Skip to content

colophon · measured from the repository at build time

How this site is built

Most sites tell you what they do. This page shows the receipts: what this site is made of, what it refuses to carry, and the numbers it holds itself to on every merge. Everything below the principles is measured from the repository itself at build time — when a number here goes stale, the caption says exactly when it was last measured.

The numbers

runtime dependencies
24
test cases
161 across 18 files
content files (all site copy)
68
architecture decision records
15
UI kits / CSS frameworks beyond Tailwind
0
tracking scripts
0

First-load JavaScript, per route

/
200 / 200 kB gz
/learn
203 / 350 kB gz
/work
194 / 200 kB gz
/work/payments-platform
197 / 200 kB gz
/work/mvc-to-api-migration
197 / 200 kB gz
/work/auth-architecture
197 / 200 kB gz
/work/async-image-pipeline
197 / 200 kB gz
/work/cv-builder
197 / 200 kB gz
/deep-dives
195 / 200 kB gz
/deep-dives/why-background-jobs
194 / 200 kB gz
/deep-dives/the-worker-host
194 / 200 kB gz
/deep-dives/the-job-contract
194 / 200 kB gz
/deep-dives/the-outbox
194 / 200 kB gz
/deep-dives/when-jobs-fail
194 / 200 kB gz
/deep-dives/security-and-observability
194 / 200 kB gz
/deep-dives/production-patterns
194 / 200 kB gz
/deep-dives/claude-md-cleanup
194 / 200 kB gz
/deep-dives/nested-claude-md
194 / 200 kB gz
/deep-dives/the-hash-workflow
194 / 200 kB gz
/deep-dives/document-driven-development
194 / 200 kB gz
/d3
194 / 200 kB gz
/deep-dives/quantum-machine-learning
194 / 200 kB gz
/deep-dives/bangla-pos-tagging
194 / 200 kB gz
/deep-dives/network-anomaly-detection
194 / 200 kB gz
/deep-dives/blood-cell-detection
194 / 200 kB gz
/deep-dives/multi-output-cnn
194 / 200 kB gz
/deep-dives/startup-success-prediction
194 / 200 kB gz
/research
196 / 200 kB gz
/agents
197 / 200 kB gz
/hire
194 / 200 kB gz
/playground
195 / 200 kB gz
/cv
195 / 200 kB gz
/colophon
195 / 200 kB gz
/research/quantum-machine-learning-thesis
195 / 200 kB gz

measured at commit be92016 on 2026-07-13 · enforced in CI on every merge · gzipped eager chunks only

Principles

  • Content is markdown, period. Every sentence on this site lives in a content/*.md file behind a strict schema — components never carry prose.
  • The physics is real. The hero classifier, the playground, and every Bloch sphere read from one hand-written, unit-tested statevector simulator. No animation pretends to be computation.
  • Agent-native, honestly labeled. MCP server, WebMCP browser tools, llms.txt, fit reports with a mandatory gaps section, an AI-pitch page that discloses itself. The machine interface is documented like an API, because it is one.
  • Performance is a contract, not a vibe. Per-route JavaScript budgets are enforced in CI; the homepage carries a live quantum simulator inside the same budget most sites spend on a cookie banner.
  • Never overclaim. Results are reported with their baselines, including the ones the quantum models lost to.

Get this as a template

Everything above is being extracted into an open-source template — the agent-native portfolio: the markdown pipeline, the MCP/WebMCP tool layer, the fit report, the guestbook, the budgets CI, minus my content. If you want it when it ships, leave an email. No spam — exactly one launch email, then the address is deleted.

stored only until the launch email is sent, then deleted

Machine-readable: /colophon.json · the full machine interface lives on /agents.