Building the research layer for understanding who’s actually shipping.

Launchpad factory logs (fifteen launchpads), DEX and market listings, ecosystem directories, GitHub code search and verified-contract sync feed one candidate table — 1.35 million screened. Nothing reaches a page without passing the quality gate.

A launch record becomes a project page only on evidence: a site of its own, a repository, a recorded ship, documentation or real channels. A website several projects share is nobody’s. The gate re-runs daily over the whole catalogue and demotes as readily as it publishes.

Repositories, releases, commit activity, changelogs and RSS feeds are read on a schedule and recorded as ship events. Every meaningful ship keeps the source it came from; a repository that turns out to be borrowed becomes context, not evidence.

Shipping, Active, Quiet, Dormant and Resumed are decided by dated, source-backed activity — never by price. Still Building means verified activity through a tracked market drawdown, and is stated as an observation, not a signal.

Identity, narrative, status and last ship first; then the latest ship, the build timeline, market context, research notes and every source with how firmly it belongs to the project. Self-reported and verified never look the same.

The catalogue by what it is building and shipping: verified builders, token projects, new builders, filters for age, volume and evidence; twenty-three narratives; a map of the whole chain clustered by narrative.

The chain’s activity as it happens and as a week: every ship with its source, the rollup, who is under the radar, builders ranked by verified development and on-chain use, fifteen kinds of measured change, and an archived weekly report.

Paste a contract address and Dawnscan answers one question from its own tables and the chain: is anyone actually building this? It gives no score, grade or verdict, and it will not.

Verified contracts are synced, deployers watched for new deployments, upgrades and activity, and every tracked contract carries the date it was deployed, read from the chain.

Price, liquidity, volume, pool depth, supply and a daily index from several providers, shown behind the builder story and used only as context or a filter — never to rank. Token market status and token verification are stated on every token.

A bubble map of one token’s largest balances, with pools, lockers and burns labelled. An address on it is a point on a chart of one token’s supply — never a person, never scored, never followed.

A builder proves ownership by GitHub, a DNS record, a file on the domain, or an on-chain signature, and can post updates. What a builder says and what Dawnscan observed stay visibly separate.

Research bounties, Request Research, bonds, holder votes and Scout tasks run on $DAWN. None of it buys ranking or coverage; that is enforced in code and tested.

Every published fact as JSON with no key, a lookup by contract address for bots, RSS feeds, an embeddable badge, llms.txt, and an MCP server with twelve tools that answer in Dawnscan’s own words.

Follow projects and receive a digest. It works; few readers use it yet, and mail delivery is not switched on for everyone.

Daily invariants over the catalogue, provenance and coverage counters, an hourly platform summary, and a public status page that decays to “stale” when Dawnscan has not summarised itself recently.

Full-platform audits, then repairs: in September the publication gate was corrected and 7,957 evidence-less pages returned to indexed status; restated code-activity summaries were collapsed into one ship per repository per week. Regression tests now guard each correction.

Project → website → repository → commit → release → contract → deployment → token, with the reason each link exists recorded: a site that links the repository, a fingerprint in the code, a deployer that matches. Repository, token and contract attribution are live; making every relationship provable is the work.

Every job is idempotent and retried; what is being tightened is observability — which sources fail, why, and whether a failure is the provider or Dawnscan — so a broken read never becomes “stopped shipping”.

Dynamic titles and descriptions that state what Dawnscan observed, share cards that place a project on the chain, breadcrumbs, a complete sitemap, deep links to individual ships. Distribution comes from useful data, not from mechanics.

From one answer to the whole picture for an address: the project, the builder, the site, the repositories, releases, deployments, other contracts, and whether they are still shipping — each line with its source.

Commits, releases, new repositories, contracts, deployments and significant changes in one chronology per project, instead of a ship timeline beside separate contract events.

Accelerating, stable and slowing, on top of the statuses that exist — only if each can be explained from the underlying events. No magic score.

Liquidity events, creator-side deployments and token activity as context to the builder story. Any wallet relationship will be worded by what Dawnscan can prove — verified, observed, inferred, or unknown — and never turned into a person.

Recently shipping, newly discovered, resumed, new releases and new deployments exist as tabs and feeds; next are the most consistent builders and activity by narrative and ecosystem. No “trending” unless the metric means something.

Every project page should answer five questions in that order. Information hierarchy, evidence visibility, freshness, related projects and mobile are being reworked against that test.

Claims are live; next is a builder that spans projects and repositories, with builder-provided metadata and Dawnscan-observed evidence kept apart on the page.

Follow a project and hear when it releases, deploys, adds a repository, resumes, or goes quiet — the digest, made timely. Nothing that fires on price.

Activity, releases, deployments and builder signals as first-class endpoints, historical datasets and exports, and the MCP server grown with them, so Dawnscan is a data layer for the ecosystem and for agents.

Dawnscan where the conversation already is: name a project or paste an address in a reply, a group or a DM, and the agent answers with the same source-backed line the API gives — status in Dawnscan’s words, ships, last ship, the link back — and nothing that reads as a verdict. Built on the public API and the MCP server, so it can never say more than the site does.

“What changed this week?”, “which AI-agent projects resumed development?”, “compare these two builders.” Only on top of the evidence graph, every answer citing Dawnscan sources, and never in a page request. Nothing is wired today, by design.

Higher API limits, historical datasets, exports, alerts and early research access are the candidate uses; Scout staking and evidence challenges remain designs behind refused flags. Basic evidence stays free so the platform’s value is visible without a token.