Repository navigation
Write the retatrutide page, and stop reading screening counts as enrolment - #82
Merged
Merged
Conversation
…lment First of the seven records that had a full evidence ledger and no written page. It was noindex and is now indexable, at 6,759 words with four tables and 41 cited quotations. Demand, measured 2026-09-30: the head term is 204,176 clickstream searches at KD 57, but its entire first page is one domain, stonevillenc.org, a municipal .org hosting generated peptide pages. That is not competition, it is an unstable SERP. The winnable demand is the long tail: "retatrutide dosage" at 5,901 clicks and KD 0, "retatrutide side effects" at 7,173 and KD 10 on a SERP that returns four organic results, none from a health publisher. Three competitors fetched and scored. None of them cites a single trial registration, so the page leads with a table of every registered trial by the ClinicalTrials.gov number its own published paper declares, separating trials that have reported from the phase 3 programme that has only a design paper. It is also the only page giving the phase 2 dose-response at both 24 and 48 weeks for all four dose groups rather than quoting the best number, reporting that the starting dose and not the target changed the side-effect rate, and carrying the dose-dependent heart-rate rise that peaked at 24 weeks. Two stored sample sizes were wrong, and the bug is in the drafter, not the records. A CONSORT abstract opens with the screening count, and both patterns that read a sample size matched it first: "930 participants were screened and 537 were randomly assigned" stored 930, and "210 people were screened, of whom 72 were enrolled" stored 210. Assignment phrasing is now tried before anything else, and a candidate whose next words say it was screened is rejected. Four regression tests, 33 in total, and both rows corrected. Every other stored sample size is being re-checked separately. The duplicate-H2 rule added to check_links.py earlier today caught its first real defect here: the generated escalation block rendered "No study in this record's ledger reports dose-escalation schedules" directly beneath a written section quoting those schedules from the trial abstract. Contradiction, duplicate heading and duplicate element id, all three from one collision. The generated block is suppressed. Title and description rewritten: the old title named neither query the demand actually sits on. Registry and label documents could not be used as sources. This sandbox reaches Europe PMC but not clinicaltrials.gov, dailymed.nlm.nih.gov or api.fda.gov, so every NCT number is quoted from the paper that declares it and the registry links are pointers. The page and the audit both say so. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
First of the seven records that had a full evidence ledger and no written page. It was
noindexand is now indexable: 6,759 words, 4 tables, 41 cited quotations, heuristic score 93/100 with trust at 30/30.The SERP is not what the difficulty score says
Measured 2026-09-30.
stonevillenc.org, a municipal .org hosting generated peptide pagesThe head term's KD of 57 describes competition that is not real. The winnable demand is the long tail, and no health publisher holds the side-effects query at all.
Information gain
Three competitors fetched and scored. None cites a single trial registration.
So the page leads with a table of every registered trial by the ClinicalTrials.gov number its own published paper declares, separating trials that have reported from the phase 3 programme that has only a design paper. It is also the only page that gives the phase 2 dose-response at both 24 and 48 weeks for all four dose groups rather than quoting the best number, that reports the starting dose rather than the target as what changed the side-effect rate, and that carries the dose-dependent heart-rate rise peaking at 24 weeks.
A drafter bug, found while writing
Two stored sample sizes were screening counts. A CONSORT abstract opens with how many were screened, and both patterns that read a sample size matched it first.
pmid-42250575pmid-36354040Assignment phrasing is now tried before anything else, and a candidate whose following words say it was screened is rejected. Four regression tests added, 33 total. Every other stored sample size across all 223 rows is being re-checked separately; any further corrections will follow in their own PR.
The new CI rule earned its keep
The duplicate-H2 check added earlier today caught its first real defect here. The generated escalation block rendered "No study in this record's ledger reports dose-escalation schedules" directly beneath a written section quoting those schedules from the trial abstract: a contradiction, a duplicate heading and a duplicate element id, all from one collision. The generated block is suppressed.
Honest limit
This sandbox reaches Europe PMC but not
clinicaltrials.gov,dailymed.nlm.nih.govorapi.fda.gov. Every NCT number here is quoted from the published paper that declares it, and the registry links are pointers rather than sources. The page says so in its own words, and the audit records it as an open action. Using registry and label documents as sources in their own right needs an environment with access to them.Gates
Content validation, registry check, 33 drafter tests, loop gate at 46/46, all five link rules. Intent coverage 15/17 queries, 98% of measured demand on a rendered answer.
🤖 Generated with Claude Code