Every classification Luminos shows — Bundled, Cabaled, or Organic — is an automated, algorithmic opinion: the output of heuristics applied to public blockchain data at a point in time. It describes a pattern our system detected in how a token's supply was distributed and acquired. It is not a legal or factual finding, not a certification of fraud, and not a statement about anyone's intent. See our Terms of Service for the full legal framing.
Bundled the distribution pattern most associated with manufactured supply — coordinated acquisition, common funders, wallets funded through mixers or low-trust exchanges, large non-exchange holder clusters, or cohorts of wallets that were all created and funded together.
Cabaled distribution shaped by insiders or unusual exchange-routing without the harder signs of Bundled — single-exchange dominance, funding skewed toward regional or low-trust exchanges, or concentration among a small number of top holders.
Organic no coordinated-distribution pattern detected — diverse wallet ages, a healthy spread of holdings, and funding traced back to well-known exchanges. Organic is shown in blue, not green, deliberately: it means "no manipulation pattern found," not "safe to buy" or "will go up."
Public data only — nothing here comes from a tip, a report, or a subjective read of a project's community.
We deliberately don't enumerate the specific providers and techniques behind these inputs — the same reasoning as the thresholds below: a precise bill of materials is a blueprint for evading the detector. We don't control the accuracy, availability, or completeness of any third-party source. An outage, a gap, or an error upstream can change what Luminos sees and therefore what it displays.
Luminos runs dozens of independent signals against each token — funding sources, holder-age spread, wallet-creation clustering, gas-fee fingerprints, identical-holdings cohorts, and more — and looks for convergence: several independent signals pointing the same direction. A single signal, on its own, is deliberately never enough to decide a classification. We don't publish the exact thresholds and weights the engine uses — the same way a spam filter doesn't publish its exact rules — because doing so would hand anyone trying to manufacture an "Organic" read a blueprint for beating the detector. The signals themselves are not secret: every scan result shows which ones fired.
Two independent numbers are shown alongside the classification: a score (how strongly the token fits its assigned category, 0–100 — not a price prediction) and a confidence level (how much independent data was available to back the read). A brand-new token with thin trading history gets a lower confidence even when the signals look clear — treat those reads as more provisional.
Data quality varies token to token. Missing holder data, an unindexed brand-new token, or a temporary provider outage can all leave a scan with less to work with — Luminos surfaces this explicitly as lower confidence or a "data pending" state rather than guessing.
A classification reflects the data available at the time of the scan, not a permanent label. Re-scanning a token can produce a different result as: new on-chain activity occurs, an upstream provider corrects or backfills its data, we refine a signal or fix a bug, or an appeal (see below) surfaces something our automated read missed. There is no cost to us in a verdict changing — the goal is an accurate read, not a consistent one.
Use the appeal form to tell us why — which signal, number, or cluster you believe is mistaken, and any evidence that supports your read. Every appeal is logged and reviewed in good faith; review is manual and discretionary; and, as explained on that page, filing one does not by itself change any verdict.