GreyBrain Pulse · Published 13 Sept 2026
Forty abstracts into one review matrix
The scientist · What’s advancing · F1
A week of extraction work in one pass, with "not stated" marked honestly instead of filled in.
Why this matters for your role
The scientist’s skill: distinguish a promising signal from a practice-changing result.
Your worked example
A fictional research team notices that students who voluntarily attend extra teaching sessions score higher on a later quiz. Attendance was not randomly assigned; prior attainment and motivation were not measured.
Put it into practice
Separate observation from causal hypothesis. List confounders and compare two study designs that could test the question. Do not invent results.
Clinician prompt
Forty abstracts into one review matrix
A week of extraction work in one pass, with "not stated" marked honestly instead of filled in.
You are helping me build the evidence table for a narrative review. This is a practice exercise; I will verify every row myself. Below are abstracts I have collected: [PASTE 10–40 ABSTRACTS, SEPARATED BY A BLANK LINE] Build one markdown table with a row per study and these columns: first author and year, country, study design, sample size, population, intervention or exposure, primary outcome and effect direction, and the single most important limitation. Rules: - Take every value from the abstract text only. Where the abstract does not state something, write "not stated" — never infer it. - After the table, group the studies by design and tell me where the evidence is thin: which designs dominate, which populations are missing, and which outcome is reported too inconsistently to pool. - Finally, list the three studies whose full text I must read before writing a word, and say why each one. End with a short section titled "Your check" listing what I must verify myself before using any of this.
Synthetic practice example. Run it as-is and check the output against the supplied facts.
Your check
Open the three flagged papers yourself and re-read two rows at random against the original abstract. A single silently invented sample size is enough to discredit the whole table.
Abstracts only, from published work. The model cannot assess study quality and has not checked whether any of these were retracted.
Sources inform this fieldnote; their publishers do not endorse GreyBrain. This exercise does not confer clinical competence or a certificate.