Skill Trade

Hamlet

COMPANION

Who changes most?

Ophelia moves furthest from where she starts. Laertes follows. Hamlet moves away and back; Horatio has the smallest end-to-end shift.

Lines: distance from each character’s first qualifying act. Scores: first-to-last shift. All charts share a 0–1.1 scale; higher means more change.

  1. 1. Ophelia1.06
  2. 2. Laertes0.92
  3. 3. Polonius0.66
See all 9 characters, including Hamlet and Claudius
  1. 4. Gertrude0.61
  2. 5. Claudius0.58
  3. 6. Rosencrantz0.55
  4. 7. Guildenstern0.34
  5. 8. Hamlet0.23
  6. 9. Horatio0.18

Each act averages six themes and six actions from Jev’s speech annotations. Change is the Euclidean distance between those 12-number profiles, with equal weight per speech and dimension. Only acts with at least eight speeches count; gaps stay blank. Characters need two qualifying acts, so their start and end acts can differ. These are model judgments, not a measure of personal growth.

How this was built

The text is the public-domain Gutenberg edition of Hamlet, parsed into acts, scenes, and speeches by ordinary code. A speech is every time a character opens their mouth to speak until another character interrupts them or the scene ends. Hamlet's one-word "Whither?" and a thirty-five-line soliloquy each count as exactly one. Every annotation comes from Jev, TypeSafe's System One model, which answers typed questions instead of chatting.

Each of the play's 190 stage directions was turned into yes/no questions ("after this direction, is Horatio on stage?") to drive the Now-on-stage rail. Each of the 1,170 speeches got one request with ten questions batched: who it addresses, dominant emotion, six theme probabilities, intensity, plot weight, fame, and dramatic irony. All 11,700 judgments ran in 31 seconds and cost seven cents.

The page reads the play through two lenses, toggled in the header. Actions is the actor's lens: six essential actions chosen by the editor (act, speak, question, move, keep, love), each scored as the probability that the speaker is doing that to their listener ("act" meaning performs or feigns; "speak" declares; "question" interrogates; "move" stirs toward feeling or deed; "keep" guards or withholds; "love" cherishes). Themes is the scholar's lens: revenge, death, madness, deception, love, corruption. The highlighted colors, each speech's border, and the aggregate views all follow whichever lens is active.

Because the judgments are stored as data, the panel can also aggregate them: the Speaker view averages a character's theme probabilities up to the selected line, or over the scene, act, or play; the Between view averages every speech two characters address to each other. These are averages of model judgments, not measurements, and the Between view inherits any addressee errors.

Fifty of the speeches, spread across the play by position, plot weight, fame, and irony, carry a second opinion: the same ten questions put to a chat model (Claude Haiku 4.5) three times, averaged. Those speeches are marked "2 models", and their panels have a toggle to flip between the two models' answers.

Built with Claude for Skill Trade. Annotations are machine judgments; enjoy arguing with them.