An Evidence-First Research Portfolio

1. Research assets and public demonstrators are different layers

The catalog consists of two research families: Sovereign (30 assets) and Frontier (18 assets). The thirteen public demonstrators are an editorial access layer chosen to communicate concrete industrial problems; they are not a replacement count, and they are not one-to-one certificates for the underlying 48 assets.

An interactive demonstration is a useful technical conversation starter. It is not the same as a frozen reproducible release, independent laboratory result, deployed reference customer or industry certification.

2. Preserve identities across revisions

The same name can refer to a research lineage with materially different performance and failure behaviour across revisions. Every material claim should belong to a source or artifact revision, test conditions and dated record. Portfolio cataloguing must avoid promoting newer prototypes as equivalents of historical originals without a reconciliation.

For the Frontier family, original and newly reconstructed candidates are particularly important to distinguish. The public site provides identifiers and safe descriptions; private lineage and proof mappings are reviewed only under a defined technical scope.

3. Negative results are assets

A failed timing constraint, missing external capture, inconclusive fault injection or higher-than-expected tail is not an editorial embarrassment. It tells the evaluator where the acceptance boundary lies. A strong evidence program keeps negative observations instead of silently replacing them with the best run.

This also prevents accidental transference: a formal property checked for one module does not certify adjacent modules, and a software-only test does not establish physical equivalence.

4. Design the buyer's reading path

Start with a recognizable cost or risk, not an unfiltered inventory. Follow with a short interactive proof and a clear qualification boundary. Only then introduce the architecture family, related assets and relevant research note. Let the evaluator choose licensing, funded validation or deeper technical diligence without exposing implementation details prematurely.

The objective is not to manufacture a large readership. It is to reduce friction for qualified technical counterparties while retaining intellectual property and the ability to negotiate use rights.

5. What a successful review produces

At the end of a scoped review the parties should agree on the problem, candidate identity, evidence accepted, missing tests, evaluation rights, expected economic value and a go/no-go decision. Reproducibility and commercial clarity matter more than blanket maturity language.