Rules

These are rules, not preferences. When someone asks for a score, the answer is that we published a rule.

These are rules, not preferences. They exist because the pressure to break them will be constant and reasonable-sounding. When someone asks for a score, the answer is "we published a rule," not "we'd rather not."

1. No verdict, ever. No trust score. No category scores. No stars, grades, tiers, or shortlists. No "picks that might be interesting." No hot list. Five category scores are five verdicts with extra steps: people average them in their heads, and then we have built a rating service that denies being one.
2. Colour marks direction, never worth. A number that went up is green and one that went down is amber, because that is what the number did. No coin is ever coloured good or bad.
3. No project money. No paid listings, no sponsored entries, no referral revenue on exchange links, ever.
4. No holdings. The operators hold nothing. This removes the front-running conflict and the selection-bias problem in one move.
5. News and price never get a story. On a big move the site lists the headlines that name the coin from the last three days, and says plainly when there are none. It never says one caused the other.
6. Nothing goes live without a human. Machines detect, humans verify.
7. Every claim carries a source and a snapshot. If it can't be shown, it doesn't go on the page.
8. Retire the word "unbiased." Let the structure make the claim. Saying it invites the argument.

Corollaries that follow from the rules

• Preset sorts use factual names only ("most delivered in category", "longest silent", "most rescheduled"). No weighting anywhere.
• Competitor blocks are ordered by size, a fact, never by our opinion.
• Contract facts are presented as facts, never as a reliability verdict.
• The disclosure bot replies with the delivery record only. No characterization. Never the word "shilled."
• Addressable-market figures are factual market sizing, cited to source with date. Never "how much it will solve"; that is forecasting.
• Entries are immutable and append-only, with a reason on every change.

Design rules (they serve the rules above)

• Everything is a sentence. "No code changes in 11 months," not "334d."
• No cards. Cards imply products to choose between. A long sectioned page implies a document.
• Same layout on every project page. Learn one page, know them all.
• Light mode by default. Dark reads as trading terminal; light reads as reference.
• Generous type, boring font. Small text signals fine print and costs trust.

What counts as delivered, missed, or withdrawn

A promise is shipped when all three hold:

1. The thing described is publicly usable or publicly published (mainnet code live, a document published, a feature reachable by users). A testnet release ships a promise only if the promise was about a testnet.
2. There is evidence a stranger can open: a release page, a block explorer entry, a repository tag, an announcement on the project's own channel. That evidence goes in delivery.evidence.
3. What shipped matches what was promised. If the scope shrank, record the delivery with a note saying what was left out, and open a new promise for the remainder if the project still commits to it.

A promise is missed when its due window has fully passed and no delivery is recorded. Windows end on the last day of the stated day, month, or quarter, using the latest revision. A vague promise has no window and is never missed; it is shown as open with its own words for timing.

A promise is withdrawn only when the project itself said so, with a source. Silence is not withdrawal. Silence is an open promise.

What counts as a promise

A public, dated, checkable statement by the project about something it will build or do. "We are exploring" is not a promise. "Coming Q3" is. A roadmap item under a dated heading is a promise with that heading's date. A roadmap item under no date is a vague promise if it commits ("will ship"), and not a promise at all if it merely describes a direction.

Write plain so that someone with no crypto background can say it back to you. Never characterise. "The team will release a new consensus client" is fine. "The team finally ships" is not.