How to read a peptide forum thread: four questions, and three real vendor claims checked against 62,079 certificates

One-line glossary: a “COA” is a certificate of analysis, a third-party lab’s report on one vial. The two numbers that matter on it are purity (what fraction of the material is the labeled peptide) and tested quantity (how many mg were actually in the vial against how many the label claimed).

The seven-post thread that makes the whole case

On April 23 2025, a user posting as meathunter opened a thread on GLP-1 Forum titled “HZSB ?”. The entire thread is seven posts long. Here it is, in order, with nothing removed.

meathunter: Is HZSB a safe China source

scarywood75: Yes, best vendor out there 100% approved high quality, ships in an hour, top nutch German quality peptides

meathunter: Beginner here , as you can tell ,

meathunter: Another question. If the COA says it tested 96 percent , that doesn’t sound like a disaster . Not that far from 99 percent . Is 96 percent ok ?

Calm Logic: If you know her personality, @scarywood75 is joking or being ironic. They are another one of the many alphabet-soup vendors that no one has heard of. There are tons more listed at Made-in-China.com and Alibaba. I would not use any of them. They are cheaper (if the product arrives eventually in a month or so) but not vetted (so you don’t know what you are really getting).

meathunter: thank you

Ted R: Top NUTCH!!!

Source: glp1forum.com/threads/hzsb.3363/. 756 views. Quotes are verbatim apart from curly quotes normalised to ASCII and trailing emoji removed; no words changed.

Read what actually happened there. A person with no experience asked a direct question about whether a vendor was safe. He got back a five-clause answer containing zero information: “best vendor out there”, “100% approved”, “high quality”, “ships in an hour”, “German quality peptides”. Every clause is an assertion. None is attached to a mechanism, a test, a purchase, or a date. One clause, “German quality”, directly contradicts the question that was asked, which specified a China source.

And meathunter believed it. His next post is “Beginner here, as you can tell,” followed immediately by a genuine technical follow-up about COA purity thresholds. He was not being sardonic back. He took the answer as real and moved on to his next question.

It took a fifth post from a different user to tell him the second post had been a joke.

If you want a single argument for why forum reading is a skill and not a passive activity, that thread is it. The reply that was worthless got a reaction. The reply that was useful arrived two posts too late to stop the beginner from processing the joke as advice. And the thread is not unusual. It is exactly what most vendor-quality threads look like.

The four questions

There is a lot of generic media-literacy advice available and almost none of it survives contact with a peptide forum, because the usual heuristics (does this person seem knowledgeable, do others agree, is the tone confident) all point the wrong way here. Confidence is free. Agreement is often the same three accounts. Seeming knowledgeable is a writing skill.

Four questions do work, and they work because they ask about the structure of the claim rather than the person making it.

1. Is there a mechanism, or only a conclusion?

A mechanism is the “because”. “Their reta is good” is a conclusion. “Their reta came back 99.5% on a Janoshik test I paid for in March, and the vial weighed 10.4mg against a 10mg label” is a conclusion with a mechanism attached. You do not have to trust the second person more as a person. You have to trust them less, because the claim now carries its own audit trail.

Conclusion-only posts are not always wrong. They are unusable, which is different. There is no operation you can perform on them.

2. Did this person do it, or hear it?

Grey-market forums run on relayed claims. “I have heard good things”, “people say”, “a friend tested it”. These are not lies. They are one-hop compressions, and the compression is where the information dies, because the hop drops the thing you needed: which batch, which lab, which date, which dose.

Watch for the switch mid-post. A poster will describe a first-hand order in the first two sentences and then, without any signal, extend it into a general claim about the vendor. The first half is evidence about one vial. The second half is opinion about a company.

3. Could this claim turn out to be false?

“This vendor is legit” cannot be false, because nobody has said what would count. “This vendor’s 20mg reta vials contain 20mg” can be false and you can find out.

This is the question that most cleanly sorts a peptide forum, because the market’s marketing language is built almost entirely out of unfalsifiable predicates: premium, pharmaceutical grade, top tier, trusted, verified, lab tested. “Lab tested” is the sharpest example. It asserts that an event occurred and says nothing about the result of that event. A vial that assayed at 40% purity was lab tested.

4. Can it be checked against a certificate?

This is the one that pays. If a claim is about purity, dose, identity, or endotoxin, then somewhere there either is or is not a piece of paper. Ask which. A vendor that will not point at a specific certificate for the specific batch under discussion has told you something even when it refuses to answer.

Our corpus holds 62,079 published certificates, of which 61,295 carry a verifiable link back to the issuing lab or the original posting, 50,617 carry a purity number, and 16,351 carry a dose-deviation number. That is the surface a claim can be checked against. If you want to run these checks yourself, the compound hubs are the entry point: retatrutide, tirzepatide, BPC-157.

Now let us actually use the four questions. Three real claims, from the forums our pipeline already ingests, checked.

Check one: “best vendor out there”

The claim. scarywood75, April 23 2025: “Yes, best vendor out there 100% approved high quality, ships in an hour, top nutch German quality peptides.”

Question 1, mechanism. None. Five assertions, no because.

Question 2, first-hand or heard. Not stated either way. The post does not claim a purchase, a test, or a delivery.

Question 3, falsifiable. “Ships in an hour” is falsifiable and would be a real contribution if it were sourced. Nothing else in the post is. “100% approved” does not name an approving body. “German quality” is an aesthetic.

Question 4, checkable against a certificate. Yes, and this is where it ends.

We searched the entire certificate corpus for HZSB: the manufacturer table (name, aliases, short codes) and every text field on every certificate record. Zero rows. Out of 4,054 manufacturer entities and 62,752 certificate records covering test dates from November 2017 to July 2026, the vendor described as “best vendor out there” appears nowhere.

We also searched all 85,721 forum posts our pipeline has ingested across two forums. Exactly one post body anywhere contains the string HZSB, and it is meathunter asking the question.

So the verdict is not “this vendor is bad”. We have no evidence that this vendor is bad. The verdict is that a superlative was applied to an entity with no public testing record at all, and a beginner acted on it. Calm Logic got there without a database, using nothing but a structural observation: alphabet-soup vendor, nobody has heard of it, not vetted. That reasoning was available to anyone in the thread. It arrived fifth.

Check two: a vendor grading its own homework

The claim. On July 20 2026, an account posting as Angelia28 in an HPT vendor thread posted this, described as a customer’s private test result for the vendor’s own Reta 20mg:

Customer private test result of our Reta 20mg. 99.9% purity. Accurate labeled dosage. Zero endotoxin, fully safe. Verified by BT Labs via the scammers manipulating test results for profit at Finnrisk. Third-party lab results speak louder than words. DM for orders.

Transcription note: the original uses emoji as clause separators and carries no sentence punctuation, so the emoji have been replaced with full stops. No words are changed, added or reordered. The strange clause in the middle is not the advertiser’s writing; see the section on word filters below.

Questions 1 through 3. There is a partial mechanism: a lab is named and a specific SKU is named. That is more than most vendor posts carry. All three headline claims are falsifiable. This is a checkable ad, which is unusual and to the vendor’s credit as a matter of form.

Question 4. We hold 103 certificate rows for this manufacturer, 63 distinct assays after collapsing duplicates that reached us through more than one source. Thirteen of those distinct assays are retatrutide. Here is what they say.

Claim A, “99.9% purity”. Across the 13 retatrutide assays, purity runs from 99.21% to 99.9%, with a median of 99.8%. Six of the 13 reach 99.9% or above. So the headline number is real, and it is the vendor’s best case rather than its typical case. Nothing here is scandalous.

But hold that number against the market. Across all 9,830 retatrutide assays in our corpus that carry a purity figure, the median is 99.83% and 34.69% of them reach 99.9% or better. On this compound, at this point in the market, 99.9% purity is not a differentiator. It is roughly the modal outcome. The ad leads with the claim that is easiest to satisfy.

Claim B, “accurate labeled dosage”. This is the one that matters, and it is the one the certificates contradict. Restricting to the exact SKU advertised, 20mg retatrutide, we hold four distinct assays:

Test date	Lab	Purity	Label	Tested	Deviation
2026-01-29	Janoshik	99.548%	20mg	20.10mg	+0.5%
2026-03-18	Finnrick (LabE)	99.9%	20mg	18.7mg	-6.5%
2026-04-28	Finnrick (LabE)	99.9%	20mg	21.4mg	+7.0%
2026-06-24	Finnrick (LabE)	99.9%	20mg	19mg	-5.0%

“Finnrick (LabE)” means the certificate was published through Finnrick, which does not name the performing lab and instead labels it LabA through LabG. We carry that label through rather than guessing which lab it was.

A 20mg vial from this vendor has come back anywhere between 18.7mg and 21.4mg. That is a 2.7mg spread on a 20mg label, and it is not a rounding artifact: three of the four are off by 5% or more, in both directions. Reconstitute either vial on the assumption that it holds 20mg, draw what you believe is 2mg, and the March vial gives you about 1.87mg while the April vial gives you about 2.14mg. Nothing on the outside of either vial tells you which one you have.

Widening to all 13 retatrutide assays makes it worse, not better. The 12 that carry a deviation figure run from -6.5% to +22.6%, median +5.5%, with three of the 12 off by more than 10%.

“Accurate labeled dosage” is a claim about the exact quantity we just measured, and the measurements are in the vendor’s own certificates.

Claim C, “zero endotoxin, fully safe”. Not one of the 13 retatrutide assays carries an endotoxin result. Not a passing one, not a failing one. The field is empty on every single one.

Across the vendor’s entire 103-row certificate footprint, exactly one record carries an actual endotoxin determination, and it is a tesamorelin vial tested on June 1 2026 reading “below LOQ”. Three more read “n/a”. The other 99 are blank.

This is a general condition of the market rather than a peculiarity of one vendor. Of the 62,079 published certificates we hold, 8,897 have anything at all in the endotoxin field, and 4,241 of those say “n/a”. That leaves 4,656, about 7.5% of the corpus, with an actual endotoxin determination on file. When someone tells you a vial has zero endotoxin, the first question is not whether the number is good. It is whether a number exists.

And the thing the ad does not mention. The same vendor has a certificate, dated April 28 2026, carrying the same Finnrick lab label (LabE) as three of the four 20mg retatrutide results above, for a 20mg tirzepatide vial that assayed at 0.0% purity and 0mg of tested content. Deviation from label: -100%. It is publicly verifiable at finnrick.com/testing-certificate/tirzepatide/homopeptide/20260428-e-xcchga6 and it is in our corpus at /research/coa/23805.

April 28 2026 is also the date of the +7.0% retatrutide result in the table above. Same vendor, same publisher, same day, opposite story.

We are not going to tell you what that certificate means. A 0mg reading can be a genuinely empty vial, a mislabeled sample, a submission handling error, or a lab error, and we do not have the chain of custody. What we will say is that it sits in the same file as the results the vendor is happy to cite, and it did not appear in the ad. Selective certificate citation is the single most common form of dishonesty in this market, and it is completely invisible unless you look at the vendor’s whole file instead of the one page they hand you.

Structural note. The thread that ad sits in has 98 posts and 25,000 views. 85 of the 98 posts are by Angelia28, the seller. Seven other accounts posted a combined 13 times. That is not a discussion with a promotional post in it. It is a broadcast channel with a comment section, and the reply-count number at the top of the page will not tell you that.

Check three: what a good post looks like, and where even a good post stops

The claim. On May 14 2026, a user posting as DjJoshua wrote a vendor review of DP (Zhengzhou Depu Technology). Excerpted:

Key Results Breakdown. Chromatographic Purity: >99.9% top-tier/pharmaceutical-like. Actual Content: 20.4 mg (labeled 20 mg) 101.9% of label claim, i feel it’s accurate for mixing and dosing. Identity: Mass Spectral Match ID Confirmed, correct molecule verified. Endotoxins: <0.5 EU (below limit of quantification). Excellent. Very low endotoxin levels. I try to target <0.05-10 EU/mg or similar, being below detection is ideal. […] Third-party tested by Krause Analytical (a lab used by several peptide vendors). Recent testing (May 2026). Full transparency with HPLC-UV/MS methods.

Run the four questions and this post scores about as well as a forum post can.

Mechanism: present throughout. Every assertion carries the measurement it rests on.

First-hand: clearly marked. “I feel it’s accurate for mixing and dosing”, “I try to target”, “this kit”. He is describing a vial he had, and he says so.

Falsifiable: completely. He names the lab, the month, the label weight, the measured weight, the ratio, the method, and his own acceptance threshold before giving the verdict. Someone who disagreed would know exactly which number to attack.

Checkable: in principle yes, and here the check comes back mixed, which is the interesting part.

Our corpus holds three certificates from Krause Analytical in total, out of 62,079. All three belong to a different manufacturer. The specific certificate DjJoshua describes is not in our corpus at all. He gave every detail needed to verify it except a link, and the certificate does not appear to have been posted anywhere our pipeline reaches.

What we can check is the vendor. We hold 11 distinct assays for DP, three of them retatrutide:

Test date	Lab	Purity	Label	Tested	Deviation
2025-10-29	Janoshik	99.271%	(not printed)	9.39mg	not computable
2025-10-29	Janoshik	99.604%	(not printed)	20.01mg	not computable
2026-06-30	Janoshik	99.771%	10mg	11.39mg	+13.9%

Nothing here contradicts him. But nothing here reaches 99.9% either. DP’s highest purity on any assay we hold, any compound, is 99.864% (a TB-500 from October 2025). And the one DP retatrutide assay in our corpus that has a printed label to compare against came back 13.9% over label, seven weeks after he called the vendor’s dose accuracy good.

So the honest reading of the best post in this survey is: it is well built, it is probably true about the vial he had, and it is a sample of one that got extended into “one of the best-looking COAs so far for Retatrutide in the research peptide space”. That last sentence is the failure mode that survives all four questions. A claim can be sourced, first-hand, falsifiable, and still be a single vial talking.

It also inherits the market’s blind spot. He treats >99.9% purity as evidence of a top-tier operation. On retatrutide, a third of everything tested clears that bar. The number he is impressed by is the number that no longer separates vendors.

Answering meathunter’s actual question, fifteen months late

Buried in that first thread is a real question that never got an answer:

If the COA says it tested 96 percent, that doesn’t sound like a disaster. Not that far from 99 percent. Is 96 percent ok?

It is a good question, it was asked in the right place, and it got a joke instead of a number. It takes one query to answer.

Across the 50,617 published certificates in our corpus that carry a purity number, the median is 99.73% and the mean is 98.87%. 1,030 certificates, 2.03% of them, come in below 96%. A further 3,040 sit between 96% and 99%, so 8.04% of everything measured falls below 99%.

Read that carefully, because it says something slightly different from what you would guess. His instinct that 96 and 99 are near each other is half right: both are below the median, and neither is a normal result. A 96% vial sits in the bottom 2% of everything tested. A 99% vial sits in the bottom 8%. What is normal on this market is 99.7%, and the practical answer to “is 96 percent ok” is that it is unusual enough to want a second certificate on the same batch before you decide anything.

The honest caveat, and it is a big one: purity is one axis. A 96% vial that contains the mg it claims may serve you better than a 99.8% vial that contains 78% of the label. That is why the dose column exists in every table above, and why the checks in this piece spent more time on tested quantity than on purity.

Three things the four questions cannot fix

The thread you are reading may be one person. 85 of 98 posts in that HPT thread came from the seller. There is no indicator on the page for this. Before you read a vendor thread as a body of opinion, scroll it and count how many distinct names appear.

The words may not be the author’s. GLP-1 Forum appears to run a word filter that substitutes an editorial phrase for a testing company’s name. The string “the scammers manipulating test results for profit at” appears 66 times across 49 posts written by 37 different accounts, all on that one forum, from June 2 2025 to July 20 2026. Thirty-seven separate people did not independently type the same eight-word phrase. It is why the HPT advertisement quoted above contains an accusation against a testing lab in the middle of its own sales copy, which no advertiser would write on purpose.

To be precise about what we can and cannot show: we have the strings, not the forum’s filter configuration. The word that follows the phrase is not stable either (two spellings account for all 66 hits, 54 and 12). We infer a substitution from the distribution and from the surrounding sentences, which read as ordinary references to a lab (“I sent one to”, “here is the email”). We cannot prove the mechanism from outside.

That is a small thing with a large consequence: you cannot always tell, from inside a forum, which words in a post were chosen by its author. Anything load-bearing needs to be checked outside the forum.

Sarcasm does not survive the format. The entire HZSB thread turns on a joke that one reader could not detect and that a regular had to translate. Text on a screen from a stranger with no shared context is a bad medium for irony, and grey-market vendor threads are full of it, because regulars get tired of answering the same question. If a reply reads as too enthusiastic and contains no specifics, the possibility that it is a joke ranks alongside the possibility that it is a shill.

The checklist

Print this or do not, but run it.

1. Count the names. How many distinct people are in this thread? If the answer is one plus a handful of replies, it is an advertisement.
2. Find the because. If a post asserts quality without saying why, it is a vote, not evidence. Votes do not aggregate into facts.
3. Find the “I”. Did this person do the thing they are describing? Unmarked hearsay is the default state of a forum.
4. Ask what would prove it wrong. If nothing would, stop reading and move on.
5. Demand the batch. A vendor claim about “our product” is worth nothing. A claim about a batch with a certificate is worth checking. Ask for the batch and the certificate link, in that order.
6. Check the whole file, not the page they handed you. One good certificate is a screenshot. The distribution of a vendor’s certificates is a track record. A vendor with one excellent COA and a 0mg COA on the same day is not a vendor with an excellent COA.
7. Check the field exists before you check the number. Only about 7.5% of certificates in our corpus carry a real endotoxin determination. “Zero endotoxin” usually means untested.
8. Weight the boring post above the confident one. In everything we read for this piece, the useful information had hedges in it and the useless information did not.

How we ran these checks

Every number in this piece comes from a query against data/coa.db (62,752 certificate records, 62,079 published, test dates 2017-11-13 to 2026-07-27) and data/forums.db (85,721 posts across GLP-1 Forum and MESO-Rx, 2010-09-28 to 2026-07-31), both read-only, on 2026-08-01. Certificates that reached us through more than one source are collapsed to one assay per distinct combination of compound, test date, purity and tested quantity, which is why the row counts below are smaller than the raw record counts.

Corpus-level figures:

-- corpus size and coverage
SELECT COUNT(*), SUM(published = 1) FROM coa_records;
SELECT MIN(test_date), MAX(test_date) FROM coa_records
 WHERE published = 1 AND test_date LIKE '20%';
SELECT COUNT(DISTINCT manufacturer_id) FROM coa_records
 WHERE published = 1 AND manufacturer_id IS NOT NULL;
SELECT SUM(verify_url IS NOT NULL AND TRIM(verify_url) <> ''),
       SUM(purity_pct IS NOT NULL),
       SUM(quantity_deviation_pct IS NOT NULL),
       COUNT(*)
  FROM coa_records WHERE published = 1;

-- purity distribution (answers the "is 96 percent ok" question)
SELECT COUNT(*),
       SUM(purity_pct < 96),
       ROUND(100.0 * SUM(purity_pct < 96) / COUNT(*), 2),
       SUM(purity_pct >= 96 AND purity_pct < 99),
       SUM(purity_pct < 99),
       ROUND(100.0 * SUM(purity_pct < 99) / COUNT(*), 2),
       ROUND(AVG(purity_pct), 2)
  FROM coa_records WHERE published = 1 AND purity_pct IS NOT NULL;
-- median computed in Python over the same 50,617 values: 99.73

-- endotoxin coverage
SELECT SUM(endotoxins_status IS NOT NULL AND TRIM(endotoxins_status) <> ''),
       SUM(LOWER(TRIM(IFNULL(endotoxins_status,''))) = 'n/a'),
       COUNT(*)
  FROM coa_records WHERE published = 1;

-- retatrutide market context
SELECT COUNT(*),
       SUM(purity_pct >= 99.9),
       ROUND(100.0 * SUM(purity_pct >= 99.9) / COUNT(*), 2),
       ROUND(AVG(purity_pct), 3)
  FROM coa_records
 WHERE published = 1 AND LOWER(peptide) = 'retatrutide' AND purity_pct IS NOT NULL;
-- dose spread on the same compound (4,148 rows carry a deviation figure;
-- median +2.0%, share with |deviation| > 10% computed in Python: 35.37%,
-- of which 12.44% light and 22.93% heavy)
SELECT quantity_deviation_pct FROM coa_records
 WHERE published = 1 AND LOWER(peptide) = 'retatrutide'
   AND quantity_deviation_pct IS NOT NULL;

Check one, HZSB:

SELECT COUNT(*) FROM coa_manufacturers
 WHERE UPPER(name) LIKE '%HZSB%'
    OR UPPER(IFNULL(aliases_json,'')) LIKE '%HZSB%'
    OR UPPER(IFNULL(short_codes_json,'')) LIKE '%HZSB%';          -- 0

SELECT COUNT(*) FROM coa_records
 WHERE UPPER(IFNULL(manufacturer_raw,'') || '|' || IFNULL(client_raw,'')
           || '|' || IFNULL(sample,'') || '|' || IFNULL(test_id,'')) LIKE '%HZSB%';  -- 0

Check two, the advertised vendor (coa_manufacturers.id = 406, Homopeptide, short code HPT):

-- distinct retatrutide assays
SELECT test_date,
       GROUP_CONCAT(DISTINCT IFNULL(lab_name,'(none)')),
       purity_pct,
       REPLACE(LOWER(IFNULL(label_qty,'')),' ','')  AS label_q,
       REPLACE(LOWER(IFNULL(tested_qty,'')),' ','') AS tested_q,
       ROUND(MAX(quantity_deviation_pct), 2),
       COUNT(*) AS rows_merged
  FROM coa_records
 WHERE manufacturer_id = 406 AND published = 1 AND LOWER(peptide) = 'retatrutide'
 GROUP BY test_date, purity_pct, label_q, tested_q
 ORDER BY test_date;                                    -- 13 distinct assays

-- the advertised SKU only
SELECT test_date, lab_name, purity_pct, label_qty, tested_qty,
       quantity_deviation_pct, verify_url
  FROM coa_records
 WHERE manufacturer_id = 406 AND published = 1 AND LOWER(peptide) = 'retatrutide'
   AND REPLACE(LOWER(IFNULL(label_qty,'')),' ','') = '20mg'
 ORDER BY test_date;                                    -- 4 distinct assays

-- endotoxin coverage for this vendor
SELECT IFNULL(endotoxins_status,'(null)'), COUNT(*)
  FROM coa_records WHERE manufacturer_id = 406 AND published = 1
 GROUP BY 1;                       -- '' 21, NULL 78, 'n/a' 3, 'below LOQ' 1

-- the 0mg tirzepatide certificate
SELECT id, test_date, lab_name, purity_pct, label_qty, tested_qty,
       quantity_deviation_pct, verify_url
  FROM coa_records WHERE id = 23805;

Check three, DP (coa_manufacturers.id = 525, Zhengzhou Depu Technology):

SELECT peptide, test_date, lab_name, purity_pct,
       REPLACE(LOWER(IFNULL(tested_qty,'')),' ','') AS tested_q, COUNT(*)
  FROM coa_records WHERE manufacturer_id = 525 AND published = 1
 GROUP BY LOWER(peptide), test_date, purity_pct, tested_q
 ORDER BY test_date;                                    -- 11 distinct assays

SELECT COUNT(*) FROM coa_records WHERE lab_name LIKE '%Krause%';   -- 3, none from DP

Forum structure:

-- who is actually posting in the vendor thread
SELECT author_handle, COUNT(*) FROM forum_posts
 WHERE thread_id = (SELECT thread_id FROM forum_posts WHERE id = 2755117)
 GROUP BY 1 ORDER BY 2 DESC;              -- Angelia28 85 of 98

-- the word filter
SELECT COUNT(*), COUNT(DISTINCT author_handle), MIN(posted_at), MAX(posted_at)
  FROM forum_posts
 WHERE body_quoted_stripped LIKE '%scammers manipulating test results for profit%';
                                          -- 49 posts, 37 authors, 2025-06-02 to 2026-07-20

What this piece does not establish

We are checking claims, so it is worth being explicit about what our own checks do and do not support.

We are not saying HZSB is a bad vendor. We are saying it has no testing record we can find. Absence of certificates is absence of evidence in both directions. It is a reason not to act on a superlative, not a reason to believe a negative.

We are not saying the advertised retatrutide is unsafe. Its purity record is good. Its dose record is loose, by its own certificates, and its endotoxin claim is unsupported because nobody tested for endotoxin. Those are the specific findings and they do not extend past themselves.

We do not know what the 0mg tirzepatide certificate means. We have the certificate and the verify link. We do not have the chain of custody, and a 0mg reading has several possible causes, not all of them the vendor’s fault. We publish it because a vendor citing its own good certificates while a 0mg certificate exists in the same file is the exact reading problem this article is about.

We could not verify DjJoshua’s certificate. It is not in our corpus. His post is the best-structured claim in this survey and we still could not close the loop on it, which is a limitation of our corpus rather than a criticism of him.

Our corpus is not the market. 62,079 certificates is a large sample of what gets tested and published. It is not a sample of what gets sold. Vials that were never tested, and certificates that were never posted publicly, are invisible to every check in this piece.

Research and education only. Nothing here is medical advice, and nothing here is a recommendation to buy from any vendor named or unnamed.