Inquiries →
Research philosophy

We publish. We protect. We say which is which.

A research company that says nothing is indistinguishable from one that has done nothing. A research company that says everything has nothing left to license. This page is where we draw the line, in public, on purpose.

Disclosure

Where the line is, and why it is there.

Everything on this site describes what the research is establishing and why it is hard. None of it describes how. The boundary is stated rather than merely observed, because a stated boundary is itself evidence that there is something behind it.

Published openly
  • The granted patent in full: claims, abstract and prosecution history, all matters of public record.
  • Priority and grant dates, which establish precedence without revealing method.
  • The questions the program is pursuing, and why each is difficult.
  • Capability at the level of outcome, with conditions and limits stated.
  • The company record: founding, structure, financing, restructuring.
  • Technical writings that document the research as it develops. Not peer-reviewed papers, but a dated record of the thinking.
  • The existence and scope of unpublished work, named, dated and bounded.
Held as trade secret
  • System architecture, pipeline design and model composition.
  • Training corpora, annotation schemes and data provenance.
  • Feature extraction, lexicon construction and filter design beyond the granted claims.
  • Benchmark figures, which reverse-engineer an approach as readily as they demonstrate it.
  • Deployment configurations and any work performed under confidentiality.
  • Subject matter of contemplated and unfiled applications.

Parties with a substantive interest in licensing, research collaboration or diligence can be given considerably more under an executed confidentiality agreement.

How the work is conducted

Applied research, held to a documentary standard.

01

Priority before publication

Novel subject matter is documented and, where appropriate, filed before it is discussed outside the company. Disclosure follows protection; it does not precede it.

02

Negative results retained

Approaches that failed are documented with the same care as those that worked. A research record that reports only its successes cannot be audited, and cannot be trusted by anyone doing diligence on it.

03

Bounded claims

Capability is stated with its conditions and its limits attached. We do not describe performance without describing the circumstances under which it was obtained.

04

Independent collaboration

Specialist engineering, language-technology and research partners are engaged under agreement for defined scopes of work, with intellectual property boundaries settled in advance.

05

Ethics inside the design

Consent, provenance and the limits of voice synthesis are treated as constraints on the architecture rather than as a policy layer added at the end.

06

Independent financing

Venture financing supported the company's early years. Since 2018 the research has been directed and financed primarily by the founder, which is why its direction has not been set by a funding cycle or a product roadmap.

Under agreement

There is considerably more behind the line.

Licensing, research collaboration and diligence requests are answered under an executed confidentiality agreement.