T27.AI

The record

Everything citable, with its identifier.

One page holding everything that can be cited or checked: the preprints, the archived datasets and their DOIs, the patches sent to other people's projects, and the accounts that are actually mine. It exists so a reviewer does not have to take a claim on trust or reconstruct the trail from a CV.

What is on it

Three arXiv preprints

arXiv:2605.28405, arXiv:2606.05017 and arXiv:2606.09686. Listed with their identifiers, so the version you read is the version being referred to.

Ten Zenodo records, each with a DOI

The GoldenFloat format description and the Trinity B001 to B007 collection, among others. A DOI resolves to a fixed deposit, which a repository link does not.

Six patches sent upstream

Changes offered to openXC7 and nextpnr-xilinx, numbered #109 to #115, covering timing constraint parsing, clock buffer placement and IDDR initialisation. Their state is shown on the page as it stands, not as it was hoped.

Accounts and identities

ORCID and the channels that are genuinely mine, so a name collision elsewhere does not get attributed here.

Why it is arranged this way

The corpus carries a date

The links were checked on a stated date, which is shown on the page. A list of references with no check date says nothing about whether it still resolves.

Known discrepancies are counted, not hidden

The page computes its own broken and contradictory entries and shows how many there are. A reference list that never admits a bad entry has simply never been checked.

Identifiers over links

A DOI or an arXiv id survives a repository being renamed, moved or made private. Where one exists, it is what is given.

Cite the DOI or the arXiv identifier rather than a repository URL — those survive a rename. If something here does not resolve for you, that is worth an email: the discrepancy list is only as good as the last time someone looked.