εEPSILON

Public impact record / measured conservatively

Reach is visible.
Impact must be earned.

EPSILON separates anonymous browser signals from server-verified research use and independently reviewable contribution. A page view is unverified reach. A historical evidence run is counted only after the server completes it. A substantive external challenge becomes impact only when it remains publicly inspectable.

Anonymous lab signals—

Deduplicated, unverified browser sessions that opened the instrument.

Verified historical experiments—

Unique real-data evidence configurations completed and recorded by the server.

Challenge-link signals—

Unverified sessions that opened the external review channel.

Evidence hierarchy

Not every number means the same thing.

  1. 01 / Reach

    Someone encountered the work.

    Visits, impressions, and video views are useful distribution diagnostics. They are not counted as adoption or educational impact.

  2. 02 / Use

    The server completed a research action.

    Only completed historical-data evidence runs are shown as verified product use. Browser openings and link clicks remain explicitly unverified signals.

  3. 03 / External challenge

    Someone tested the method itself.

    A review counts only when a real person identifies a specific assumption, evidence gap, interpretation risk, or product weakness in a publicly linkable record.

  4. 04 / Change

    The criticism altered the work.

    Accepted challenges link to the resulting experiment, issue, correction, or product change. Unfavorable findings remain visible.

Falsification challenge / open

Find the assumption
we failed to expose.

Start with a question about a backtest conclusion. Describe the claim, the assumption you distrust, and an optional public example. A code contribution is not required. Completed experiments and feedback-led changes need their own records.

Bring a question ↗File an independent reproduction →

Participation requires no stars or endorsements. No investment claims. No account identifiers in the measurement ledger; coarse source categories only; rolling 90-day measurement window. Unavailable measurements display a dash, not zero.