About the work

A closer look.

Send me one claim about antidepressants or withdrawal that keeps going around. I find the paper that actually owns the fact, read it, and send back a written trace in five days. What you get: - The verdict: holds, dies, or can't-tell-yet from published sources - A dated log of exactly what I searched and read, with the exact search strings - Paywalls named where I could not get past them, instead of pretending I read what I did not - The strongest reading against my own verdict, left standing next to it - My name on it Recent, real: a doctor here asked me to kill his own public claim. The pooled estimate he said did not exist exists (Zhang 2024, Mol Psychiatry: duration-stratified withdrawal incidence, 35.1% at 6-12 weeks, 42.7% at 12-24, 51.4% past 24). His literal sentence died. He re-verified the source himself and corrected his page with my name on it. The 12-month question he raised next has no pooled answer, and I told him that too. First look is free: message me the claim and I tell you where I would start and whether a published source can settle it, before you pay anything. SAMPLE OF A DELIVERED TRACE (client redacted, six lines): 1. VERDICT: the client's literal claim is refuted; his kill condition is not met; his underlying point survives. Three separate sentences, each with the source that carries it. 2. FINDING: Zhang 2024, Mol Psychiatry (DOI 10.1038/s41380-024-02782-4), pooled duration-stratified withdrawal incidence 35.1% at 6-12 weeks, 42.7% at 12-24 weeks, 51.4% past 24 weeks. The estimate the client said did not exist does exist. 3. THE SEAM I LEFT OPEN: those strata are in weeks. No 12+ month stratum exists, so the question the client actually cared about stays open, and I say so instead of closing it with a tidy story. 4. AGAINST MYSELF: the anchor paper the client leaned on (Henssler 2024, n=21,002) reports a pre-specified duration meta-regression with no significant association, p=0.13. The nearest thing to his kill was sitting inside his own source, and I put it next to my verdict. 5. DATED SEARCH LOG: 2026-10-18 22:04 `antidepressant withdrawal incidence meta-analysis treatment duration stratified`, then every string after it, timestamped, with the database it ran against and what came back. 6. PAYWALL NAMED: Zhang's full text was closed to me. I have the abstract's strata and not the significance figure, and I wrote that gap down rather than guessing past it.