03 · Principal limitations
Formal Rubric, Interpretive Precision
The Greatness rubric is defined, but the resulting scores are still judgments. The difference between an 8.9 and a 9.0 is not a scientifically measurable interval. No rubric can prove that the artistic distance between those values is exact. The decimals should therefore be read as tools for ordering close comparisons, not as claims that artistic value can be measured to the tenth or hundredth with objective precision. The framework disciplines judgment. It does not remove the judge.
The Ease Trade-Off
Because Ease contributes 20% of Total, long, demanding, bleak, structurally difficult, or highly contextual films may rank lower in the balanced leaderboard than they would in a pure artistic hierarchy. This is intentional. A lower Ease score does not reduce Greatness. It records the conditions of encounter. Films such as Lawrence of Arabia, Barry Lyndon, Seven Samurai, or Jeanne Dielman may therefore receive lower Total scores because the default ranking includes practical recommendation strength. Users seeking the closest equivalent to a traditional canon should sort by Greatness rather than Total.
Inherited Biases and Curatorial Counter-Bias
GCML inherits some structural biases from the external lists it incorporates. Certain countries, periods, genres, filmmakers, and traditions arrive with much stronger institutional visibility than others. GCML Curated can counter some of those distortions by admitting films independently of the tracked lists. Direct curation, however, introduces another source of bias: the knowledge, access, preferences, comparisons, and blind spots of the curator. Expansion therefore does not make the canon neutral. It changes where its subjectivity enters.
Finite Coverage
Seven hundred thirteen films constitute a large canon. They do not constitute a comprehensive history of cinema. Some traditions are mapped deeply. Others are represented lightly. Some important films are already well mapped by existing canons. Some belong to routes this project has not emphasized. Some have inevitably been overlooked. Absence should therefore not automatically be read as a judgment of unworthiness. The list is intended to remain revisable as films become available, reputations change, restorations alter access, and gaps become more visible.
Why 713?
I built the canon around two deliberate choices: the films were selected through explicit curatorial criteria, and I set the overall size through an editorial boundary. That boundary is fixed by design, and it can grow or shrink as future evidence warrants. No film canon has a "correct" size—whether it contains 100, 500, 1,000, or 713 titles, the number is always a line someone decides to draw. Round numbers only feel natural because list culture has made them familiar, not because they carry any special methodological weight.
GCML began with the films that appear across Sight & Sound, IMDb, and AFI, and then expanded through direct curation. I set 230 curated additions as the scale of that expansion. I chose that number openly and for personal reasons: twenty-three has long mattered to me because of Michael Jordan, and my birthday is March 2. I've also always preferred specific integers to round ones.
After accounting for overlap among the three external lists and the curated additions, the working canon reached 713 films. That total isn't meant to represent the "right" size of a film canon. It's simply a clear, documented boundary with a traceable history, and any future additions or retirements should be justified by the films themselves—not by the desire to preserve a particular number.
Eligibility and Boundary Cases
GCML primarily evaluates feature-length cinematic works. Narrative features, documentary features, and animated features are eligible by default. Extended, anthology, or hybrid works may be admitted by exception when their critical or historical standing is substantially cinematic and when the work can be evaluated principally through cinematic form rather than merely possessing cinematic qualities. Conventional episodic television, standard miniseries, video games, filmed stage productions, music videos, interactive works, and gallery installations are normally outside scope. Boundary cases should be labeled explicitly rather than silently treated as ordinary feature films. Twin Peaks: The Return remains an acknowledged Extended Work exception, not a rule that dissolves the boundary.
Coverage Audit Is Still Incomplete
GCML can report parts of its coverage clearly, but it does not yet quantify every blind spot with equal precision. A stronger transparency audit should continue to measure where the canon is dense and where it is thin across country, period, language, genre, format, filmmaker concentration, and GCML Curated status. Low representation does not automatically prove a curatorial failure. It identifies a place worth examining.
Usefulness Remains Testable
GCML describes itself as a canon designed to be used. That claim is ultimately empirical. The existence of Film Finder, Tonight, Change the Lens, Collections, Double Feature, Cinema Passport, and other tools does not by itself prove that visitors find them useful. Usage, repeat use, recommendation-to-film clicks, feature depth, and return behavior can eventually provide evidence about which parts of the project actually help people choose and explore films. Until that evidence accumulates, practical usefulness should be treated as a design goal that can be tested, not as a fact established by the site’s existence.
Viewing Mood Standardization
Viewing Mood descriptors vary in specificity. Some categories are relatively precise, while others cover broader emotional territory. That makes the system useful for practical discovery but less formally standardized than Greatness, Engagement, or Ease. Future revisions may refine the taxonomy as actual usage reveals which distinctions are useful and which merely create noise.
Aging External Ranks
The IMDb snapshot from August 3, 2026, Sight & Sound’s 2022 poll, and AFI’s 2007 list are fixed historical reference points. IMDb will continue to change. Sight & Sound and AFI will remain fixed until new institutional editions appear. External rankings should therefore be read as dated evidence, not permanently current measurements of consensus.
Metadata and Editorial Revision
Film descriptions, mood labels, inclusion arguments, metadata, and even curatorial judgments may continue to benefit from revision. A large hand-curated project inevitably contains unevenness. The appropriate response is not to pretend the dataset has reached a final state, but to document corrections and improve it when the evidence warrants doing so.
