How this site works
What this site changes from the base
--surfacevar(--ground)--surface-mutedvar(--exhibit-ground)--surface-accentvar(--pen-soft)--bordervar(--hairline)--border-strongvar(--ink)--textvar(--ink)--text-mutedvar(--ink-2)--text-faintvar(--ink-3)--ringvar(--ink)--radius-control0.5rem--radius-pill1rem--composer-height3rem--chip-radiusvar(--radius-pill)--chip-pad-blockvar(--space-2xs)--chip-pad-inlinevar(--space-sm)--chip-height2rem--arrive-distancevar(--space-md)--arrive-durationvar(--dur-short)--thread-measure44rem--chapters-width12rem--label-column12.5rem--label-gap3rem--block-headingvar(--text-h3)--block-heading-lhvar(--lh-h3)--heading-tracking-0.02em--card-image-ratio4 / 3--exhibit-measure36rem--exhibit-height22rem--rail-indent1.5rem--rail-marker0.25rem--grip2.5rem--scrimcolor-mix(in srgb, var(--ink) 64%, transparent)--on-scrimvar(--ground)--underline-offset0.25em--underline-offset-hover0.375em
The questions
- Q01 — Identity and level
- Q02 — Fit and logistics
- Q03 — Work history
- Q04 — Show me the work
- Q05 — Complexity made legible
- Q06 — Outcomes, honestly framed
- Q07 — Ownership and zero-to-one
- Q08 — Decisions and trade-offs
- Q09 — Research that changed a decision
- Q10 — Systems: built and governed
- Q11 — AI: how he uses it
- Q12 — AI: what he has designed for it
- Q13 — Close to code
- Q14 — Working with PM, engineering and leadership
- Q15 — Leading and raising others
- Q16 — How this site works
- Q17 — Contact and links
The blocks
statement
A block type
Statement
The plainest block: an optional eyebrow label, an optional heading, and a passage of prose underneath. No rows, no options, no image — just the words, carried straight from wherever they were written.
facts
Facts
- Shape
- A two-column list — the label hugs, the value fills the rest of the row.
- A row can link out
- to a source, a page, or a document
- Used for
- a quick rail of role, team, timeline, and similar short answers
exhibit

ledger
What I claim
A ledger block can also point at an existing case study's claims by name, rather than writing its own.
Documented rows have a dated artifact I can produce on request.
decision
Why there is no confidence percentage
- The constraint
- A rules engine that matches a question to an answer by chip, phrase, or keyword has no real confidence score to show. An early sketch put a percentage next to the answer anyway, borrowed from the shape of a familiar AI product.
- The options
- Show an invented confidence percentage next to the answer
- Show a similarity score without saying which tier produced it
- Show the mechanism instead — which rule fired, or that none did and a similarity search took over
- The call
The percentage was rejected in favour of a plain line: the matched question, why it matched — a chip click, a phrase, a keyword, or, when no rule fires, a similarity score labelled as exactly that — and the answer it opened. A rule hit still opens its answer at once. When nothing matches, the closest few questions by meaning are offered as chips under a line saying there is no exact answer, never opened automatically, because the top pick by similarity alone was wrong often enough that automatic was the wrong default.
- The cost
- The line that replaced the percentage reads more like an engineering log than a product feature, and it is one clause longer. Every word in it is checkable.
the system's own notes on how a question gets matched to an answer
timeline
Timeline
- FirstA dated or ordered entrywith a short note underneath, if one is worth keeping
- ThenAs many entries as the story needs
- LastIn the order they happened, not the order they were written
diagram
chart
The numbers behind this chart
| Question type | Postings of 43 |
|---|---|
| Working with PM, engineering and leadership | 37 |
| Ownership and zero-to-one | 30 |
| Systems, built and governed | 29 |
| Research that changed a decision | 28 |
| Complexity made legible | 25 |
exhibit — compare

