6–8 hour dosing sessions already exceed Medically Unlikely Edits (MUEs) for standard psychotherapy CPT codes. Category III codes (0820T–0823T) exist but are tracking codes only — not reimbursable under Medicare, Medicaid, or commercial insurance. No Category I CPT status until 18–24 months post-FDA approval at minimum. "Investigational," "no medical necessity," and "protocol non-adherence" will be the dominant denial vectors. Payers are already building internal task forces — prior authorization will be their first instrument, and it will be more complex than any existing BH PA schema.

The corpus doesn't exist yet and can't be built. Stratum's thesis is precedent-based intelligence — actual denial decisions and clinical necessity reasoning encoded across thousands of cases. For psychedelic therapy, those cases don't exist. Oregon is 100% self-pay. No insurance claims means no denials means no corpus. "First corpus in psychedelic denial intelligence" is a category error — you'd be building documentation infrastructure for a market that has generated zero insurance documentation.

Stratum has no revenue yet. H1 (knowledge asymmetry proves real) is unvalidated. Adding an 18-month speculative extension before locking paying customers in the primary BH market is how founders lose focus at exactly the wrong moment.

The timeline risk is real. MDMA got rejected. Ibogaine has cardiac risk concerns. If COMPASS hits a manufacturing issue or safety signal, NDA approval slips to 2028. Commercial coverage follows 12–18 months after that. Medicaid coverage (the dominant BH payer) could be 5+ years from now in most states.

Large AI RCM platforms already have an implicit corpus. Waystar, Availity, Cohere, and R1 have processed tens of millions of claims. At that volume, pattern-level denial intelligence is a statistical byproduct — they don't need a curated corpus because the signal emerges from volume alone. If they've processed enough Spravato and Auvelity claims, they effectively already have what Stratum is planning to build.

Their data has two structural weaknesses Stratum can exploit. First, BH claims have historically been processed through carved-out managed behavioral health organizations (MBHOs) with separate data architectures — the large RCM platforms have meaningfully worse BH data than their general medical data. Second, their implicit corpus degrades at low volume and high complexity: novel treatment categories, new drug approvals, cases where n is small. For COMP360 at launch, nobody will have the pattern data — and whoever builds it first has an 18-month lead before the volume players catch up.

The real H3 hypothesis refinement. The original H3 (corpus depth creates compounding returns) needs a sharper assumption: that a curated, structured, precedent-annotated corpus produces better outcomes than a statistically-derived one at the same volume. In BH specifically, this is a reasonable bet — clinical necessity logic is more narrative and less codifiable than medical. But it needs to be true, not just assumed. The Spravato corpus is where you validate this.

The competitive clock runs faster than the approval clock. The window isn't "get there before psilocybin approval in 2027–28." It's "get there before Waystar processes enough Spravato and Auvelity claims to make their implicit corpus good enough." That clock is already running and has a shorter fuse than the FDA timeline.

The SUD + psychedelic LOC intersection identified separately (PHP/IOP authorization for patients who received ketamine or psilocybin as part of SUD treatment) is a genuinely new denial category where this dynamic is most visible. No volume player has pattern data here yet. The category is forming now. This is the highest-leverage surface for Stratum to plant a flag — not because the market is biggest, but because the lead time against implicit corpus competition is longest.