About this publication

What we publish

Two kinds of work. Analysis is a daily stream: each piece takes one new item of primary research, a paper, preprint, grant award or registered trial, and works out what it changes. The standing explainers, beginning with what is an organoid?, are meant to stay correct for a long time and are revised when the field moves. An analysis is a dated reading of one result at one moment, and we do not quietly rewrite it later.

Where the material comes from

We maintain a research library that ingests continuously from arXiv and bioRxiv, the NIH RePORTER grants database, ClinicalTrials.gov, NSF award records, journal feeds and vendor technical documentation. Each incoming item is checked for duplicates, summarized and scored for relevance before it can be written about. Every published analysis names the specific record it came from and links to the primary source. A ledger records every source consumed, so a piece that has been covered cannot resurface.

How each piece is produced

Analyses are researched and drafted by an automated research system, a coordinated group of large language models led by a reasoning model working to a fixed editorial brief, and are validated against publication rules before anything goes live. We think readers are entitled to know that, so we are saying it plainly here rather than burying it.

What that process is required to do: retrieve and read the primary source rather than working from a title or abstract; state explicitly, in the article, when a full text could not be obtained, and bound the claims accordingly; separate what a study demonstrates from what it asserts; identify the load-bearing assumption and stress test it; and give the genuine objection alongside the genuine opportunity. Pieces that fail the publication checks are rejected rather than patched into shape.

What it is not: peer review. An analysis is an interpretation of someone else's research by a reader who did not run the experiment. Where our reading is uncertain, the article says so.

Sourcing and citation standards

Every factual claim traces to a named source. References carry a title, venue, date, a resolvable link and the date we accessed it. We do not cite a study we have not opened, we do not reproduce a number we cannot attribute, and we do not use a secondary write up in place of the underlying work.

We report what a model was validated against, not what it was advertised to do. Sample sizes, donor counts and passage numbers matter for reproducibility, so we quote them where the source does and note their absence where it does not. Preclinical results are reported as preclinical: an effect in a dish is not an effect in a patient, and we do not blur that line to make a headline.

What makes this subject hard to report

The recurring failure in this field is generalization: a result obtained in one lab's organoids, from one donor line, at one passage, presented as a property of the tissue type. We read for that specifically, and an analysis here will usually tell you how far the claim travels before it needs re-testing.

Corrections

If a piece misstates a result, misattributes work or misreads a source, we want to know and we will correct it. Corrections are made on the page itself and noted there rather than applied silently, because an analysis is dated evidence of what we thought at the time. Send corrections, and any question about how a piece was sourced, to mg@brandadvertisers.com.

Publisher

Organoid Biofoundry is published by Brand Advertisers LLC. It is one of four related titles covering this field: Organoid Intellect for the concepts and theory of biological computing, Organoid Grid for platforms and ethics, Organoid Biofoundry for organ models and drug discovery, and Organoid Array for electrode hardware and specifications.