Two doors, and both of them start with a question.
SYKQD is a questions business before it is anything else. For an organisation that means one of two beginnings: find the cause you should actually be standing behind, or read what is really going on inside the room.
Find a cause.
Answer as plainly as you would in a meeting you weren't presenting at.
01 of 11
Who are we talking to?
Also here
Questionnaire Lab.
Build a new questionnaire for your audience, or tighten one you already use — versioned questions, a stated measurement intent, bias checks and an export you can put on your own page. It has a room of its own.
Behind it
What this is actually built on.
Six frameworks we work to, each with the belief it is testing and the thing we watch to find out whether the belief survives. None of them is a finding yet.
Framework 01
The specificity ladder
We start at the smallest defensible unit — one object, one route, one Tuesday habit — and only climb toward a general statement when the same shape shows up in unrelated places. A finding that begins general has nowhere to fall back to when it is wrong.
Belief under test: a specific, recognisable detail earns more attention and more accurate self-reporting than a well-written general claim about the same subject.
Framework 02
Ordinary life as the sample frame
Work, commutes, objects, language, rituals and small frictions are the sampling grid. We do not ask people to describe themselves; we ask what they did, in the order they did it, and let the description assemble itself out of enough ordinary points.
Belief under test: behaviour questions about yesterday are more reliable than identity questions about yourself, particularly where an answer carries social cost.
Framework 03
The relationship graph
Place → person → material → season → behaviour → decision. Content is stored as nodes with declared edges rather than as a pile of articles, so a cloth can be reached from a city, a month, a maker, or a purchase someone is about to make.
Belief under test: the edge carries more predictive value than the node — knowing what a thing is connected to explains a decision better than knowing what the thing is.
Framework 04
Questionnaire design as an instrument
Every question carries a stated measurement intent, a version number and a bias note before it is ever asked. A changed question is a changed dataset; the old version stays. This is the discipline the Questionnaire Lab exists to make visible.
Belief under test: a short, concrete, option-bounded questionnaire written in the respondent's own register produces more usable variance than a longer scale-based one.
Framework 05
Declared provenance vs perceived authorship
How much of a piece was machine-assisted is a fact we record. Whether it felt machine-made is a separate opinion we ask for. They are stored apart on purpose, because the interesting research sits in the gap between them.
Belief under test: readers cannot reliably detect AI involvement, but their guess tracks something real — register, specificity and rhythm — rather than the actual production method.
Framework 06
Slow as error correction
Publication speed is not the constraint we optimise. Review state, rights status, source lineage and owner are recorded before anything becomes public, and a draft can sit in the repository indefinitely without becoming a promise to anyone.
Belief under test: a small corpus with clean provenance is more useful to a partner than a large one without it, and the difference shows up the first time someone asks where a claim came from.
And the machinery underneath it.
The enterprise tools are not a layer on top of the public properties. They are the same objects, read differently.
The repository
- Typed content objects with property, section, status, owner, editor and review state.
- Source lineage per object: where it came from, who touched it, which version is live.
- Draft and planned items are stored in the same table as published ones and are readable only by the owner.
The model seam
- Model assistance runs server-side and is checked against a schema before anything is shown.
- Every generated artefact has a deterministic template fallback, and the fallback is labelled as one.
- Prompts and methods are recorded with the object, not lost in a chat window.
The boundary
- Public reads stay public; anything about a person is read and written server-side only.
- Reader perception, tags and reactions attach to a profile only when the reader chooses to connect them.
- External page fetching is opt-in per request and bounded, never a background crawl.
Put a piece of SYKQD on your own site.
The writing, the strips, the Atlas and the questionnaires can run as a panel or a feed inside your own pages. Pick one and take the code, or ask us to send it.
The belts and the urban field, on your page.
We ask for a name because we'd like to know where our work is showing up. Nothing is published about you.
Or simply say something.
The same enquiry that has always been here. A person reads them.