Evidence changes the thesis
A patent examination, reader response, founder conversation, customer experiment, sale, renewal, or failed test matters when it changes what the firm believes or does next.
The firm
About the firm
Intellectual Frontiers turns unique human insights into evidence, and evidence into decisions.
Most companies are organized around things they already know how to sell. Intellectual Frontiers is organized around things it is not yet sure are true. Work starts with an observation, not an instruction to build a company. The first question is what the firm might know, see, control, or be unusually capable of doing that matters to somebody else.
That possible advantage is Native Alpha. It is not proved by a clever idea, a patent, expertise, important relationships, or access to AI. Those may be signals. Evidence is what eventually proves or disproves the claim.
We read AI the same way. The obvious use of AI is the same work, faster or with fewer people. We look for the work that did not exist before: every denial read the day it arrives, a device's evidence standing before the study is written. That is what produces patents, companies, and impact worth owning.
IP develops the thesis. Press makes it inspectable. Capital asks whether resources should move and searches for founders already pursuing it. Studios runs the cheapest useful experiment when no better-positioned builder exists. Network connects the people needed at every point. Discoveries move in every direction, and each unit can change the work of the others.
What counts as progress
The best builder may be elsewhere
If a founder has stronger evidence, customer access, distribution, or execution, Capital can back that founder. The point is to support the best expression of the thesis, not recreate it.
Building is conditional
Studios builds when the search has failed, the evidence remains compelling, and a cheap experiment can answer the next important question. A company is not the default destination.
How work enters the firm
- 01
A researcher on our staff hits a problem in the course of a research program: a readmission report nobody trusts, a device feed that drops overnight, a title process that takes three days, and logs it with the date and the source.
- 02
Before any money moves, we write down what would prove the idea wrong. If nobody on the team can name that test, the idea is not ready.
- 03
The test runs against real data or a real operating environment: an insurer's authorization history, a device's telemetry feed, a trial site's enrollment records. The result is recorded whether it holds or not.
- 04
One named person decides: continue, change, back someone else, build it ourselves, or stop. The decision gets a date and a written reason.
- 05
Whatever comes out of it, a granted claim, a dataset, a working integration, a company, goes back into the shared record so the next one is cheaper.
A thought finds its best form
The result may be a publication, a patent or licence, a partnership, an investment in somebody else, a small experiment, software, a new company, continued research, or a decision to stop. The firm does not manufacture activity. It improves the decision about which form the evidence supports.