October 2024 — Monthly Briefing¶
1. What was read¶
823 research articles from 33 of the 33 journals on the roster.
A further 165 records (17% of the 988 this month's harvest returned) carry no abstract and are not research: news, correspondence, editor highlights and obituaries. PubMed types them Journal Article and nothing else, so the publication-type filter cannot see them — across the whole index they are 22.7% of it and 35% of general, and 3,801 of them are JAMA. Every count below is over the full harvest and is inflated by this; the figure above is not.
| General | Cancer | Immune | Bioinfo | Sysbio |
|---|---|---|---|---|
| 450 | 154 | 87 | 271 | 26 |
Largest contributors: Nature 124, Nucleic Acids Res 96, Science 83, Cell 52, Bioinformatics 50, N Engl J Med 45.
2. Net change on the thread curve¶
| Thread | This month | Cumulative | Status |
|---|---|---|---|
agentic-ai |
+1 | +5 / −0 | forming too-early → forming |
cell-therapy |
+8 / −3 | +136 / −42 | forming |
ctdna-mrd |
— | +44 / −22 | forming |
foundation-models |
+7 / −3 | +67 / −45 | splitting |
immunotherapy-resistance |
+5 / −1 | +146 / −60 | forming |
interception |
+2 | +77 / −16 | forming |
method-credibility |
+1 / −12 | +7 / −279 | crisis |
neoantigen-vaccines |
+1 | +54 / −23 | forming |
neuro-immune-microbiome-axis |
+2 | +102 / −13 | forming |
next-gen-checkpoints |
+1 | +48 / −30 | splitting |
single-cell-3d-genome |
+1 | +45 / −15 | forming |
suppressive-population-inversion |
+1 | +34 / −45 | contested |
Status is derived by rule, not asserted. Fewer than 5 cards → too-early; then refutes ≥ 2×supports → crisis; refutes > supports → contested; supports ≥ 2×refutes → forming; else splitting.
3. The evidence that moved it¶
A randomised controlled trial of a large language model, in October 2024. 39009780 (Nat Med): outpatient reception via collaboration between nurses and an LLM, randomised. This corpus spent 2025 recording the absence of randomised clinical evidence for language models and building arguments on that absence. The absence was not real.
Pathology foundation models reach Nature. 39232164: a pathology foundation model for cancer diagnosis and prognosis prediction. With 39039250 (Nat Med, clinical-grade computational pathology and rare-cancer detection), the imaging branch of the foundation-model thread is publishing in the highest-tier venues a full year before the single-cell branch attracts the same attention.
4. Refutations¶
(a) Landing on a tracked thread¶
39465898 (NEJM): large language models and the degradation of the
medical record. The harm is located in the clinical artefact rather than in
model accuracy — a different axis from every benchmark result, and the axis that
survives a model getting better. Filed as a refutation on foundation-models
because the thread's claim is about transfer into clinical use, and this says the
transfer damages the substrate.
(b) Method credibility¶
Neither of this month's two AI-methods papers is a credibility result, and both are here as notes rather than as ledger entries.
39333790 (Nat Comput Sci) is STAMP, which splits genetic-perturbation prediction into three ordered subtasks — which genes change, in which direction, by how much — and reports substantial improvement over existing approaches on single perturbations, multiple perturbations and transfer across cell lines, along with the ability to recover regulatory genes and interaction types from small samples. An improvement with no adverse finding attached.
39443777 (Nature) is SynthID-Text, a production text watermarking scheme that modifies only the sampling procedure, detects without running the model, and was assessed over nearly 20 million live Gemini responses with no measured loss of quality. Provenance infrastructure, which is what a field builds when it expects not to be able to tell machine output from human output by inspection — but a tool for establishing where text came from says nothing about whether published methods are reliable.
Corrected 2026-08-27
Both entries overreached, in opposite directions. STAMP was carded refutes
as an audit and read as evidence that "the task is already being rebuilt
around its own weaknesses — the benchmark's authors had the objection before
the null result was published", tied to a June 2025 result about simple
controls. The word "benchmarking" in the title is the only thing in the paper
that attaches to; the narrative was supplied by a reader who knew what 2025
would say. And the watermarking paper was the single supports card among
151 on method-credibility, carrying that thread's claim that published
methods and reagents are reliable enough to build on without re-checking —
which it neither states nor contradicts. Its stance alone moved the thread's
supports count off zero. Both cards are now neutral, and STAMP has lost
method-credibility.
5. Open items¶
Nothing carried forward.
6. Meetings¶
No abstract book published this month. The abstract layer runs 2023-02 to 2026-08 and is seasonal by nature — meetings are annual, so most months are empty here and that is the expected shape, not a gap in the harvest.
7. Regulatory¶
| Event | Count |
|---|---|
| trial-first-posted | 592 (of which 512 genuinely oncology) |
| fda-approval | 4 |
4 FDA actions — dated from each notice's own prose, not the table's posting date, which disagrees on 7 of 190 rows and once by enough to move an approval into the wrong month:
- 2024-10-03 — FDA approves neoadjuvant/adjuvant nivolumab for resectable non-small cell lung cancer
- 2024-10-10 — FDA approves inavolisib with palbociclib and fulvestrant for endocrine-resistant, PIK3CA-mutated, HR-positive, HER2-negative, advanced breast cancer
- 2024-10-18 — FDA approves zolbetuximab-clzb with chemotherapy for gastric or gastroesophageal junction adenocarcinoma
- 2024-10-29 — FDA grants accelerated approval to asciminib for newly diagnosed chronic myeloid leukemia
8. Trade press¶
44 on-topic of 362 items across 3 feeds (n-stat 342, n-aacr-blog 16, n-acir 4). Headlines are selected by subject keyword, not recency — most of what these feeds carry is health policy and industry finance the corpus does not read.
- 2024-10-31 — STAT+: These 10 scientists are leading a new generation of gene editors developing CRISPR medicines —
n-stat - 2024-10-30 — Editors’ Picks, October 2024: Head-to-head CAR T, a Colorectal Cancer Blood Test, and More —
n-aacr-blog - 2024-10-29 — STAT+: Amid divorce from Dana-Farber, Mass General Brigham touts its own cancer institute —
n-stat - 2024-10-29 — STAT+: Warren wants close FTC scrutiny of drug distributors buying cancer medical groups —
n-stat - 2024-10-29 — STAT+: Dreams of cancer vaccines are becoming more real. Here are 9 scientists making it happen —
n-stat - 2024-10-28 — STAT+: Novartis, Seagen veteran David Epstein is back, leading a new cancer startup —
n-stat - 2024-10-28 — STAT+: Feud between Prime and Tessera, two gene-editing companies, spills out into public —
n-stat - 2024-10-28 — STAT+: A drug may cause problems with chemotherapies in cancer patients with obesity. Many doctors are in the dark —
n-stat - 2024-10-27 — Opinion: Blood-based biomarkers for Alzheimer’s present complex challenges to health care system —
n-stat - 2024-10-25 — The AACR Industry Roundtable 2024: Detecting and Treating Cancer Early —
n-aacr-blog - 2024-10-24 — Pfizer’s activist troubles, gene therapy questions, RFK Jr.’s MAHA ambitions —
n-stat - 2024-10-24 — STAT+: Gilead needs a win with Arcellx CAR-T data. Here’s how to know if it scores —
n-stat - … and 32 more on-topic items.