Why rDNA.ai?

In 2020, the SEC's FOIA office stopped releasing newly-expired FOIA documents and, separately, changed the rules so that biopharma companies were no longer required to send unredacted versions of alliance contracts to the SEC. The combined effect was to draw a line under a four-decade record. Contracts already in the SEC's possession in 2020 with as-yet unexpired confidentiality would never be FOIA-released; new alliances would not arrive at the SEC in unredacted form to begin with. As a practical matter, the most recent FOIA-released versions of biopharma alliance contracts are, and will remain, for deals commenced around 2017 or earlier.

Two years ago in this space, I argued that AI applied to SEC interactive financial-statement notes could fill the gap. Public biopharma filers had been describing material alliances in their quarterly notes for more than a decade, and AI could harness those descriptions at scale to substitute for what unredacted SEC-filed contracts had previously provided.

It turned out not to work — not for the deals that mattered most.

The problem is one of materiality. The biopharma originators most likely to enter a transformational alliance are also the ones for whom a single deal is rarely material to the financials of the consolidated entity that files the note. A $50 million upfront payment that would have produced 30 pages of unredacted contract detail under the old regime now produces a paragraph in a 10-Q — sometimes less. Multiply this across the 1,500 biopharma SEC filers, and the resulting corpus is too thin to support the comparability, trajectory, and counterparty-discrimination analysis that the old contract record made routine.

I should have anticipated this. "The FOIA-disclosed contract record was load-bearing precisely because it carried the terms — the royalty rates, the milestone schedules, the diligence triggers, the offset and stacking mechanics" — and those terms are exactly what self-reported notes don't carry, by intention.

So what survived 2020?

The patent record did. Patents are filed publicly, prosecuted in the open, examined against prior art that is itself disclosed, and litigated on a docket. Across 30+ jurisdictions, every meaningful biopharma asset leaves a trail of composition-of-matter claims, method-of-use claims, divisional filings, oppositions, IPRs, family-member geographies, and citation neighborhoods. None of this depended on FOIA. None of it has gone dark.

What the patent record contains, and what biopharma BD&L decisions actually require, turn out to overlap more than is generally appreciated. A composition-of-matter claim with a 2042 expiry and a thin prior-art neighborhood predicts a different alliance term spread than a method-of-use claim with three fast-followers filed within twelve months. A pattern of divisional filings concentrated in the top seven pharma markets predicts a different post-signing royalty trajectory than a single PCT with no national-phase entries beyond the US and EU. An assignee's track record of defending its in-licensed IP through IPR and EPO opposition predicts a different innovator royalty-offset exposure than an assignee who has historically settled.

These are observable, structurally available, and legally durable signals. They survived 2020 because they were never under FOIA in the first place.

rDNA.ai is being rebuilt around them.

The new product is a structured read of the patent record applied to specific BD&L decisions: which asset, which partner, on what terms. The starting point is composition-of-matter scope, claim distinctness, prior-art density, enablement quality, and family geography for any biopharma asset whose primary patents are findable in the public record. The output is a partner-and-terms read keyed to comparable historical deals where the IP profile was similar.

There's an honest version of what this is and isn't. It isn't the synthetic substitute for the FOIA-disclosed contract record I was hoping AI on financial-statement notes would produce. It is, more modestly, the most objective signal still available about a biopharma asset's commercial trajectory and a counterparty's post-signing behavior — the patent record, read with the rigor that contract terms once warranted.

What the patent record tells you about the deal.

I'll be writing about this on LinkedIn over the coming weeks — five articles tracing the argument from 2020 forward, into how the patent record substitutes for what was lost, and into the specific BD&L decisions where the patent read materially changes the answer. The first article goes up shortly. As before, I welcome feedback on whether this approach is useful and reliable.