Checked against primary sources 2026-08-24
Fifty numbers, and what stands behind every one of them
Every change list in this market prints its destinations with the same flat confidence, and some of those destinations are wrong. This one grades each row and names the five it could not settle.
On this page
What twenty-nine dollars buys
50 rows. Each one is a number you already know, and each one answers four questions about it: where that rule sits in the 2026 edition, what now occupies the number you left behind, how confident we are, and which document we read it in with the date we read it.
The fourth is the reason it costs anything. Anybody can tell you that load calculations moved from Article 220 to Article 120. Very few will tell you that the row came off page 70-78 on NFPA free access on 23 August 2026, that the page header was read directly rather than lifted from somebody else's change summary, and that an independent numbered change list from the Minnesota Department of Labor and Industry says the same thing.
- 21 rows read in the 2026 book itself, on the date the row shows.
- 24 read in one of NFPA's own code-making panel draft reports, with the document and the revision number printed on the row.
- 5 we could not settle, printed with both readings rather than left out.
- 6 numbers that still exist in the 2026 book and now hold different text, which is the set that costs money on a job.
- 8 rows that say a number did not move, which is worth the same as a move and takes less time to act on.
Fourteen of the 50 ship inside the page and work with the network off. The other 36 arrive together for twenty-nine dollars, once.
The free half is a count, not a sample
A smaller version of a paid thing is not a reason to buy it. So the tool asks what you work in, in buttons, 12 areas of the book, and answers with a tally out of the 50 rows we hold: how many touch your work, how many moved, how many are graded against a document, how many are unsettled. Then it gives you the single row in your set most likely to catch you out, whole, with its provenance attached.
Three of those 12 areas hold nothing at all. Somebody who works hazardous locations finds that out in one click, and no price appears on that path. One more area holds two rows and both of them say nothing moved, and that path says so and offers nothing either. Losing those sales is the point: a tool that sells to somebody it cannot help is evidence against every row in it.
The 14 free rows are the five moves every change summary in the market already carries, plus one sting from each area that holds anything. Nobody pays here for what is already everywhere.
The confidence grade is the product
Four grades, and every row carries one.
| Grade | What it means | How many |
|---|---|---|
| Read in the book | Read in the 2026 edition on NFPA free access, on the date the row shows | 21 |
| Read in a draft | Read in one of NFPA's own code-making panel draft reports. The row names the document and the revision number | 24 |
| Inferred | Not read. Derived from how the article renumbered as a whole, and the row says so | 0 |
| Not settled | Two sources disagree and nothing we hold settles it. The row prints both readings | 5 |
The last row is the one nobody else publishes. Two of those exist only to say that a move you will find printed confidently elsewhere is one we could not stand up. One is a section move this site published for a day and then withdrew, and it sits in the product with that history attached rather than quietly deleted.
The draft grade needs a caveat and the tool carries it on every row that has it. A panel draft does not always number sections the way the finished book does. A first draft carried the new over-1000-volt grounding article under the working number 252 and it published as 270. Every number read from a draft here was cross-checked before it landed in a row, which is why the grade names the document and the revision rather than saying "verified" and stopping.
What it does not do
It is not a code lookup. There is no text box, no route that answers an arbitrary section number, and no code text anywhere in it. Every row tells you which number to turn to and you read the rule in your own book. This site will not build a searchable code database and this is not a quiet version of one.
It is not complete. Fifty numbers is what we have read, not what exists in a book of nine chapters. The whole grid, locked rows included, is visible before you pay, so whether your number is on the list is a question you can settle for nothing.
It only covers 2023 to 2026. We have not read the 2017 or 2020 editions against anything, so if your book is older than 2023 then everything here still applies and there is more we have not mapped.
It is not a subscription, it does not expire, and it does not know who you are. One payment. No account, no password, no mailing list, and nothing on our side that could identify you if somebody asked.
If it is not what this page says it is
Ask for the money back and you get it. Write through the contact page inside ninety days of the payment, quote the Stripe receipt, and the refund goes back to the card. No reason needed and nothing to fill in.
The rows already on your machine stay there. There is no account to close and no way for us to reach into your browser and take them back, which is the same reason we cannot tell whether you used any of it. That cuts one way on refunds and we would rather say so than pretend the honesty is free.
Twenty-nine dollars, once
Stripe takes the payment and emails the receipt. What comes back to your browser is a signed token saying a payment happened, and nothing else. The payment record lives with Stripe, we run no account system, and a cleared browser loses the token, so the receipt is the way back in.
Everything free stays free. The edition checker, the free rows, and every page on this site are unaffected by whether you ever buy anything.
- ToolOpen the crosswalk
Fourteen rows free, no payment, no account
- ToolCheck your edition first
Free, and it answers a different question
And the thing this site covers next, at no charge, is the part no crosswalk can carry: the amendments NFPA issues after a book prints, which change the code without changing a printed page, and the corrections log where our own withdrawn rows are recorded with dates.
Why any of this needs saying
Texas adopts the 2026 National Electrical Code on 1 September 2026 under 16 TAC 73.100. From that date an inspection runs against section numbers that in many cases are not the numbers in the book on the truck, and the expensive ones are not the numbers that went empty. They are the ones that were refilled. Text at the old address reads as though it belongs.
The market response has been a wave of confident lists. We read one that put a delayed effective date at a section number where the panel draft puts a reconditioned equipment heading. We wrote one ourselves that had a section move backwards for a day. The difference between a list you can act on and a list you cannot is not tone. It is whether each line tells you where it came from.
What this page cites
- 16 TAC 73.100 The Texas adoption of the 2026 National Electrical Code, effective 1 September 2026. Adopted by unanimous Commission vote on 28 July 2026. source
- NEC 450.5 Overcurrent protection of transformers at its 2026 number, with 450.2 Listing Requirements and 450.3 Reconditioned Equipment reserved ahead of it. Second revision 8259 in code-making panel 9 second draft, which numbers sections at their 2026 numbers. source
- NEC 410.2 Listing Requirements in Article 410, with Reconditioned Equipment at 410.3 and no section behind them renumbered. The clearest evidence that the edition template did not shift every article it touched. source
- NEC 120.13 General lighting unit load for dwelling branch circuits, read in the 2026 book on 23 August 2026 and held in this repository as a dated primary read. source