus Opus.
Bend, Not Shift
Explore scaling laws and what would make a learner better at learning.
The Second Loop
Explore the control loops that could help AI maintain long-running plans and commitments.
The Limits of Connectivity
Experiment with the trade-off between communication and independent discovery.
The Geometry of Disagreement
Build an opinion-dynamics model, then experiment with trust, network structure, and polarization.
New essays by email
Writing on systems, machine learning, and data infrastructure.
Earlier writing collects older essays, projects, and experiments. The RSS feed includes both selected and earlier writing.