Purpose: the vocabulary of cannabis device safety. Most of these terms don’t exist in the current literature, and that absence is itself the problem — you can’t regulate, peer-review, or certify what you can’t name. The Institute coins, defines, and stewards this terminology so that the field has shared language to build on.
Status: founding draft. Each entry is open for review and refinement. Entries will be cited in CDSI papers and protocols by their coinage tag (e.g., “LSO testing per CDSI Lexicon §3.1”).
Conventions: - Coined — first published use traceable to CDSI; original to this institute - Adopted — borrowed from an adjacent field (food science, indoor air, breath analysis); CDSI specifies its meaning for this domain - Reframed — existing word, but our usage narrows or sharpens the meaning
1. Foundational concepts
1.1 Hardware-side safety [reframed]
The category of cannabis consumer risk that originates in the device — the heater, the chamber, the materials, the geometry — as distinguished from concentrate-side safety (pesticide residue, microbial contamination, residual solvents, adulterants in the cannabis material itself). Existing state cannabis regulation almost exclusively addresses concentrate-side safety. CDSI exists because hardware-side safety has no institutional home.
1.2 Concentrate-side safety [reframed]
The category of risk originating in the cannabis material loaded into a device. Already extensively regulated by state cannabis bureaus. Mentioned in CDSI work primarily to clarify what we are not duplicating.
1.3 The Hardware Vacuum [coined]
The structural absence of a federal or state safety standards body for cannabis consumption hardware. A descriptive name for the gap CDSI exists to fill. Distinct from “regulatory gap” because it is institutional, not legislative — even if Congress passed a hardware safety statute tomorrow, no body currently exists to enforce or interpret it.
1.4 Builder-grade standard [coined]
A safety standard co-authored by the engineers, machinists, and ceramicists who build the hardware in question, then peer-reviewed by chemists, toxicologists, and materials scientists. Distinguished from academic-grade standards (no builder input) and trade-association standards (no independent review). The CDSI signature methodology.
2. Testing concepts
2.1 Loaded-state off-gas (LSO) [coined]
The thermal volatilome of a vaporizer operated with cannabis concentrate present in the chamber. Distinguished from dry-fire signature (§2.2). Most third-party testing of cannabis hardware to date has been dry-fire only — testing the empty heater. LSO testing reveals interactions between the heater material and the concentrate matrix that dry-fire cannot see.
2.2 Dry-fire signature [coined]
The characteristic off-gas profile of a heating element operated empty, at typical operating temperature. A material-only fingerprint. Useful as a baseline against which LSO results are compared. Some devices produce a worse dry-fire signature than LSO signature, which is itself a finding.
2.3 Thermal speciation profile (TSP) [coined]
The temperature-resolved set of compounds released by a device under load — i.e., what compounds appear at 350 °F vs. 480 °F vs. 650 °F. A device that passes acceptance criteria at one temperature may fail at another. TSP testing produces a curve, not a single value. CDSI-001 protocol specifies a three-temperature TSP minimum.
2.4 Heater-matrix coupling [coined]
The interaction between a specific heating-element material (ceramic, quartz, kanthal, titanium) and a specific concentrate matrix (distillate, live resin, rosin). Coupling effects mean that a heater certified safe with one matrix may not be safe with another. This is why CDSI-001 requires three matrices per device.
2.5 Pyrolytic reservoir effect [coined]
The phenomenon where residual concentrate, charred onto the heating element from prior sessions, re-cooks at higher temperatures during subsequent draws and contributes a measurably different off-gas profile from a fresh load. Reservoir effects are why “burn-off” cleaning cycles matter, and why session-cycled hardware needs distinct testing from out-of-box hardware.
2.6 Combined-loading testing [coined]
A testing protocol that exercises a single device across multiple concentrate matrices in sequence (or across replicate units, one per matrix). Distinguished from single-matrix testing, which is the industry default. Combined-loading is harder, slower, and more expensive — and is the only way to reveal coupling and reservoir effects.
2.7 Volatilome [adopted]
The full set of volatile and semi-volatile compounds released by a system. Borrowed from food science and breath analysis. CDSI usage: the device volatilome is the totality of compounds released by a vaporizer under specified operating conditions, measured by GC-MS and ICP-MS jointly.
2A. The device as instrument — precondition concepts
This section states the Institute’s foundational scientific claim: hardware characterization is not only a consumer-safety activity, it is a precondition for valid research into inhaled cannabis at all. Sections 1 and 2 describe what CDSI measures. This section describes why anyone else’s measurements depend on it.
2A.1 Device confounding [coined]
The systematic error introduced into any study of inhaled cannabis when the delivery device’s own chemical contribution to the inhaled stream is uncharacterized. In an uncharacterized device, every measured outcome — a contaminant, a carbonyl, a clinical effect, an adverse event — has at least two candidate sources: the cannabis material and the hardware that heated it. Device confounding is not a bias to be corrected statistically; it is an attribution failure, and no sample size fixes it.
Every other measurement science treats this as elementary. No analytical chemist reports a result without characterizing the blank. No inhalation pharmacologist would dose subjects through an uncharacterized nebulizer. Cannabis research routinely does the equivalent, and the reason is §1.3 — no institution was ever responsible for the device.
2A.2 Attribution floor [coined]
The concentration below which a device’s own contribution cannot be distinguished from background for a specified analyte, under a specified operating envelope. The attribution floor is the rigorous replacement for the phrase “tests at zero.”
This distinction matters more than it looks. “Zero” is not a measurable quantity — it is an artifact of an insufficiently sensitive instrument. Every device emits something; the question is whether the emission is below the limit of quantitation (LOQ) for the compounds that matter, across the temperature and draw range the study will actually use. A device is never “clean.” A device is characterized, and quantified below the attribution floor, for a named analyte panel, within a stated envelope.
Any CDSI document, talk, or interview that says a device “tests at zero” is making a claim chemistry cannot support and a reviewer will correctly reject. Say below the attribution floor and state the panel, the LOQ, and the envelope. The weaker-sounding claim is the one that survives.
2A.3 Reference device [coined]
A device whose contribution to the inhaled stream has been characterized across its operating envelope and quantified below the attribution floor (§2A.2) for a defined analyte panel. A reference device is not a safe device — it is a scientific instrument: a delivery vehicle whose own signal is known and bounded, so that what a study measures downstream can be attributed to the material rather than the hardware.
The analogue is a calibrated instrument, a certified reference material, or a validated assay blank. Cannabis science has certified reference materials for the plant and none for the device it passes through.
2A.4 Device-qualified study [coined]
A human or analytical study of inhaled cannabis conducted on a reference device (§2A.3), such that observed effects are attributable to the material under study. Its opposite — the current default — is a device-blind study: valid-looking work whose delivery vehicle is an uncontrolled variable.
The claim CDSI makes is narrow and testable: a device-blind study of inhaled cannabis cannot cleanly separate the effect of the medicine from the effect of the machine. This is why hardware characterization is upstream of efficacy research, not adjacent to it. You cannot honestly ask whether an inhaled cannabis medicine helps someone until you can account for what the device contributed to the dose.
2A.5 The instrument-first sequence [coined]
The Institute’s ordering principle, and the through-line of the whole papers pipeline:
Characterize the device → establish a reference device → then, and only then, ask what the medicine does to the human.
Each step is a precondition for the next, and the sequence cannot be run out of order — efficacy findings generated on uncharacterized hardware are not merely noisy, they are unattributable. This is the reason CDSI is a hardware body that takes human outcomes seriously: not despite starting with devices, but because starting anywhere else forecloses the answer.
2B. The attribution chain — the full sequence
Section 2A establishes that the device must be characterized before the human question can be asked. This section carries that ordering to its end: the device is the first of four characterizations, and the third of them is not derivable from the others. Vocabulary for P-012.
2B.1 The attribution chain [coined]
The four characterizations that must be closed, in order, before an observed human outcome can be attributed to an inhaled cannabis medicine:
(1) the device alone → (2) the material alone → (3) the two together under heat → (4) the human outcome.
Each link supplies a term the next one requires; the chain dissolves if run backwards and is tractable if run forwards. The instrument-first sequence (§2A.5) is the first half of this chain stated as an ordering principle; the attribution chain is the whole of it.
2B.2 Interaction products [coined]
Compounds present in the inhaled aerosol that are produced by the reaction of the material with the device at operating temperature, and are therefore absent from a characterization of either one in isolation. Heat is not a transport mechanism; it is a reagent, and a hot element is a reaction surface rather than an inert bystander.
Adulterant amplification (§4.3) is a documented special case — the adulterant alone does not predict the harm, the device alone does not predict it, the pairing does. The pyrolytic reservoir effect (§2.5) is a second case, and shows the interaction has a memory: the reaction vessel is not the vessel it was 200 draws ago. The general phenomenon requires no adulterant to be present.
2B.3 The non-additivity problem [coined]
The property of the attribution chain that makes Link 3 irreducible: the inhaled stream is not the sum of the device’s inventory and the material’s inventory. Because interaction products (§2B.2) are generated by the measurement conditions themselves, they cannot be recovered by characterizing device and material independently and subtracting. Link 3 must be measured as a system, loaded and at temperature, under realistic draw regimes — which is what loaded-state off-gas analysis (P-002, CDSI-001) exists to do, and why a dry-fire proxy is not a substitute.
The converse holds and matters equally: Link 3 without Links 1 and 2 is a measurement of a black box — you know what came out, not what put it there. Links 1 and 2 are what partition a Link 3 result into machine, material, and interaction. Link 3 alone is a measurement without an interpretation; Links 1 and 2 alone are an interpretation without a measurement.
2B.4 Chain integrity / the weakest-link rule [coined]
A study is only as attributable as the weakest link in its chain.
A randomized, blinded, adequately powered, competently analyzed trial run on an uncharacterized device with an uncharacterized product has impeccable statistics and absent attribution — its chain is open at three of four links. Statistical rigor and attribution rigor are independent properties; a study can max one and score zero on the other.
2B.5 The regulatory gradient [coined]
The observation that the legal gates on the attribution chain run in the same direction as the chain itself: Link 1 requires no license, no schedule, and no human subjects (it is materials chemistry); Link 2 requires a cannabis testing license and ISO/IEC 17025 accreditation; Link 3 requires both plus an emissions bench; Link 4 requires IRB review and federal controlled-substance compliance.
The operative consequence: the link with no legal gate is the link everything else depends on, and it is the one that was never built. The unblocking step in cannabis efficacy research required no policy change and nobody’s permission — only a bench, a method, and an institution whose job it was. The hardware vacuum (§1.3) is why no such job existed.
3. Standards and certification concepts
3.1 Pay-the-lab discipline [coined]
The structural norm that manufacturers pay testing fees but never buy outcomes. Operationally enforced by three rules: (a) testing fee is published and uniform, (b) every report is published regardless of pass/fail, (c) manufacturer responses are appended verbatim to the public report. The discipline is what distinguishes a credible testing institute from a pay-to-pass certifier.
3.2 Provenance certification [coined]
Certification that includes traceable disclosure of who performed the test, when, on what unit, with what concentrate, under what protocol — in addition to the pass/fail determination. Most certification marks today are pass/fail bumper stickers; provenance certification is the test report itself, summarized into a mark.
3.3 Open-data certification [coined]
Certification conditional on the manufacturer’s agreement that all raw data, instrument logs, and chain-of-custody records will be public. The CDSI Tier 2 mark requires open-data; closed-data testing exists in the industry but is not certifiable by CDSI.
3.4 Shadow standard [coined]
A de facto standard with no formal standards body behind it. The current state of most cannabis hardware safety claims. “FDA-grade ceramic” is a shadow standard; “medical-grade stainless” is a shadow standard. Shadow standards proliferate in any industry where the institutional vacuum (§1.3) is unfilled.
3.5 Tier-graded certification [coined]
A certification system with multiple levels of rigor, each clearly labeled. CDSI’s three tiers: - Tier 1: Reported — manufacturer submitted device, results published, may have failed - Tier 2: Certified — passed all current acceptance criteria - Tier 3: Gold Standard — passed Tier 2 plus committed to ongoing batch testing and full materials disclosure
Distinguished from binary certification (the typical bumper sticker), which collapses meaningful gradations into yes/no.
4. Failure-mode concepts
4.1 Coil failure [reframed]
Mechanical or electrical failure of the heating element. Already a term in vaping engineering. CDSI usage adds: any mode of coil failure that produces a non-baseline off-gas profile — not just “doesn’t heat” failures.
4.2 Glaze leach [coined]
The migration of trace metals or unreacted glaze constituents from a ceramic coating into the gas stream during operation. The 2021 V5 ROHS analysis was a glaze-leach investigation. Now a generalizable category.
4.3 Adulterant amplification [coined]
The increase in toxicity profile that occurs when an adulterant in the concentrate (vitamin E acetate, propylene glycol, polyethylene glycol, untested terpene additive) reacts with hot heater material at operating temperature. Distinct from the adulterant itself — adulterant amplification is the device-mediated harm.
4.4 Thermal runaway (cannabis-specific) [reframed]
In battery engineering, thermal runaway has a specific meaning. Cannabis-hardware thermal runaway is a separate phenomenon: the heating element exceeding setpoint by enough to alter the off-gas profile catastrophically (overshoot of 50 °F+ at high duty cycle). A device that runs away under load may pass dry-fire testing.
4.5 Session drift [coined]
The progressive change in off-gas profile across a single user session as the device, concentrate, and chamber heat-soak. Most testing measures the first few draws; session drift may not appear until draw 15-20.
5. Institutional concepts
5.1 Builders, not lawyers [coined — repeatable line]
CDSI’s positioning shorthand: standards written by people who’ve built the hardware (engineers, ceramicists, chemists working in industry) rather than by lawyers, lobbyists, or trade-association staff. The phrase is intentionally sharp; it signals which side of an institutional choice CDSI sits on.
5.2 Pay the lab, not pay to pass [coined — repeatable line]
The two-clause expression of pay-the-lab discipline (§3.1). Used as a one-line description of CDSI’s funding and reporting model.
5.3 The Founding Artifact [reframed]
The 2016 ALS Environmental off-gas report on the Divine Tribe V3 (service request P1605022). Used in CDSI documentation as both a literal reference and a methodological seed. Every CDSI protocol cites it.
5.4 Humboldt model [coined]
A proposed institutional pattern: a builder-led applied-testing institute partnered with a university research institute via formal MOU, headquartered in the same town, jointly hosting an annual symposium. This pattern is described as a future possibility, not a current arrangement. CDSI has no academic partnership in place at this time. The term is reserved for the day, if it comes, when such a partnership exists and proves replicable.
6. Reserved / under development
The following concepts are in active development and not yet published. Listed here to claim coinage and prevent re-invention.
- Volatilome divergence index (VDI) — a quantitative metric for how different two devices’ off-gas profiles are
- Material disclosure tier — the depth of bill-of-materials transparency a manufacturer commits to
- Closed-loop certification — certification whose criteria are themselves open-source and forkable
- Chemovar-aware testing — testing protocols that account for terpene profile differences across cultivars
Citation format
Reference entries in CDSI papers as: (CDSI Lexicon §X.Y, v. 2026-04). The version date is the lexicon’s overall draft date, not the entry date. Older usages remain valid; newer revisions add or refine without invalidating prior citations.
Coinage is consequential. Whoever names the field shapes the questions the field can ask. CDSI takes that responsibility seriously.
— Matt Macosko, Founder 2026-04-24