A publication by Onur Oncer // EST. 2026
The Signal Report.
Signal, not noise. Science and world events, sourced and explained: defense tech and electronic warfare, lab science, supplement reality checks, luxury home security, and energy storage. Every claim cited. Where I have a stake, the report says so, and what I link is what my research and experience point to. How this works →
Reports 183 · page 11 of 19
The Battery Fire Rate Fell 97 Percent
It is a real number from a real database, and it is the most quoted statistic in grid storage. The same document reports something the number cannot: of the 81 incidents recorded at the time of writing, only 26 held enough information to say what had gone wrong. The rate is measured. The reasons are mostly missing.
Read the report → Report 082 · Supplements · 2026-08-12 · 8 minMethylene Blue Is a Drug, Not a Supplement
A synthetic dye became a wellness staple, taken by mouth for energy and focus. The only FDA-approved version is an injection, for one blood disorder, carrying a serotonergic interaction warning with a 72-hour washout and a contraindication most people cannot rule out without a blood test. The largest trial of this chemistry enrolled 891 people and was negative.
Read the report → Report 081 · AI in the Lab · 2026-08-12 · 9 minThe AI Read the Spectrum. Now Count the Guesses.
The most-quoted accuracy figure for AI structure elucidation from NMR is 69.6%. Two things about it: that is the rate at which the right molecule appears somewhere in fifteen guesses, and it was measured on spectra a computer generated. On 106 real spectra, the same framework scored 33.0%. Getting structures out of spectra is my day job, so here is what the metric means.
Read the report → Report 080 · Lab Science · 2026-08-11 · 9 minWhat NIST Traceable Actually Buys You
The bottle your instrument is calibrated against carries its own uncertainty, on the order of a couple of percent for some elements, with a coverage factor printed per analyte because it is not always 2. Its certification expires on a date, is nullified by improper storage, and NIST's own policy says establishing the traceability of your result is your job, not theirs.
Read the report → Report 079 · Defense Tech · 2026-08-11 · 9 minRemote ID Is a Beacon, Not a Sensor
The FAA's drone rule requires a one-per-second broadcast on unlicensed spectrum, accurate to 100 feet at 95 percent, from aircraft that choose to comply. It also writes in the places where no broadcast is required at all. An electronic warfare officer on why identification has never been how anyone finds anything.
Read the report → Report 078 · Luxury Home Security · 2026-08-10 · 9 minWhat UL Rated Actually Means on a Safe
Three different UL standards hide behind that sticker, and they test unrelated failure modes. Until 2016 the residential burglary rating was one pass/fail test: five minutes, one person, a drill and a hammer. The fire standard is stranger still, because its own published scope declines to say whether the safe still works afterward.
Read the report → Report 077 · Energy Storage · 2026-08-10 · 10 minWhy a Grid Battery Ships Oversized
PJM's reference battery is rated 200 MW for four hours and specified with 1,030 MWh of cells. Thirteen percent of them are installed to be consumed, with cell replacements booked for year five and year ten. The national planning model handles all of it by prepaying the wear at a rate that assumes one cycle a day.
Read the report → Report 076 · Supplements · 2026-08-09 · 9 minDoes NMN Actually Do Anything?
The NMN pitch is built from three true statements, which is what makes it worth taking apart slowly. Take the capsule and a blood marker really does rise: two independent meta-analyses agree. Look at what that marker was supposed to predict and the same two find nothing. One writes it out plainly: an exaggeration of the benefits may exist in the field.
Read the report → Report 075 · AI in the Lab · 2026-08-09 · 9 minWhere AlphaFold Quietly Breaks the Chemistry
A lab made the wrong protein by accident. Measured, it had twice the helix of the original and came in threes instead of ones. Four AI predictors, given that same sequence, handed back the original to within an angstrom, with charged atoms parked in the greasy core where charge is not allowed to sit. Why the wrong answer looked so right.
Read the report → Report 074 · Lab Science · 2026-08-08 · 10 minReproducible and Replicable Are Not the Same
Nature published a package of meta-research on 1 April 2026. Two papers landed on a number near half and the coverage merged them. They measured different failures, and the more basic one could only be tested on a quarter of the sample.
Read the report →