The AI claim, stated at its actual strength
The most AI-relevant item in this month's biotech flow is short. Radical Numerics says AI-designed viruses have treated the startup's first patient, with an experimental treatment targeting drug-resistant E. coli.1 That is the whole of what our dossier holds. It gives no trial size, no outcome, no regulatory pathway and no description of the model that designed the viruses. A first patient treated is a start, not a result.
Readers should also know how well this source has held up. The publication that carried the claim is a newsletter digest, and our measurement of that outlet in this corpus is that 6% of 16 checked claims held up.1 That is a small sample and it does not mean this claim is wrong. It does mean the claim should be read as the company's own statement, not as an established fact.
The field matters because phage-style approaches aimed at drug-resistant bacteria would be a very different business from the cancer and kidney programmes that fill the rest of the news. Nothing in the dossier lets us size that business or say how far along Radical Numerics is. What we can say is that an AI-designed therapeutic has, by the company's account, moved from design into a person.
The same sources show how loosely 'AI' and 'rejuvenation' claims get repeated
The same newsletter family carried a second, more extravagant claim. A company called Generation Lab says its treatment "blocks the systemic spread of aging in the bloodstream, reawakens the body's own repair mechanism and restores health and youth to multiple tissues."2 The same sentence appears, near-verbatim, across several issues of that newsletter.3,4,5 The product is described in one piece as an injectable combination of two existing drugs which the company calls 1 Generation.3 The piece's author describes being approached by the company with an offer to write about the treatment, and even to receive it.3
The dossier contains no clinical evidence for the quoted claim. The repetition shows how widely the company's marketing language spread, and it is not corroboration. We include it as a calibration point for how to read AI-adjacent biotech claims. One source quotes a company's fact sheet in several places, and only the company's wording is repeated.
The same newsletter edition that carried the Radical Numerics item also carries a headline about OpenAI's "biology data bid".6 The dossier holds only that headline. We cannot say what the bid involves, so we do not characterise it.
Where the checked data sits: conventional drug development
The rest of the late-2026 pipeline news is not AI-designed medicine. It is conventional, and much better documented.
Vor Bio announced final two-year Phase 3 results for telitacicept in IgA nephropathy on 22 October 2026. The annualised kidney-function slope was -0.6 against -5.7 mL/min/1.73 m² per year for placebo, a difference of 5.1 with p<0.001. The company also reports 89% and 81% reductions in the risk of confirmed 30% and 40% kidney-function declines.7 This is the company's own release, and it is the kind of readout that AI-designed drugs will eventually be compared against.
On 26 October 2026 Incyte presented Phase 1 data at ESMO for its KRAS G12D inhibitor INCB161734 in metastatic pancreatic cancer. The data cut-off was 6 July 2026 and the dossier gives n=54. Combinations with gemcitabine/nab-paclitaxel (n=26) and mFOLFIRINOX (n=27) showed what the company calls encouraging efficacy and favourable safety.8 Incyte also announced new collaborations to test the drug alongside navlimetostat from BMS and SNV1521 from Synnovation.9 These are early, small-sample results.
Early-stage timelines elsewhere are company guidance, not outcomes:
- Ernexa says it remains on schedule to submit its IND application for ERNA-101 in the third quarter of 2026, with a first-in-human Phase 1 trial anticipated in the fourth quarter.10
- Akari describes ADC-refractory patients as "a significant, growing unmet need" and an exclusive opportunity for its novel payload.11 The release it came from puts planned Phase 1 development for AKTX-101 in mid-2027.11
- Prelude expected an IND submission for PRT13722 by mid-2026 and a Phase 1 start in ER-positive breast cancer in the second half of 2026.12 The dossier does not confirm whether that submission happened.
- OPKO reported that its OPK8801001 showed roughly 20-fold greater growth hormone receptor antagonism than Pegvisomant, but only in vitro.13
Several of these sources are press-release wires. Our measured reliability for the main wire in this dossier is 53% of 6,322 checked claims, and for the 'catalysts to watch' piece it is 34% of 3,660.7,10 Company timelines are intentions, and they slip.
What audited numbers say about who can fund this
For the hard numbers, we turn to Incyte, the one company in the dossier with checked SEC-filing figures. Its net income was $32.6 million in fiscal 2024 and $1.287 billion in fiscal 2025, nearly 40 times higher. Earnings per share rose from $0.15 to $6.41.14 Operating cash flow went from $335 million to $1.413 billion.14 Cash at year-end rose from $1.688 billion in 2024 to $3.098 billion in 2025, and was $3.461 billion at the end of Q1 2026.14 The dossier does not say what drove the jump, and we do not guess.
The relevance to AI is practical. A discovery platform, AI-based or not, is only worth building if someone can fund trials. A company with over $3 billion in cash and R&D spending that was $1.628 billion in fiscal 202314 can afford several simultaneous early programmes, such as the pancreatic-cancer combinations above. The dossier gives no equivalent figures for Radical Numerics.
The royalty side shows the same pattern. BioArctic reported a Leqembi royalty of SEK 179 million for Q2 2026, up about 10% from Q2 2025. Excluding a one-time Chinese stockpiling effect of around JPY 5.3 billion in the prior-year quarter, royalty revenues grew about 43%.15 Revenue like this funds pipelines, and it comes from drugs that took years to reach the market.
Talent moves, and what they tell us
Idorsia appointed Begoña Carreño-Gómez as Chief Business Development Officer and Executive Committee member, effective 1 November 2026.16 The dossier does not say why, so we read nothing into it beyond the fact itself. We also found no checked evidence in the dossier of AI-specific hiring at any of these companies.
What to watch
- Radical Numerics: any disclosure of a trial registration, patient numbers or outcome. Without one, the milestone stays an assertion.
- The IND-to-Phase 1 sequence: Ernexa's third-quarter IND and fourth-quarter first-in-human target10 are the nearest hard tests of whether stated timelines hold.
- Incyte's pancreatic programme: whether the BMS and Synnovation combinations9 generate larger datasets than the n=54 so far.8
- Evidence standards for AI claims: a single-patient start does not compare with a 104-week randomised readout such as TELIGAN.7 The gap between those two is the real measure of how early AI-designed therapeutics still are.