I want to talk about the thing underneath all of this, because I don’t think I’ve ever laid it out plainly in one place, and it’s the actual reason the Cannabis Device Safety Institute exists.
Start with the only question that really matters. Not is this legal, not is this clean, not what’s the terp profile. The question is: does this help?
Does it help her sleep. Does it help his nausea during chemo. Does it help the seizures. That’s the question patients ask, it’s the question that justifies most of the money that’s ever gone into cannabis research, and I’m not a doctor — I build vaporizers. But I’ve come to think the people trying to answer that question have been blocked by something on my side of the fence, and nobody on either side has been in a position to say so.
Here it is. When you inhale cannabis, two things happen at the same time. The cannabis goes into your lungs. And the machine that heated it — the coil, the ceramic, the solder, the wick, the glue, whatever’s in there — is also hot, and it’s putting its own chemistry into the exact same stream. Same puff. Same lungs. Same second.
So when a study measures something afterward — a contaminant in the blood, a symptom that got better, a symptom that got worse — there are two possible sources for it. The plant. Or the machine.
And nearly every study ever published describes the device in about one sentence. A brand name. Maybe a wattage.
Why more studies don’t fix this
The first thing people say when I explain this is: fine, so run a bigger study. More subjects. That’ll wash it out.
It won’t, and the reason took me a while to be able to say properly.
Statistics are built to handle noise — stuff that’s random with respect to what you’re studying. Add more subjects and noise averages out. That’s what statistics are for.
But the device’s contribution isn’t random. It shows up every single time you dose, because dosing is heating the device. It arrives in lockstep with the medicine, from the same puff. It doesn’t average out with more people. It averages in.
And it’s worse than that, because it tracks the exact things a study cares about. Higher dose means more draws means more device. Higher temperature means more cannabinoids coming off and more off-gassing from the heater. So if you run a dose-response curve on an uncharacterized device, you’re measuring two dose-response curves stacked on top of each other and reporting the sum as one number.
This isn’t a precision problem you beat with sample size. The information you’d need to separate the two was never collected. That’s a different kind of problem, and it has a name: it’s an attribution failure. Two causes, one measurement, no way to tell them apart afterward.
I want to be careful here, because it would be easy to read this as me calling the existing research junk. I’m not. Those studies aren’t wrong — they’re under-determined. Their findings are consistent with more than one explanation, and they don’t contain what you’d need to pick between them. The people who ran them aren’t at fault for a gap they inherited. Nobody was ever responsible for the device, so nobody characterized it.
The four things
So here’s the whole sequence, and once I saw it in this order I couldn’t unsee it.
One: the device by itself. What does the hardware put out at temperature with no cannabis anywhere near it? That’s just materials chemistry.
Two: the medicine by itself. What’s actually in it — cannabinoids, the full terpene picture, pesticides, solvents, metals, whatever got added. “Cannabis” isn’t a specification. It’s a category with an enormous range inside it.
Three: the two of them together, under heat. What actually comes off when you put that characterized material on that characterized device and hit the button. Not what went in. What comes out.
Four: the human. Now — and only now — you put it in a person and believe what you measure, because you can finally say where the signal came from.
Each one is the precondition for the next. Run it in that order and every step is doable. Run it backwards and each step dissolves into the one before it. Which is why I keep saying, to people who think it’s a strange thing for a hardware guy to care about: you don’t get to the human question by skipping the machine question. You get there through it.
The part that surprised me
Here’s where I thought I understood my own argument and didn’t.
If you’ve measured the device, and you’ve measured the material, why bother with step three? Just measure what comes out and subtract the two things you already know. Right?
That doesn’t work, and the reason is kind of beautiful in a horrible way: heat isn’t a delivery mechanism. Heat is a reagent.
That hot coil isn’t a bystander watching the cannabis go by. It’s a reaction surface. The material breaks down, the pieces hit hot metal and hot ceramic, and they react there — and they make things that were in neither the device nor the material to begin with. You can’t subtract your way to those. They didn’t exist until you pushed the button.
And we already have a famous example of this. EVALI. Vitamin E acetate got blamed for the 2019 outbreak, and that’s right as far as it goes. But vitamin E acetate sitting in a bottle at room temperature is basically inert stuff — it’s in your skin lotion. What hurt people was what it became when it hit a hot heating element. The oil alone doesn’t predict it. The device alone doesn’t predict it. The combination does. Sixty-eight people died in the gap between those two facts, and there was no institution whose job it was to look there.
That’s not a special case. That’s the general rule showing its hand on a bad day.
The part that makes me want to put my head through a wall
I made a table of what each of those four steps legally requires. It goes like this:
Step four — the human one — needs an ethics board, informed consent, and federal controlled-substance compliance. It’s the hardest door in the building, and honestly, it’s not ours. We’re not a clinic and we’re not going to pretend to be.
Step three needs a cannabis license and a lab bench.
Step two needs a cannabis license.
Step one needs nothing. No license. No scheduling registration. No human subjects. No permission from anybody. It’s a bench, a method, and somebody whose job it is.
The legal gates run in exactly the same direction as the science. The one step in the entire chain that requires nobody’s permission is the step that everything else depends on — and it’s the one that was never built. Not because it was blocked. Because it was nobody’s job.
I find that genuinely maddening. I also find it encouraging, because a missing thing that requires no one’s permission is a thing you can just go build.
Where we actually are
Two papers went up on the site today laying this out properly, with the argument done carefully instead of conversationally:
- P-011, The Reference Device — why characterizing the hardware is a precondition for cannabis efficacy research, not a side quest.
- P-012, The Attribution Chain — the four links, and why the third one can’t be shortcut.
Now the honest part, which I’d rather say myself than have someone find.
We don’t have a bench yet. CDSI has been federally recognized as a 501(c)(3) for about two weeks. We have no lab, we’ve certified nothing, and every paper on that site is a founding draft with an open invitation to tear it apart. The two new ones are arguments — analysis, not measurements — and they say so on their face.
I don’t know how big the interaction term is. Everything above says those combination products exist. None of it says how much they matter. They might turn out to be small and boring. Unmeasured isn’t the same as dangerous — it’s just unmeasured, and I’m not going to dress up a question mark as an alarm bell. That’s the first thing a bench should be pointed at, and I’d rather find out I was worried over nothing than keep guessing.
And I build the stuff. I own hardware companies. That conflict is on the cover page of every single thing the Institute publishes, including this post, because a hardware safety body run by a hardware guy either discloses that everywhere or isn’t worth anything.
What I keep coming back to is this. It all starts with the devices — not because the devices are the most important thing here. The person is the most important thing. Always. But the machine is the part you have to understand first in order to understand any of the rest of it honestly.
Somebody has to go stand upstream and do the unglamorous part. That job’s been open for fourteen years. We took it.
The Cannabis Device Safety Institute is an independent California nonprofit, recognized by the IRS as exempt under IRC § 501(c)(3) and classified as a public charity under § 509(a)(2) (determination letter dated June 29, 2026; EIN 42-2429365). Recognition of exemption is a determination of federal tax status — it is not an endorsement of the Institute or its findings by the IRS or any agency. The founder holds ownership interests in cannabis hardware businesses (Divine Tribe / Nice Dreamz LLC), disclosed on the cover page of every CDSI document. Methodology, papers, and the public record: cdsi.click.