THE INTERNET FORGETSTHE RECORD DOESN'T100 CREATOR FILESREAD-ONLY BETA
CREATORSON RECORDFind a creator
THE CHANGE LOG / NOTHING QUIETLY DISAPPEARS

We keep receipts on ourselves too.

Meaningful changes to methodology, policies, evidence and product behaviour are recorded here from the founding alpha.

15material changes on record

Product decisions, withdrawn models and methodology changes stay visible after the new version ships.

CURRENT RELEASE

Creators On Record read-only beta

Renamed the public service from Channelmark to Creators On Record and moved its canonical home to creatorsonrecord.com. Public proposals, voting, reviews, nominations and web forms are disabled in both the interface and the server routes while identity, moderation and coordinated-campaign controls are tested. The 100 researched profiles remain public; the 1,000-name candidate queue is no longer shown in the public directory. The News and politics ratings and category chart are held for separate editorial review. Corrections, privacy requests and general contact now use dedicated email addresses.

The public decides the record

Anyone can now add an event to a creator's history and vote on everyone else's, with no account. Each proposal is checked twice over: whether it happened, and whether it helped or hurt. Sourced events need 8 confirmations and unsourced ones need 15, and in both cases the agreement has to clear a statistical floor, so a close split never passes however many people vote. Movement scales with the size of the crowd and starts slow on purpose: one voter shifts a line by about a quarter of a point, eight by roughly one and a half, sixty produce half the maximum effect, and no single event can move a creator more than 14 points. Editors are deliberately the weakest lever, able to halve or amplify an event by half and to hold an adjustment of at most eight points, with no power to set a score or reverse the public's direction. Open voting can be organised against, so every vote count is published beside the number it produced.

Modelled perception scores withdrawn

Removed the generated perception score from every profile without researched audience reaction, leaving two. The model inferred movement from the wording of each event and could not tell subject matter from conduct, so it penalised a mortician for writing about death, a scambaiter for investigating scams and a broadcaster for presenting an investigation series. Those profiles now show their dated, sourced events with no score attached. Trust and content sub-scores are also gone: they were never researched separately and only ever repeated the overall figure.

Events integrated into the curve

Removed the detached row of event ticks. Documented events without their own scored audience outcome now appear as hollow circles at the line's existing level. Added a confidence ribbon whose width distinguishes low-confidence modelling from stronger reaction research, and filled KSI's missing Sidemen formation and album milestones.

Post-event reaction slopes

Replaced hard vertical score changes with short coloured reaction slopes that begin on the event date. Quiet periods remain flat and grey. Same-date clusters divide the reaction window, allowing a backlash to fall in red before an apology or correction recovers in green.

Wider career highs and lows

Recalibrated provisional event movements so routine announcements remain small while major recognition, sustained successes, serious disputes and public failures create visibly different career highs and lows. The common 50 baseline and hand-researched KSI and Logan Paul points are unchanged.

Event-timed perception movement

Changed every creator curve to hold flat between documented events, then rise or fall on the event date. A controversy discovered years later now produces a sudden drop at the investigation or announcement rather than a misleading decline across the quiet years before it.

Comparable 100-creator histories

Expanded the curve to all 100 researched creators with 849 plotted points. Every record starts at the same 50 baseline on its real YouTube Joined date. Overall position combines public trust and content quality, green segments improve and red segments decline, and each active year is capped at three major points. Provisional public-record movements are labelled separately from reaction-researched points.

Public perception curves

Moved creator history to the centre of every profile and replaced the subscriber line with a scrollable audience-reaction curve. The latest three years open first. Only separately researched outcomes move the line; unmeasured events remain context markers. KSI and Logan Paul are the first completed reconstructions.

Science chart research tranche

Added sourced profiles and dated records for Tom Scott, Hank Green, Sabine Hossenfelder, Steve Mould and Brady Haran. The Science chart now has ten researched creators.

Creator profile visual pass

Added a distinct dossier-style profile header, category colour system, live YouTube channel avatars, visual channel-scale cards and clearer event panels.

Thirty-profile launch set

Added 30 editorial profiles across technology, science, entertainment, gaming, food, lifestyle, beauty and reporting. Every profile has a verified YouTube channel ID, research focus and visible source trail.

Launch hardening

Added reporting, privacy requests, community rules, submission limits and a closed editor allowlist. Reworded candidate and shortlist claims.

YouTube importer connected

Added staggered channel-statistics and recent-upload refreshes. Imported data is shown separately from audience opinion.

Founding candidate index

Published a 1,000-record Wikidata seed for verification. It is not described as a list of the largest creators.