Every entry answers the same four questions, in the same order, so you always know what you're getting: what we found, what we got wrong, one thing worth checking at home, and what we're looking at next. The letters live on this page permanently — subscribing just means you don't have to remember to come back.
September to November 2026 — California's fiberglass ban leaves some glass in
Published 2026-11-12 · Entry 2 · covers three months
There was no letter in September or October. The promise at the top of this page says one a month, and it was broken twice, so this entry covers all three.
What we found
Three California rules for mattresses start on 1 January 2027, and this time we went back to the code text for each one instead of trusting our own summaries. Two read the way we had them. The PFAS limit for bedding and other textiles drops from 100 to 50 ppm total organic fluorine, and the 1,000 ppm cap on covered flame retardants stops exempting the parts of an adult mattress that aren't foam. Regulatory
The third is the fiberglass ban, and it is narrower than every summary we had read, ours included. The law defines "textile fiberglass" and then excludes two things from it: thread used to stitch mattress parts together, and yarn whose glass filaments are corespun or sheathed by another fiber. So from January a mattress can be fully legal in California and still have glass inside it. How many fire barriers are built that way, the text doesn't say, and we haven't found anyone who has counted. Regulatory — Cal. Bus. & Prof. Code §19101.5
All three, with the code sections, now sit at the top of the regulatory tracker.
What we got wrong
Our own tracker said AB 1059 was enacted in 2024, and that any brand using a fiberglass fire sock "must reformulate." It was signed on October 8, 2023, and the exclusion above means some glass-containing socks don't have to change at all. The entry was rewritten from the code text on November 12. Three of our articles described the law as a flat fiberglass ban. They were corrected on November 13.
A tool we built to clean up citations reported nothing left to fix. Meanwhile 157 evidence tags in the body of our pages still linked to a search engine's results page instead of the paper. The tool only ever read the reference lists at the bottom, so the tags inside the paragraphs, the ones a reader actually clicks, were invisible to it. A second tool now reads them, and we are checking its matching rule before it changes a single link.
And the October re-check of every cited paper against the retraction database didn't happen. The last full sweep was in August. It runs before the next letter.
One thing worth checking at home
If a listing or a salesperson tells you a mattress is "California compliant" for 2027, take that as a statement about the law, not about glass. Whether there is glass inside is still answered by the law tag sewn into the side seam: "glass fiber" or "fiberglass" in the fill list means yes, whatever the listing says, and the cover stays zipped. The fiberglass guide walks through reading the tag.
What we're looking at next
A plain-language piece on the January deadline for anyone buying a mattress this winter (now published: what the ban covers and what it leaves in). The retraction sweep. The two questions from the last letter, why old foam sags and whether a used mattress holds onto old smoke, are still open; neither got written.
The next letter lands in December, before the deadline.
— Ken Eyjolfson. How this site is written, including the role of AI assistance, is documented on the editorial standards page. Corrections are public, always.
August 2026 — the water wasn't getting worse; the measuring got better
Published 2026-08-22 · Entry 1
What we found
This month the Exposure Ledger's drinking-water column doubled, from 17 compounds to 34. It had only ever carried EPA's newest monitoring cycle, which happens to test for PFAS and nothing else; it now carries all five cycles back to 2001, which pull in flame retardants, solvents, metals and pesticides the newer cycle never looks for. That count is our own arithmetic on EPA's public monitoring files, and it is reproducible from the published dataset.
The find inside the data is a trap we nearly fell into. PFOA was detected in 2.38% of tested water systems in 2013–2015 (117 of 4,920 systems) and in 12.52% in 2023–2025 (1,288 of 10,289). Read as one line, that's a five-fold rise. But the lowest level the method could report also fell over the same decade, from 0.02 to 0.004 µg/L. The test got five times sharper, so it started seeing water it had always been blind to. Much of the "rise" is the ruler changing, not the river. This is why the ledger keeps every cycle with its own reporting floor instead of collapsing them into one tidy trend — a tidy trend here would be a false one. Regulatory
The mirror image is a zero that means nothing: an earlier cycle found no BDE-47 (a flame retardant) in 4,140 systems — at a reporting floor roughly a thousand times above the levels where these compounds actually occur. We print that as "the method couldn't see it," never as "it isn't there." Regulatory
What we got wrong
Three corrections this month, all from turning our own accuracy gates on our own writing.
One article opened with a scene about a woman who had "the same conversation with her doctor three times." She never existed — an earlier draft invented her for color. She's gone, and the rule is now written into our editorial gates: no invented people, no invented scenes, even sympathetic ones.
Another article closed with advice written in the first person that the person behind the pronoun would not actually give. Rewritten in the site's voice, and the rule is now explicit: advice never ships as "I" unless it is genuinely his.
And the most useful emissions study we cite on memory foam — the 2022 chamber measurements behind our reassuring one-year numbers — was written entirely by authors at an industry consulting firm. That fact was sitting in citation fine print. It now sits in the paragraph where you weigh the number, where it belonged all along: a reason to want replication, not a reason to throw out the measurement.
One thing worth checking at home
If a mattress in your house is thin — under six to eight inches — and its cover is sewn shut with no zipper, it is likely made of rebonded foam: chipped-up recycled scrap pressed together with adhesive, the same material as carpet underlay. That is the tier where the persistent-smell complaints cluster, and no certification logo tells you this. Nothing to panic about. Just know which kind you own; the off-gassing guide covers what to do either way.
What we're looking at next
Why old foam stops holding you up — the polymer-fatigue literature behind sagging, which almost nobody outside materials journals has read. A reader-prompted question from a thirdhand-smoke researcher: is a used mattress a smoke reservoir? And the quarterly re-check of every paper we cite against the retraction database, due in October — because a citation verified last year can name a paper withdrawn this year.
The next letter lands in about a month. Sooner only if something crosses the line in the promise at the top of this page.
— Ken Eyjolfson. How this site is written, including the role of AI assistance, is documented on the editorial standards page. Corrections are public, always.