About
Most anime rankings either average user votes (MyAnimeList) or follow popularity (YouTube top 10s). Shonen Codex applies a single critical rubric — six criteria with explicit weights — to 100 widely-known shonen titles. The scores will sometimes disagree with the community consensus. That's the point. Disagree with the rankings; the rubric is published.
The rubric
| Criterion | Weight |
|---|---|
| Story & narrative | 25% |
| Character writing & growth | 25% |
| Themes & emotional resonance | 15% |
| World-building & power system | 15% |
| Animation & direction | 15% |
| Cultural impact | 5% |
Scores are 1.0–10.0 with one decimal allowed. A 10 means “definitive of the medium” and is used sparingly. A 7 means “good but flawed.” Below 5 means “significant weaknesses.”
What's in the catalogue
100 shonen anime whose source manga ran in a major shonen magazine (Weekly Shonen Jump, Bessatsu Shonen Magazine, Weekly Shonen Magazine, Weekly Shonen Sunday, Monthly Shonen GanGan, Shonen Ace, and similar) and which meet at least one recognition threshold — 100K+ MAL members, a major streaming license, 50+ episodes, or significant Western cultural footprint.
Multi-season franchises are represented by a single season — the one that best captures the franchise's peak quality or cultural inflection point. The editorial choice is documented per show.
How rankings are generated
The rubric is mine. I picked the six criteria, set the weights, and wrote the scoring scale — a 10 means definitive of the medium and gets used sparingly; a 7 means good but flawed; sub-5 means real, namable weaknesses. Story, character writing, themes, worldbuilding and power-system rigor, animation and direction, cultural impact — these are the lenses I argue with friends about. Whether Bleach earned its Big Three slot. Whether AoT was overrated by its 2013 moment. Whether Demon Slayer's animation is doing more work than its writing.
What's automated is the application, not the taste. Each show runs through Claude (Anthropic's LLM) against the same published prompt — same scoring anchors, same requirement to cite specific arcs, episodes, sakuga, and directorial choices. Hinokami Kagura. Levi's debut spin against the Female Titan. Joe versus Mendoza. Killua's transformation. Generic praise gets rejected; the model has to show its work. Every ranking row is stamped with the model and prompt version, so the catalogue stays internally consistent and a rubric revision regenerates only the affected entries.
The selling point isn't AI-with-opinions. It's a single critical lens applied to 100 titles without drift — something no human reviewer has the patience or attention span to do honestly. The taste is mine; the throughput is the machine's.