Tacomancy

Prototypes

Twelve surfaces, drawn before a line of code.

Each was designed from a prompt written to stand alone: the overall shape first, then eight surfaces, then three dashboards. They are live below. Tabs switch states, rows select, filters filter. Sample content is placeholder research; the mechanics are the point.

00 · The shape

Navigation model and visual language

Not a screen. The rail, the surfaces, how a surface occupies the window, and the type and colour everything after it inherits. Runs once; every later prompt assumes it.

Open full size

01 · Surface

Question Inbox

Where two-keystroke captures land, each with its provenance visible. Age is a sortable fact, never a state. No badge, no red, nothing owed. Shown at three years and three hundred questions.

Open full size

02 · Surface

Reader

PDF reading with highlights that have identity, so a note can link to a passage. Selecting text offers a note, a question, or a link. Ingest review shows what came back from the iPad, including anything that could not be re-matched.

Open full size

03 · Surface

Research Question view

A promoted question, answered by reading. Supporting and opposing sources are structurally separate. The working answer is a position held at a time, and its history reads as changing your mind, not as a diff.

Open full size

04 · Surface

Hypothesis view

A promoted question, answered by testing. Criteria are recorded before evidence arrives, and the state is computed from them. There is no status control. Falsified looks as solid as supported.

Open full size

05 · Surface

Experiment view

A run designed here and executed elsewhere. Artifacts are the one thing in the vault that exists nowhere else, so plots and snippets are laid out to be looked at, not filed as chips.

Open full size

06 · Surface

Scout Queue

Proposals from scheduled agents, sorted like a stack rather than read like mail. Two lanes: Review is a queue and Skim is a feed. One card in hand, one key per decision, ranking that never hides anything.

Open full size

07 · Surface

Vault

Ordinary notes, links, backlinks, tags, and folders. Deliberately unremarkable. The only differences: questions and source stubs are first-class in the tree, in autocomplete, and in backlinks.

Open full size

08 · Surface

Home

A landing, not a command center. One column of sentences and routes. The only numbers are ones you can click through, and the only thing that raises its voice is a Scout that has quietly stopped seeing anything.

Open full size

09 · Dashboard

Loose Ends

What is incomplete or broken, as a punch list. Counts per group, never a total. Every row resolves in one click, and one of those clicks is always mark deliberate.

Open full size

10 · Dashboard

Scout Activity

Are the agents earning their keep? Accept rate on Review-lane items only, source health per Scout, and every row edits its brief in place. The one dashboard where a line chart is the right shape.

Open full size

11 · Dashboard

Question Map

Where is my work thin, and where am I relative to the field? A sorted coverage matrix, a labeled attention scatter, ranked lists. Not a force-directed graph, and never will be.

Open full size