trenches — A cross-chain crypto trenches digest: what changed since the last report, which narratives are moving, which tokens belong to them, and what the evidence actually shows.

Methodology

trenches is a publication, not a terminal. Each report is a dated document describing what was observed in the crypto trenches across whichever chains were actually covered in that report's observation window, which narratives were moving, which tokens belonged to them, and what the evidence did and did not show.

What a report is

A report is one validated JSON document. Every public representation (the HTML page, the Markdown, the plain text, the JSON and the feed entry) is generated from that same document by a small static generator. No representation is written by hand, so they cannot disagree.

Report ID
Derived from the publication minute in UTC, e.g. 2026-09-09T1200Z. IDs are stable and several reports can be published on the same day.
Publication time
When the report was published (published_at). It is not the time the data was observed.
Observation window
The period the observations cover (observation_window.start to observation_window.end). Each metric additionally records its own observed_at and, where relevant, a measurement window.
Previous report
The report a new report compares itself against. The first report has none, and therefore claims no changes or trends.
Revisions
Corrections are published as a new report whose ID appends -r2, -r3 and so on, with a revision note. The original stays online unchanged and is marked as superseded. Nothing is silently overwritten.

Coverage

Every report lists the chains and networks that were actually observed, with the sources used for each and the scope of what was looked at. Mainnet and testnet are labelled explicitly. Chains that were not observed are not listed, and coverage gaps are stated with their impact. The trenches are not assumed to be a single chain.

Narratives and lifecycle labels

A narrative is a theme that tokens cluster around. Each narrative has a stable id (used for anchors and for comparison across reports), a description, catalysts, member tokens with roles, and a lifecycle label:

Token roles inside a narrative are leader, member, derivative, copycat-suspected or unknown. A copycat or originality claim is only made with evidence; otherwise it is marked suspected and given low or unknown confidence.

Facts versus interpretation

Each narrative separates observed items (facts tied to evidence) from inferred items (interpretations, each with its own confidence and basis). Contradictory observations are listed as such rather than dropped.

Confidence levels are high, moderate, low and unknown. They apply to narratives, inferences, actor activity and relative-strength statements. Claims about wallet relationships, promotion or coordination that lack evidence must be marked low or unknown.

Metrics

Every metric preserves its value and unit, the time it was observed, the measurement window (or that it is a point-in-time snapshot), and the source. Market cap and fully diluted valuation are distinct metrics. A change is only shown when a baseline (value, time, source and usually the previous report) is recorded alongside the current value; without a comparable earlier observation no change is claimed.

A value that could not be obtained is null in the JSON and rendered as unknown. It is never written as zero.

Token identity and addresses

Tokens are identified by chain id (CAIP-2 style, e.g. eip155:8453 for Base mainnet or solana:5eykt4UsFv8P8NJdTREpY1vzqKqZKvdp for Solana mainnet) plus the contract or mint address, never by ticker alone. EVM addresses are shown as observed without normalising checksum case; Solana addresses are base58 and case-sensitive.

Checks are not verdicts

Reports never call a token safe or clean. Instead each token lists the specific checks that were run (for example mint authority, freeze authority, holder concentration, liquidity lock), what was observed or not observed, when, and from which source. Anything not checked is unknown.

Sources and research workflow

Reports are produced by a research workflow that discovers and invokes data tools through the Frames catalog (token discovery, market data, holder data, social activity, launchpad feeds and web retrieval), then writes the report JSON by hand or with agent assistance. Every source consulted is listed in the report, including sources that failed or returned partial data, and Frames tools are cited by their catalog tool id.

Retrieved web pages, social posts and token metadata are treated as untrusted source material. Their text is quoted only as escaped excerpts and never rendered as HTML. Nothing on this site calls models, tools or paid data providers when a page is loaded: pages are static files.

Formats and routes

JSON and Markdown routes are served with permissive CORS headers and need no authentication or cookies.

How to cite

Cite the permanent report URL, for example https://trenches.party/reports/<report_id>/, optionally with a fragment such as #evidence-<id> or #narrative-<id>. Latest aliases change on every publication and are not citable.

curl -s https://trenches.party/latest.json | head -c 400
curl -s https://trenches.party/reports/<report_id>/index.json

Everything here is observational reporting with stated uncertainty. It is not a recommendation to buy or sell anything.