Meet InkTide

A good research result deserves a clear argument and a figure that makes its evidence visible. InkTide connects the reading, writing and checking needed to get there.

Learn what a good paper makes possible

A strong paper can teach how to pose a question, order evidence or make a comparison visible. InkTide records the job each choice performs and the conditions under which it transfers. A new manuscript still begins with the author's own findings, data and intended audience.

The project retains inspected observations on argument, language and visual design, then asks whether a lesson fits a new author task. It keeps evidence for the scientific claim separate from evidence for the writing choice.

Different roles, connected responsibilities

Coordination, curation, writing, figure design and independent review have distinct responsibilities. Writing produces original text and an evidence map; figure work preserves units, transformations and comparisons. The reviewer reads a frozen output against the author materials and returns located repair points.

What the actual comparison showed

The whole-manuscript study generated 26 complete drafts from fixed synthetic facts, with 17 primary paired reviews and five position-reversed secondary reviews. The held-out test completed four of five registered groups. InkTide was preferred in three, but only two satisfied strict quality-win criteria; one passing pair tied and one input was blocked.

All four held-out English-score pairs tied. The current evidence is strongest in particular argument and repair cases: explaining what a finite test shows, preserving a theorem’s quantifier, and correcting a causal conclusion that fluent language had obscured.

Why these changes matter

In one held-out case, both drafts preserved zero disagreement after complete lineage removal. The guided draft separated that observation from the deterministic operation argument, keeping the conditions on the learner visible. In another development case, a missing minimum/every quantifier invalidated a conclusion. The repair restored that condition in the conclusion and made the decision branches explicit in Methods.

The failure record matters too. One guided conclusion kept its numbers and fluent English but crossed the causal boundary of the supplied ablation. That case reached a tie after repair. InkTide’s value includes a way to locate and revisit such choices, alongside the cases that improved.

A growing library, grounded in the author’s work

The current snapshot contains 466 main-paper PDFs, 459 reading cards and 416 admitted study references. Lessons retain source identity and transfer scope. Personal style remains open to real author annotation; the present comparison uses model-only judgments and has not established general writing or professional-English superiority.