Microsoftfor Startups partner

Lumen / Applied research intelligence

The record that changes your program is already public.

Founders at Series A and B deep tech companies burn hours on patent searches, IP reviews, and competitor news, and still cannot tell which of millions of records touch their program. Lumen puts autonomous researchers on it instead, and tells you exactly what changed.

Run the system Book a walkthrough
Scroll to open the field
Seven days of public output

204,000 records were published last week.

Every cell is one patent, paper, grant, trial, or regulatory event that entered the public record in seven days. Colour is the source. Nothing here is secret and nothing here is addressed to you.

01 / 03
204,000
New public records a week, across six streams
0.003%
Of that week reaches a decision brief
6
Sources read as one evidence layer
24 / 7
The perimeter is never not being read
The founder's week

Nobody misses the announcement. Everybody misses the evidence.

You already know the drill. A founder or a lead scientist disappears for two days into patent databases, IP reviews, journal alerts, and competitor news, and comes back with a folder of maybes.

There are millions of records and no honest way to tell which of them touch your program. Not one is decisive on its own, so each gets read, shrugged at, and forgotten. Meanwhile the filings, papers, awards, and hires behind a rival program have been public for a year.

The usual answers cost more than the problem. An in-house intelligence team runs into six figures and is a hire you cannot justify before Series C. A consultant charges thousands for a report that starts ageing the day it lands, and you still do not get the underlying records.

The question Lumen answers

What is moving around this program that we have not connected yet?

The system, end to end

From public record to decision.

01 / Intake

Everything public arrives as evidence.

Four streams enter one layer: patents, papers, awards, and news. No source is read in isolation, because no single source is conclusive on its own.

What a scout is

A scout is not a saved search.

A scout is a standing instruction to watch one part of your field. You give it an objective in plain language: a technology, a claim area, a set of organizations. It decides which sources to read and what counts as a change worth reporting. A saved search returns the same query forever. A scout keeps working out what to look at.

Cross-finding convergence

One record is noise. Five that converge are a finding.

Five ordinary public records, none of them decisive alone. Watch them converge on one organisation moving in one direction.

Public evidence / ordinary alone
  1. PATENTUS 18/604,215
    Continuation adds a 2 to 5 nm coating range
    PUBLIC+0.26
  2. PAPERbioRxiv 2411.08
    Continuous conversion at the same cascade ratio
    PUBLIC+0.21
  3. GRANTNIH R01-EB0
    Scale-up funding awarded to the same group
    PUBLIC+0.18
  4. HIRINGORG / ROLE
    Formulation science lead appointed
    PUBLIC+0.13
  5. CLINICALNCT 06812774
    First-in-human protocol registered
    PUBLIC+0.19
Below threshold

Nothing here is a signal yet.

Every one of these has been public for weeks. Read alone, each is ordinary, which is why each gets read alone and forgotten.

PROGRAM IMPACTPENDING
STATEWATCHING
How it works, screen by screen

The whole system, in the order you use it.

Seven screens, from defining a program to acting on a brief. Every one of these is the running product, not a mockup.

The project breakdown screen, showing the domains, key terms, and entities Lumen extracted from a CRISPR program's own documents.
01 / You brief it once

Give it the program, in your own documents.

Upload what already describes the work. Lumen reads it and pulls out the domains, key terms, entities, IP flags, and risks, then keeps that breakdown as the thing every scout later judges a record against. Edit any of it inline, and rebuild when the program moves.

How the system behaves

Four rules it does not break.

01traceable

Every claim keeps its receipts.

No finding exists without the records underneath it. You can open the filing, the paper, and the award that produced the conclusion, and disagree with it if you want to.

02relative

Relevance is defined by your work.

The same patent is material to one program and noise to another. Lumen scores against your technology and roadmap, not against a keyword list that any competitor could guess.

03continuous

A search is a snapshot. This is not.

A freedom-to-operate search starts ageing the day it is delivered. A watched perimeter does not, because the landscape is re-read continuously and you are told what moved.

04auditable

You can see why it was surfaced.

Every finding carries the trail that produced it: when the record was ingested, what it was screened against, what it was cross-referenced to, and the moment it was escalated. Nothing arrives as a verdict you have to take on trust.

Built for R&D teams in deep tech

Find the next one while it is still evidence.

Bring one program and the organizations you already worry about. We will show you what the last ninety days of public record say about them.

Book a walkthrough See pricing