The engine under all four rooms
The Triple Lens
Three questions, asked of every object on this site. SOURCE — where did this come from? POWER — who holds the power here? PATH — what happens next? Nothing on this page calls a language model. The engine reads records that exist, shows you the query it ran, and refuses when there is nothing on record.
One · The three lenses
Three questions that survive contact with a real archive.
Each lens holds a fixed set of fields. That is what makes an answer checkable: you can point at the field that produced it, and at the field that was missing when the engine refused.
| Lens | The question | What it holds |
|---|---|---|
| Ssource | Where did this come from? | Who made it, whether a person or a machine made it, and how complete the record is. Tiers S1 primary document → S6 unverified. |
| Ppower | Who holds the power here? | Creator's share, who controls access, audit right, term length, cost to leave, and what the other side may do with the work. |
| PATHpath | What happens next? | The one next action, how long it takes, whether it can be undone, and what it forecloses. |
The refusal principle
When a required field is absent, the answer is NO RECORD — refused, printed as a result rather than swallowed as an error. A refusal is a correct answer. It tells you exactly which field is missing, which is a smaller, truer thing than a confident paragraph assembled out of nothing. Run cost-of-waiting below on any object to see it: this site has never recorded a deadline, so that lens refuses every time, and it stays in the menu so the gap stays visible.
Why there is no model in the path
A language model would answer all nine questions instantly and plausibly for objects where the record is empty. That is the failure mode this project exists to argue against — the same failure mode as a court declaring a right unclear because no earlier court happened to write it down. The engine can only say what is written. Where nothing is written it says so, and the gap becomes work rather than prose.
Two · Run a lens
The nine specialty lenses, against this site's real objects.
Sixteen objects are seeded here — titles, records, court cases, chapters, corrections, evidence items and one room. Several have gaps on purpose. Pick a lens and an object, or type a question and watch the keyword router show its work.
Or pick a lens
Three · The behaviour that matters most
Absence is not zero.
If you only take one rule from this room, take this one. An absent revenue_share_to_creator means nobody read the contract. A recorded 0 means somebody read it and the creator's share is nothing. Collapsing the two would publish a claim about a person's income out of a hole in a spreadsheet.
Rule 3 · the pledge
human_pledged = false renders human_made: "undeclared", never "false". A creator who did not take the pledge is not a machine. Chapter 6 in the object list is seeded that way — run who-made-this on it.
Rule 4 · asymmetry math
Balance is the mean of the known components only. Severity is a max, never a sum, so one unread field can never dilute a bad term into a decent score. The arithmetic is printed under every would-i-sign-this answer.
Rule 5 · [] is a statement
An empty derivative_claims: [] means the terms were read and none are claimed. An absent one means nobody looked. A filter for "claims no AI-training rights" must never quietly include the unchecked ones, so the two render differently and sort differently.
Four · The binding for this site
Which object types carry which lenses — and the promise behind each one.
A binding is a commitment, not a configuration. Every "never" below is enforced in the engine, so it cannot be quietly relaxed later by someone in a hurry.
| Object type | Carries | Why, as a human commitment |
|---|
A room never carries POWER
The DISCUSS room is in the object list, and its POWER segment reads never. A room is people. Scoring the power balance inside one would be surveillance wearing an accountability costume — and it would be the most tempting feature in the whole system, which is exactly why it is refused in the binding rather than left to good manners. The room still carries SOURCE and PATH, so you can ask who runs it and what has to happen before it opens.
serves_minors — the flag and this site's honest answer
An app marked serves_minors sweeps its query logs at 30 days rather than 180, and no engagement-derived signal may enter an answer. On this site serves_minors = false: the reader, the atlas and the register are built for adults working with court records, and the sibling portfolio apps that do serve minors carry the flag instead. The 30-day rule is stated here anyway, because a rule you only mention where it applies is a rule nobody can hold you to. The no-engagement rule applies on this site regardless of the flag — nothing here reads dwell time, scroll depth or streaks, and room-health says so in its own output.
Five · Every object, every badge
The badges are the data, not decoration.
The same compact S / P / PATH badge appears on the app tiles on the front door, on the atlas rows and on the register rows. Dashed segments read not stated; faded segments are lenses the object type does not carry. Click any badge for the full block.
| ID | Object | Type | Tier | Triple Lens |
|---|
Six · Specialty triads used on this site
The three-part frames a documentary about doctrine actually needs.
The portfolio keeps a shared triad library. These are the ones that bear directly on qualified-immunity work, with the constraint that comes attached to each.
LENS-003 · Unsupported / Contested / Verified
Status follows evidence, not preference. This triad is the atlas's spine: a holding is verified, a reading of it is contested, and an inference nobody has sourced is unsupported no matter how sensible it sounds.
LENS-009 · Facts / Inferences / Questions
Never merge an inference into a fact. This is the same discipline as the four record labels used across the site, applied inside a single paragraph rather than across a table.
LENS-002 · Problem / Common Ground / Solution
Common ground cannot erase documented harm. Useful in the ACT door, where officers and plaintiffs sometimes want the same procedural fix — and where a warm middle column must not quietly retire the harm.
LENS-006 · Past / Present / Future
Forecasts must be labelled as forecasts. Forty years of doctrine invites prediction; every predicted outcome on this site is an editorial conclusion with a date on it.
LENS-005 · Harm / Context / Protection
Never identify a child in an analytical export. Relevant wherever a case record involves a minor, which in this subject matter is more often than the citation list suggests.
LENS-012 · Creator / Audience / Market
Rights and consent override monetization. The reason the rights gate on the front door is a checklist of unchecked boxes rather than a launch date.
Also in the shared library and not yet bound to an object type here: LENS-001 Worst / Middle / Best · LENS-004 Risk / Neutral / Opportunity (show assumptions and decision thresholds) · LENS-008 Urgent / Important / Watch (urgency requires a named trigger) · LENS-010 Red / Yellow / Green · LENS-011 Child / Parent / Professional (age-appropriate, private and non-diagnostic).