How work arrives

No case studies. Here is the actual shape of a delivery.

This page does not carry client logos, invented metrics or borrowed case studies. Client work is confidential and fabricating proof would defeat the point. What it does carry is the exact anatomy of what lands on your side, including the completion receipt that closes a Herman-led, Brick-led or paired engagement alike.

What actually lands

Three artefacts, every engagement

These are the documents you can hold up in a review. They are the same shape whether Herman or Brick led the work, and the same shape whether it was a data pipeline or a brand system.

Before you commit to anything

The scope note

Contains

Deliverables, listed individually
Explicit non-goals
Acceptance criteria written as checks, not adjectives
Risks and unknowns visible on the first page

It is the cheapest possible way to find out whether we disagree about what "done" means, and it costs you nothing.

Throughout, and again at handover

The evidence log

Contains

Each command run, with its actual output
Test runs, including the ones that failed first
Measurements with their collection method attached
A dated record of decisions and what changed them

Claims are checkable or they are marketing. This is what makes the difference legible.

At acceptance, whoever led

The completion receipt

Contains

The accountable lead, and the challenger if there was one
Acceptance criteria checked off with the evidence against each
Known gaps, stated as gaps rather than smoothed over
The named human who accepted it, and the date

It is the same receipt for a Herman-led, a Brick-led and a paired engagement. One standard means you can compare two of our deliveries to each other, which is the only comparison that would expose us.

At delivery

The handover pack

Contains

The artefact itself, deployed or deployable
Runbook covering routine operation and known failure modes
Acceptance criteria checked off one by one
A written list of known gaps and what it would take to close them

Work that only its author can operate has not been delivered, it has been parked.

One proof standard

The shared completion receipt, line by line

Herman-led, Brick-led or paired, you get the same receipt. There is no lighter version for the mode that happened to be cheaper, and no line that is quietly dropped when it would be inconvenient.

Completion receipt

Specimen · structure only, no engagement data

Accountable lead: ⟨Herman or Brick — named in the scope note before work starts⟩ The operator who owned the work, named before it started.
Challenger: ⟨the other operator, or “not used”, with the reason stated⟩ The operator who tried to break the result, and what they found. Says "not used, and why" when there was no challenger.
Acceptance criteria: ⟨each criterion from the scope note, checked off individually⟩ Every criterion from the scope note, checked off one by one, with the command or measurement recorded against each.
Evidence: ⟨commands and their real output, tests including failures, measurements with method⟩ The commands run and their real output, the tests including the ones that failed first, and the measurements with their collection method.
Known gaps: ⟨anything unmet, stated as unmet, with what it would take to close it⟩ Anything unmet, stated as unmet, with what it would take to close it. A receipt with no gaps section has not been checked.
Human approval: ⟨Jon or Anthony, who accepted it, and when⟩ The named human who accepted it on our side, and the date. No delivery is signed off by an operator alone.

This is the blank form, not a record of work. Every line below is answered in writing at handover, and the same six lines close a Herman-led, a Brick-led and a paired engagement alike.

If a receipt reaches you with the known-gaps section empty, that is a claim that nothing is outstanding — challenge it. It is the line we get asked about most, and the one worth reading first.

Operating model

The rules the work is built to

Evidence over assertion

Anything presented as fact carries the measurement or command that produced it. Anything presented as judgement is labelled as judgement.

Reproducible or it did not happen

Builds run from a clean checkout. Analyses re-run from raw inputs. Assets regenerate from source. If it only worked once, it is not done.

Failure reported at full volume

Broken builds, wrong turns and dead ends are reported as they occur, not discovered by you at handover.

One accountable lead, never a committee

Every engagement names one operator who owns the outcome. Where a second operator is involved, they are there to challenge the result, not to dilute who answers for it.

Written down beats remembered

Scope, criteria, decisions and gaps live in documents, not in conversations neither of us will recall accurately in six weeks.

Publicly checkable

What you can inspect right now

Client work is confidential, so it is not on this page. Our own output is not, and you are welcome to audit it.

Evidence ledger

The public artefacts and proof standard this operation can substantiate now, without implying that a separate logbook route already exists.

Open the evidence ledger

This site

Built as a demonstration piece: a custom Hugo theme, two original vector marks drawn from geometry, two complete themes, self-hosted assets, zero third-party requests, and an automated test suite that gates the deploy.

Read this site’s source