Method

Our method

Identified sources. Documented facts. An analysis you can examine.

Origin, claim, interpretation

ARKOSCOPE separates where information comes from, what the source claims and the interpretation drawn from it. An official announcement establishes that an organization made that announcement; it does not automatically confirm everything it claims.

An analysis should be open to examination: which sources support it, which elements remain uncertain and what could call it into question?

How an account becomes a source

ARKOSCOPE separates a project’s official sources from approved expert sources. For an official source, its link to the project must be demonstrated. For an approved expert source, the review covers its identity, expertise and limits; it is not presented as the project’s spokesperson. A name, a logo, a platform badge or a follower count is not enough.

Each registered source carries a permanent identifier, an HTTPS address, verified proof (of its link to the project for an official account; of its identity and expertise for an approved source), a method and verification and review dates. The details of the proof stay internal; the link to the original post stays visible in the app.

An attribution that becomes doubtful is removed from the feed until it is reviewed again. A revocation removes the official badge without erasing the history.

What “official account” and “approved source” mean

These badges describe provenance. They guarantee neither the future accuracy of a claim, nor the future identity of an account, nor the absence of fraud or loss.

Who publishes the feed

SCOPE FEED editorial posts are selected and published by ARKOSCOPE from registered sources. Users and outside accounts cannot post to it freely.

An official account may itself publish promotional content: its provenance is accurate, but its content remains a statement by the project.

Corroboration and contradiction

Our reading rule: several reposts of the same press release are not several independent confirmations.

In the feed, the “Confirmed information” badge is reserved for cases where several reliable elements agree; an incomplete element carries the “To verify” badge and carries little weight in the analysis.

Contradicting evidence from primary sources counts against a thesis; when it becomes significant, the thesis moves to the “Invalidated” state.

The role of automation and AI

Collection and the organization of the service are automated; registering a source requires verified proof: of its link to the project for an official account, of its identity and expertise for an approved source.

A language model summarizes, translates and classifies the selected posts; its output is checked against a fixed format before it is stored. It does not decide what is sensitive: regulatory or legal topics, financial status, capital transactions and absolute promises are detected by fixed rules before it runs and flagged for human review, and the model cannot lift that flag.

Not every card is reviewed by a human, and no human approval is simulated. An automated analysis is not infallible: its traceability makes it possible to examine and correct it.

How a Thesis is built

Identifying the source of an announcement and independently confirming its content are two separate checks. A dossier built only on the project’s own announcements must be read within that scope. Adding new elements to the dossier does not, on its own, demonstrate any improvement in the project’s fundamentals.

A Thesis starts from the verified facts gathered on a project. Its statement, expected confirmation and invalidation point are an ARKOSCOPE analysis, labelled as such: it is derived from those facts and adds no new fact.

A thesis is proposed only once a minimum number of facts has been observed over a sufficient period. These thresholds decide whether a dossier deserves to be opened; they are not reliability scores.

The statement is fixed when the thesis is created; its state is then re-evaluated from the facts (states computed by the engine; how they are displayed may vary between app versions): “Proposed” when the thesis has just been opened, “Strengthened” when new verified facts have been added since its creation, “Weakened” after a period without any verified fact longer than its threshold, “Invalidated” in case of significant contradictions. These states describe the body of facts, not a project’s value.

The evidence base counts verified facts, the share of primary sources and the observation period. It predicts nothing and measures no probability.

How Radar is selected

Radar 20 is a twenty-place editorial selection. It explains why a dossier is followed and studied, and why the selection changes.

No team can pay its way in. Radar is neither a performance ranking nor a recommendation: it does not take your financial situation into account and guarantees no project’s success.

The twenty projects you can follow in SCOPE GUARD with VISION PRO are a separate count: they never change Radar.

Updates, corrections and reporting

The Timeline keeps the dated changes of an analysis and its corrections. “Followed over time” does not mean “up to the second”: the app shows the timestamp of what it displays.

To report an error, write to support@arkoscope.com with the content concerned (link or title), the date and the point you dispute. We never ask for a password, Apple code, private key or financial data. No response time is guaranteed.

Interests and advertising

ARKOSCOPE makes verifiable commitments: no paid place in Radar, original links kept, analysis kept separate from announcements. The first version contains no advertising; in V2 trials, demonstration advertising stays separate from editorial content and never becomes a source.

These commitments claim neither absolute independence nor the absence of any possible conflict of interest.

What the method does not guarantee

ARKOSCOPE provides general information. It guarantees neither the permanent accuracy of data, nor the continuous availability of the service, nor any return. A project listed in the Observed Universe, in Radar or in a dossier is not a buy recommendation.

To see this method applied to a real case, read the Aptos dossier.