For researchers working across fields

Your library is full of bridges you haven't found.

Unfathomed sounds your whole corpus overnight and surfaces the connections between papers in fields that never cite each other — a shared mechanism, a transferable method, a quiet contradiction. Every one arrives with the sentence it came from.

Start a spaceRead the argumentFree to 200 papers. No card.
PubMed · arXiv · OpenAlex · Semantic ScholarA locator on every numberYour library, never training data
Sounding · 142 papers06:10
EndocrinologyMicrobial ecologyEnergy-sensing kinase cascade
DEEPEST BRIDGEno citation path between them

Metformin's AMPK cascade and nitrogen fixation under soil stress run on the same energy-sensing kinase homologues.

“Energy-sensing kinase homologues regulated nitrogen fixation under stress.” — p. 7, Fig. 4

What a bridge is

Three kinds, and none of them are keyword matches.

A bridge is a claim about two papers, scored for how far apart their fields sit and defended with the sentences it rests on. You can dismiss it — that teaches the space what you care about.

01 · Mechanism

The same thing, two vocabularies

Two fields describe one cellular state under different names. Nobody searching either term would surface the other.

Immunology calls it T-cell exhaustion. Nutrition science calls it metabolic fatigue. The effect magnitudes are within 10%.

02 · Method

A solved problem, one field over

The limitation one paper names in its discussion is the thing another paper's instrument already resolves.

A neurology assay could not resolve clearance below 30 minutes. A sleep-medicine protocol gets to four.

03 · Contradiction

The disagreement you would rather find early

Two papers you trust assume opposite things about the same intervention — often from the same parent trial.

A stratified re-analysis kills the effect below BMI 27, which the renal cohort models as adiposity-independent.

Depth, not relevance

A link inside your own field is not news.

Every bridge is scored by how far apart the two literatures sit — whether they share vocabulary, whether anyone has ever cited across them. You set the floor. Raise it and the obvious disappears; the feed gets shorter and stranger.

Most tools optimise for the closest match. That is the one you would have found anyway.

Near

Same field, shared vocabulary

Useful, and you would have found it. Turn the floor up and these stop arriving.

Deep

Adjacent fields, translated terms

A concept mapped across vocabularies before the comparison is made.

Deepest

No citation path in either direction

Structural analogy rather than shared entities. Rare, and the reason people stay.

How it works

Import once. Sounded nightly. Four ways to work.

01

Bring your library

Connect Zotero, drop a folder of PDFs, or name a question and we assemble a starter library from PubMed and arXiv. Paywalled papers wait in your inbox until you upload them.

02

The nightly sounding

Every paper is read against every other paper in every space you own. Candidate bridges are scored for depth and confidence, then defended with the sentences they rest on.

03

Decide, and it learns

Pin a bridge to a matrix, ask about it, or dismiss it. Dismissals are training signal: tomorrow's sounding is shaped by today's judgement.

Bridges

The feed

Three at a time or a scannable ledger. Every card can be re-viewed through a lens.

Matrix

Extraction

Prompt a column in plain language. Every populated cell keeps its page and table.

Synthesis

The studio

Toss cells onto an evidence board and compile a review, a grant section or a brief.

Notebook

Questions

Cited answers, and the occasional polite interruption when another space knows something.

Provenance

Nothing appears without a locator.

Every extracted number carries the page, table and sentence it came from. Ask for a value that isn't in the paper and the cell stays empty and flagged — we would rather hand you a hole than a guess.

Which is also why the matrix survives a PRISMA audit, and why a draft compiled here can be checked line by line.

PaperNΔ IL-6
Smith 2024450−42%
Okafor 20241,204−18%
Lindqvist 202362−29%
Davies 2025——————

Smith 2024 · p.4, Table 2

Serum IL-6 declined 42% (95% CI 31–51) at week 26 versus placebo.

Who it is for

People whose work does not fit in one field.

Solo researchers

A review across 200 papers, or a dissertation drawing on three intersecting literatures. Your projects do not have to live in one folder to talk to each other.

Labs and research orgs

Everyone keeps their own spaces. The sounding notices when a colleague's corpus touches yours and tells you both — sharing stays an invitation you send.

Analysts and science journalists

Private documents, published papers and grey literature in one corpus. Traceable answers, and structured datasets extracted on demand.

The bridge you need is already in your library.

Free to 200 papers, no card. Import tonight; the first sounding is waiting in the morning.