2026.08 — Cycle 07
The instrument, not the answer
Weighing honors thesis directions. Not the topic — the shape. Whether the
question is one I already know how to build my way through, or one that
resists being solved. The distinction matters more than the field.
2026.07 — Cycle 06
Research as connective tissue
Coursework and apps stopped being separate concerns. Same falsifiability
instincts, same statistical rigor. The academy asks whether a claim survives
its own tests. The products ask whether a user survives the same. The overlap
turns out to be larger than the syllabus admits.
2026.06 — Cycle 05
Verse dispatches
Notes from inside Epic's Verse language. Strong typing, effect specifiers,
transactional computation — a functional-adjacent runtime nested inside
someone else's engine, someone else's economy, someone else's platform. A
useful humility exercise for a web-first builder.
2026.05 — Cycle 04
The other side of the fence
Time as an AI trainer, generalist: evaluations, red-team probes, quality
signals across domains. Instructive to see model development from the input
side rather than the API side. Different physics on the way in.
2026.03 — Cycle 03
Eight signals, one frequency
Portfolio site is live. Eight applications in parallel development — each one a
different answer to the same question. Notes on building multiple complex systems
simultaneously without losing coherence.
2026.03 — Cycle 02
On architecture as archaeology
The best code architectures aren't designed — they're excavated. You remove what
doesn't belong until the structure reveals itself. Thoughts on letting the problem
space dictate the solution shape.
2026.02 — Cycle 01
First transmission
Initial signal. Setting up the workbench, calibrating the instruments. Notes on
choosing a stack for a portfolio that needs to feel both ancient and impossible.
Why static files can carry more weight than a framework.