Research intelligence for people who cannot afford to miss it.
Open access solved availability. It did not solve discoverability. Roughly three million research articles are published every year, and the number climbs annually.
Every sheet is a paper. One of them was written for you.
A physician finishing a full day of patients cannot track the literature relevant to their practice. The finding that would have informed a diagnosis goes unread.
Duplicated effort is the quiet tax on every lab. Work published in a neighbouring field six months ago stays invisible until it is too late to matter.
The method that halves your compute, the material that solves your failure mode, the model that outperforms yours. All published, none surfaced.
Free, machine readable, no gatekeeper.
Dollars per paper two years ago.
Long form audio already fills the commute.
We read the world's research continuously, score every paper against the specific context of your work, and keep only what is genuinely relevant. A digest you can read in ten minutes, or a briefing you can listen to on the way in.
Everything else, you never have to see.
Eight sources, deduplicated by DOI with merge rather than drop, ranked by per-domain calibrated profiles, summarised once and shared.
Recommendation behaviour is fully specified by per-domain profiles. Cardiology, condensed matter physics and macroeconomics run through the same ranked pipeline with different calibration. Adding a field is a data change, not an engineering project, so coverage compounds while cost does not.
Scholar alerts fire on string matches and get ignored within a month. Single-vertical products serve one audience and cannot cross the boundary where the most valuable findings usually sit. ReadSearch is broad and precise at once, which is only possible because the engine is parameterised.
Entry is bottom-up and consumer priced, which removes the procurement cycle entirely. Expansion runs upward into team seats, institutional licences and an API tier, where the same engine sells at materially higher price points. Population figures are working estimates, to be validated.
No subsidised cohort, no conversion funnel to fix later, and no ambiguity about whether the product is worth paying for.
Summaries and audio are produced per topic-set and per paper, then served to everyone following those topics. Adding a user to an existing cluster costs close to nothing.
Annual plans that cut payment fees and lift retention, team and institutional seats, and an API or alerts tier for embedded use.
| Unit | Cost | What drives it |
|---|---|---|
| One carousel | $0.0013 | Distinct papers. Cached forever, shared globally. |
| One episode | $0.0275 | Distinct topic-sets × 30 days. Twenty-four hour cache. |
| One search parse | $0.0010 | Queries per user. |
Measured against 56 real cached carousels and a verified production episode build on 15 August 2026: 1,641 characters, 105 seconds. The episode figure replaces an earlier estimate that was 64% too high, because it carried an old text-to-speech rate we no longer use.
| Line | Per user | Basis |
|---|---|---|
| Revenue | $3.99 | Single tier, monthly |
| Payment processing | ($0.42) | 2.9% + $0.30, exact |
| Audio briefings | ($0.46) | 555 topic-sets × 30 days × $0.0275 |
| AI summaries | ($0.15) | ~117k distinct carousels × $0.0013 |
| Search parsing | ($0.02) | ~15 structured queries per user |
| Infrastructure | ($0.20) | Itemises to $0.12. Pad retained. |
| Gross profit | $2.74 | 69% gross margin |
Audio cost scales with distinct topic-sets, so clustering is the entire margin story. Today's ratio of 1.8 users per set comes from nine seeded accounts on a legacy nine-topic picker; the live taxonomy now exposes 144 followable leaves, so early diversity will run worse and push toward the downside row. The 30% overlap assumption is modelled, not measured. Even at the floor the business clears 53%, and annual billing removes most of the payment drag, which is the largest single cost at this ticket.
At a $3.99 ticket the acquisition channel has to be cheap by construction, so growth is built around organic and content-led distribution rather than paid performance. Annual billing is the primary margin lever: it removes most of the payment-fee drag, lifts retention, and pulls a year of cash forward per user.
Somewhere in today's research is the paper that changes how you work. It exists. It is published. The odds you will ever see it get worse every year.
We are building the thing that puts it in front of you, while it can still make a difference.