One Adoption, Seven Filing Numbers

If you have ever been handed a filing number for a Texas rule and told that is where the requirement comes from, this one is worth ten minutes.

We read the March 2026 adoption of the Texas consumable hemp rules end to end today — the whole page, fetched raw and searched rather than skimmed. About a hundred thousand characters. We went looking for one thing and found something else, and the something else is a mistake of our own that has been sitting in our published work for five weeks.

The adoption is not one document

It is seven.

The rules that took effect on 31 March 2026 were filed as seven separate documents, one for each subchapter of Chapter 300. Every one of them was filed with the Secretary of State on the same day, 2 March 2026. Every one of them took effect on the same day. And every one of them has its own filing number.

• TRD-202601064 — Subchapter A, General Provisions — sections 300.100 to 300.103
• TRD-202601065 — Subchapter B, Manufacture, Processing, and Distribution — sections 300.201 to 300.208
• TRD-202601066 — Subchapter C, Testing of Consumable Hemp Products — sections 300.301 to 300.303
• TRD-202601067 — Subchapter D, Retail Sale of Consumable Hemp Products — sections 300.402 to 300.407
• TRD-202601068 — Subchapter E, Registration for Retailers — sections 300.501 and 300.502
• TRD-202601069 — Subchapter F, Enforcement — sections 300.601 to 300.606
• TRD-202601070 — Subchapter G, Restrictions on Sale to Minors — sections 300.701 to 300.702

So a TRD number does not identify the chapter. It identifies one subchapter's filing. Hand someone TRD-202601066 as the source for a registration fee and you have sent them to the testing rules, where there is no fee to find.

We did that

TRD-202601066 is the testing filing. It is the right citation for the twelve COA fields and for the total delta-9 threshold, which is where we first used it and where it is correct.

Then it spread. Our own fact base cited it for the fee rule at 300.502, for the label rule at 300.402, and for the sale-to-minors subchapter at 300.701 and 300.702. None of those sections is in it. Those three entries are corrected as of today.

We also went and audited our published articles against the mapping above, because a correction to a private record that leaves the public pages standing is not a correction, it is bookkeeping. Twenty-nine of our fifty-one articles cite a filing number, and all twenty-nine cite the same one. Three of them name no section that filing covers at all. Thirteen name a testing section it does cover and then sections from other subchapters whose filing is never mentioned, so one number quietly stands in for four. Thirteen use it correctly and only for testing sections.

Those pages are not fixed yet. We are saying so here rather than waiting until they are, because the alternative is a silent gap between what we know and what we have published, and that gap is the thing this article is about.

How one right citation becomes six wrong ones

This is the part worth taking away, because the mechanism is not specific to us.

The first use was correct. Everything after it was a copy. Nobody re-derived the citation, because there was nothing obviously to re-derive — the number was already in the record, attached to the right chapter, with the right filing date and the right effective date. Two of the three fields were right. The filing date is genuinely identical across all seven documents, and so is the effective date, which is exactly what makes the wrong one so hard to see: everything around it checks out.

And a filing number is not the kind of thing a reader pushes back on. If a sentence says the fee is $50,000 somebody writes in. If it says the fee is $5,000 under TRD-202601066, the number that is wrong is the one nobody can check at a glance.

We tried to check it the way a reader would, and could not. Asked to resolve TRD-202601066 against section 300.502, a search answer returned the correct fee and then said plainly that it could not locate that filing number and suggested going to the Texas Register archives directly. That is the honest answer. It is also why a wrong filing number costs something: the reader who tries to verify it gets nothing back, and a citation that cannot be checked is doing decoration rather than work.

What we went in looking for: the fee

The question underneath all of this is one an operator types in a plain form. What does it cost to register a shop.

We published the figures a few weeks ago — there are four of them in circulation, all real, all from different documents, at gethempdash.com/learn/both-numbers-are-real. We are not going to restate that here; it does the job and nothing about it has changed.

What we did not have then, and have now, is the regulator's own account of why the number is what it is.

Here is the proposal, published in the Texas Register on 26 December 2025, quoted as printed:

> A retail hemp registration or renewal fee of $20,000 [$150.00] for each location

The square brackets are the Register's convention for text being struck. So that single line contains both the proposed figure and the figure it was going to replace. The companion section proposed $25,000 per facility in place of $250.

And here is the adopted rule, 25 TAC 300.502(g)(1):

> A retail hemp registration or renewal fee of $5,000 for each location is required before the sale of consumable hemp product.

Between those two sentences sits a document almost nobody opens. The adoption carries a response-to-comments section, and this is what it says, quoted in full:

> Comment: Most commenters oppose the proposed fees for consumable hemp manufacturers and retailers. Response: DSHS agrees with this comment. DSHS revised §§300.202(c)(1), 300.202(c)(2)(A), and proposed 300.502(f)(1) (renumbered to subsection (g)(1)) to set the fee for a manufacturer's license at $10,000 per facility per year, and a retailer's registration at $5,000 per location per year.

That is the whole explanation, in the agency's own words, and it exists in exactly one place. It is not in the proposal, which predates it. It is not in the adopted rule text, which states the figure and not the reason. It is not on any guidance page we have read. If you want to know why a Texas rule says what it says instead of what it was going to say, the response to comments is the document, and it is the one people skip to get to the rule.

The same response goes on to give the cost basis — it names Health and Safety Code 12.0111 as the cost-recovery authority, and lists inspector salaries and travel, laboratory testing, and related legal and administrative-hearing costs among the estimates. It also contains this sentence, which is worth knowing about and worth not over-reading:

> DSHS will reevaluate licensing fees periodically to ensure fees do not exceed the amount required to administer the program.

That is a statement that fees get looked at again. It is not a pending proposal, and we are not going to dress it up as one.

Two smaller things from the same page

The subsection letter moved. Read that response again: proposed 300.502(f)(1) was renumbered to 300.502(g)(1). The fee lives at (g) in the rule and at (f) in the proposal. A subsection letter carried over from the proposal era points at different text in the adopted rule, and someone quoting 300.502(f) is quoting a document that was superseded in March.

Adopted with changes, adopted without changes. Each of the seven filings lists which of its sections were adopted with changes, and therefore republished, and which were adopted without changes, and therefore not. Sections in the second group have their operative text only in the proposal. We hit this last week on the minors subchapter and it is not a quirk of that rule; it is how the Register works.

The number that is still being served

We asked the cost question four different ways this morning, to see what an operator actually gets told.

Two phrasings returned the $20,000 as something still coming. One of them put it in as many words: that these are still proposed changes and have not yet been finalized. The same answer also gave the correct current amount payable. So the reader is handed the right price for today and a warning that it is about to multiply, and the second half has been untrue since March.

This is a durable kind of wrong and it is worth understanding why. The $20,000 figure is not invented. It was in a real filing, and in January 2026 it was extensively and accurately reported — the "13,000% increase" framing traces to the arithmetic between the struck $150 and the proposed $20,000. Those stories were correct on the day they ran. Nothing retracts a true story. The rulemaking then concluded in a different document, six weeks later, that got a fraction of the coverage, and there is no mechanism anywhere that goes back and marks the first set as spent.

The proposal page itself is part of this. We checked it today for any banner, note, cross-reference or forward link indicating that it had been adopted, changed or superseded. There is none, and there is no reason there would be — it is an archive of an issue as published. But that means a document that was a proposal in December still reads in September exactly as it read then, and nothing on it tells you the answer arrived.

One phrasing did better, and it is worth recording because this lane is not uniformly wrong. Asked directly whether the fee is going up to $20,000, the answer hedged in the right direction: these appear to be proposed rules, and the department's own page shows $5,000, so the proposal may not have been implemented as written. Naming the number invited the caution that asking the price did not.

The practical version

A filing number is a subchapter, not a chapter. Before you rely on one, check that the sections you care about are inside it. For Chapter 300 the mapping is the seven-item list above.

Two right fields do not make a right citation. Ours had the correct filing date and the correct effective date, because all seven filings share both. The one field that varied was the one that was wrong.

If a number surprises you, read the response to comments. Not the rule, and not the proposal. The rule tells you what the requirement is. The proposal tells you what it nearly was. Only the adoption's response tells you which way the argument went and why, and it is the section people scroll past.

Ask what changed between the proposal and the rule, including the numbering. A subsection letter is part of a citation, and it can move on adoption without anyone announcing it.

A proposal is published permanently and reads as current forever. There is no expiry stamp on a filing, and the coverage of a proposal will almost always outweigh the coverage of the adoption that settles it. When you find a dramatic number, the question is not whether it was real. It is whether it is what got adopted.

We read the adoption on 28 September 2026 and the proposal the same day. That is the provenance for every quotation above, and it is the field we had been leaving out.