Matthew Baggott

I'm a cognitive scientist with an interest in epistemic vigilance, and this essay that's been going around gave me pause.

I don't think it's straightforward to apply the concept of epistemic vigilance to interactions with LLMs, as this essay does.

🧵/

sbgeoaiphd.github.io/rotating_the...

‼️New paper‼️ Does your psychedelic facilitator shape your experience?

This is a key question considering that the psychedelic experience can predict treatment outcomes.

We analyzed the largest dataset of its kind, featuring 20yrs+ of psilocybin trials from @jhpsychedelics.bsky.social

Transcend redacted the 5D-ASC and at least one other end-of-session measure from their Phase 2 methylone protocol. We're no longer in the Doblin-knows-all-keeps-no-secrets era of entactogen research.

Some clarity could come if full measures are reported in the Phase 2 publication. But as competition heats up, companies may be strategically withholding endpoints that could predict or model therapeutic outcomes.

Even if R-MDMA is twice as potent, ~1.5µM is still far from classical psychedelic territory. Whether R and S-MDMA differ meaningfully on neuroplasticity or some other non-psychological mechanism remains unanswered.

A central unanswered question is what mechanism underlies R-MDMA's apparent efficacy. The 5-HT2A agonism story is appealing but shaky: racemic MDMA has an EC50 of ~3µM at 5-HT2A with only 65% Emax (Alberto-Silva et al., 2024), making it a weak agonist.

But that Lykos data analysis compared baseline to 18 weeks post-treatment. Timing difference or mechanism gap? More data required.

Phase 1 R-MDMA acutely boosted the "warm/connected" self-compassion subscales (CS) but not the "self-critical" ones (UCS). Agin-Liebes et al. proposed self-compassion as a mediator of standard RS-MDMA-AT outcomes in PTSD, and UCS reduction was the stronger predictor.

Unlike racemic MDMA, R-MDMA does not raise oxytocin, a proposed therapeutic mediator. And when given alone at these apparently required high doses, its half-life is ~14 hours. Longer post-dosing monitoring windows will likely be needed in clinical practice.

The Emotional Breakthrough Inventory, developed in psilocybin therapy, was moved by 175–225mg, with scores in the range of high-dose psilocybin. Almost no EBI signal appeared below 125mg. R-MDMA may require crossing an intensity threshold before therapeutically relevant phenomenology emerges.

In Phase 1, 225mg R-MDMA raised the Challenging Experiences Questionnaire: fear, grief, physical distress, insanity, and isolation all increased. But magnitudes were modest. Physical distress reached classical psychedelic levels; death and paranoia barely crossed p<0.1. No ego-dissolution.

AtaiBeckley's Phase 1 data are summarized in a US patent & assessed therapy-relevant constructs that Straumann et al didn't collect & that haven't been reported for this Phase 2a trial, including the Challenging Experiences Questionnaire, Emotional Breakthrough Inventory, and Self-Compassion Scale.

They frame R-MDMA as having reduced cardiovascular burden vs racemic MDMA. I'm not so certain: Straumann et al.'s crossover data suggest 225mg R-MDMA sits below the full therapeutic Lykos dose (~300mg R ≈ 125mg rac-MDMA), so the cardiovascular comparison may simply reflect lower dosing.

As with Transcend's Methylone for PTSD trial, there was no adjunctive psychotherapy. Either standard trial designs are flawed when applied to psychoactive therapeutics, or there are readily harnessable therapeutic mechanisms triggered by entactogens, or both (my guess).

Caveats: primary endpoint was safety, efficacy was secondary and exploratory, and the p-value is one-tailed. The NNT (~3) is essentially the same as Danforth et al. reported in their 2018 MDMA-AT trial for social anxiety in autistic adults (consistent, but from a similarly small exploratory base).

AtaiBeckley just announced positive Phase 2a results for EMP-01 (R-MDMA) in Social Anxiety Disorder: 49% CGI responders vs 15% placebo, 11.85-point placebo-adjusted LSAS reduction after just two doses over six weeks. Small trial, but looks like a signal.

ir.ataibeckley.com/news-release...

I'm interested in what you'll think of those two papers then.

And setting aside all these questions about 5-HT and mood, I think preventing or minimizing long term tolerance is an important goal if entactogens are to be a useful therapeutic.

But it is certainly odd how the literature doesn't have a convincing animal paradigm showing subacute mood or anxiety changes after MDMA or related drugs. The main lasting effect of high dose MDMA seems to insensitivity to later MDMA and some other serotonergic drugs.

And, in this slightly better theory, lowered 5-HT doesn't make you feel bad, but as seen in acute tryptophan depletion (ATD) protocols, it has some destabilizing effect that is expressed as depressed mood in a vulnerable subset of people.

That is certainly too simple a story. A slightly better story is that central serotonin is more about stabilizing processes and modulating flexibility.

Its release doesn't make you feel good but it does modify the effects of DA and NE to make MDMA have atypical effects for a stimulant.

A final point is that there is both fast-developing tolerance to MDMA and inhibition of metabolic enzymes. So redosing, which is common, yields higher exposures than you would guess from the experience or from single dose kinetics.

If you think transiently lowered mood after MDMA is related to 5-HT depletions, an unproven but theoretically appealing conjecture, then the lowered mood after 125 S-MDMA (equivalent in exposure to something below 250 mg RS-MDMA) seen by Straumann seems consistent with this AUC dose estimate.

I wouldn't take those estimates too seriously, since this is a quick and dirty approximation. But the truth for the population may be in this ballpark and sensitive/ vulnerable individuals/ neurons may get these effects at lower exposures.

Line graph titled "Rat 7.5×3 vs Human Equivalent Exposures." X-axis: Time (0–24 h). Y-axis: Concentration (ng/mL, 0–1600). Three curves are shown. Blue: Rat 7.5 mg/kg x 3 (doses at 0, 2, and 4 h) producing three sharp peaks around ~1100–1500 ng/mL and rapid decline to near zero by ~12 h. Green: Human matched Cmax (686 mg = 9.80 mg/kg) showing a single broad peak near 1500 ng/mL around 3–4 h with slow decline, remaining >300 ng/mL at 24 h. Red: Human matched AUC (241 mg = 3.45 mg/kg) showing a lower single peak around ~450–500 ng/mL near ~2–3 h and gradual decline to about 100 ng/mL at 24 h.

Or we can match AUC but have a much lower Cmax. Using a too simple model (1-compartment PK, 1st-order absorp, tmax = 2.4 h, t½ = 8.4 h; human Cmax/AUC ∝ dose^b with b = 1.12 based on literature Cmaxes), we can get rough ballpark estimates 240 mg or 685 mg to match rat AUC or Cmax, respectively.

But I'll foolishly try as an exercise. Aside from PK uncertainties, we also don't know what aspects of exposure to match. From this plot, we can already see that human exposures last longer, which means that if we match peak exposures the human total exposure (AUC) will be much bigger.

Line graph titled "Simulated Plasma Concentrations." X-axis: Time (0–24 h). Y-axis: MDMA concentration (ng/mL, 0–~1700). Two curves are shown. Black: Human 1.6 mg/kg PO, rising gradually to a single peak around ~250 ng/mL at ~2–3 hours, then declining slowly and remaining above ~50 ng/mL at 24 hours. Red: Rat 7.5 mg/kg IP x 3, with injections at 0, 2, and 4 hours, producing three sharp peaks approximately ~1200, ~1500, and ~1600 ng/mL, followed by rapid decline to near zero by ~12 hours. A black dashed horizontal line marks the rat peak (~1600 ng/mL), and a red dashed horizontal line marks the human peak (~250 ng/mL).

Rat regimens that lower 5-HT appear higher than typical human exposures (ignoring for now possible toxic metabolites), see my attached plot. And, in my view, given MDMA's nonlinear PK, we don't have enough PK data on high human exposures to do a great job predicting depleting exposures.

Possibly related, we also have increasingly good data that people sometimes have transiently lowered mood after higher exposures (e.g., festival attendees in Blankers et al. 2025; study participants after 125 mg S-MDMA in Straumann et al. 2024).

van de Blaak & Dumont (2022) did a nice analysis across imaging studies of high exposure users and found SERT appears lowered in a manner that plausibly suggests downregulation with a slow recovery.

What are you skeptical about? I don't think it's neurotoxicity per se, but whatever it is, it happens in essentially every mammalian species (tho mice are weird) so it seems to me a question of when not if for human exposures.