Nice response, thanks a lot! :) I might share some more thoughts soon (e.g., maybe the assumption that pain is a one-dimensional quantity could be doing some additional heavy lifting here; and I guess others have discussed problems with aggregationism, which may still apply even if all pains are comparable in the sense that you mean).

Thanks for writing this, Vasco! And thanks for inviting me to comment given that you expected me to disagree with the conclusion (which I do :)). I'm still confused about these questions but I think there are two analogies from physics/mathematics that might apply here and cast doubt on the post's conclusions, namely, (1) phase transitions and (2) topological transformations. 

NB: I recorded a voice note into Claude and then asked it to structure and clarify my thinking, which I then cleaned up.

Your argument, as I understand it, goes something like this:

1. A human body can be described by the state of all its fundamental particles.
2. For any two states that differ only infinitesimally, the pain intensities must be quantitatively comparable.
3. You can get from any state to any other state via a chain of infinitesimal changes.
4. Therefore, all pain intensities are quantitatively comparable.

I think (2) is doing a lot of heavy lifting here, and I don't think it's obviously true. The key issue is that continuous variation in underlying physical parameters does not entail continuous variation in the properties that emerge from them. This happens for example in phase transitions and topology.

Phase transitions

When you heat a ferromagnet continuously, nothing dramatic happens for a while. Then, at a specific critical temperature (the Curie temperature), ferromagnetism doesn't just decrease, but it vanishes entirely. The property "being able to attract iron" is not something that fades smoothly to zero. It disappears at a sharp threshold, even though the underlying control parameter (temperature) was varied continuously.

Similarly, when you cool certain materials, electrical resistance drops to exactly zero at a critical temperature (superconductivity). "Very low resistance" and "zero resistance" are not just quantitatively different. Superconductors exhibit qualitatively new phenomena (persistent currents, the Meissner effect, flux quantization) that materials with merely very low resistance simply do not have.

In our earlier exchange, I used the water/steam analogy, and you responded that we can still compare the temperature of water and steam. That's true, but I should have added that I think pain might not be analogous to temperature, but to some other property. Temperature is a property that varies smoothly across the phase boundary, but there are other properties that only exist in one phase. For example, steam can do mechanical work via expansion (it drives turbines); liquid water cannot. Liquid water has surface tension (it forms droplets, menisci, capillary action); steam doesn't. These properties aren't "less" or "more" across the transition. They're categorically present or absent.

So the question is: is pain intensity more like temperature (a quantity that varies smoothly across all physical states), or more like magnetism or surface tension (a property that can appear, disappear, or transform categorically / jump at critical thresholds)?

I think there's good reason to suspect the latter. For example, neurons themselves exhibit all-or-nothing action potentials. Below a voltage threshold, nothing propagates. Above it, a full spike fires. Continuous variation in input produces a binary output. If certain pain experiences depend on whether particular neural cascades fire or not, then the inference from "infinitesimal change in particle positions" to "infinitesimal change in pain intensity" breaks down at these thresholds.

Topological transformations

Consider a rubber sphere. You can stretch it, compress it, bend it, twist it however you like, and it remains topologically a sphere. But to turn it into a torus (a donut shape), you need to introduce a hole. The number of holes (the "genus") is an integer. It cannot change by 0.01. And yet this discrete invariant determines fundamental properties of the surface, like how many independent loops can exist on it.

Or consider a loop of string. You can deform it continuously all day long, but the moment a crossing is introduced and locked in, you get a knot, which is a topologically distinct object. The unknot and the trefoil knot are categorically different. You cannot "halfway" have a knot. Knot invariants are discrete, even though every local manipulation of the string is smooth.

Suppose the morally relevant feature of a conscious experience is something like a topological property of the neural dynamics: perhaps the structure of information integration, the geometry of recurrent loops, or the topology of attractors in the brain's state space. More concretely, if pain was, say, the degree of dissonance in the electromagnetic waves generated by the nervous system, then introducing a topological defect into the field could suddenly allow new levels of dissonance not possible without the defect.

If that's the case, then the argument "moving an electron by 10^-100 m cannot prevent the pains from being quantitatively comparable" would be analogous to arguing "stretching a rubber sphere by 10^-100 m cannot change its genus." For most configurations, that's true. But at a critical configuration, that tiny stretch is exactly what punctures the sphere and changes the genus from 0 to 1. The transition is sharp and local in parameter space.

So overall, I think the burden of proof is on showing that pain intensity actually behaves like temperature (smoothly varying, phase-invariant) rather than like magnetism, surface tension, or genus (categorically structured, with critical thresholds). I think it's more likely to be the latter (and that the symmetry theory of valence will be key in finding the solution). 

My intuition is that the conclusion "I would prefer warming up slightly cold patches of soil for sufficiently many nematodes over averting 1 trillion human-years of extreme torture" should be a reductio ad absurdum for the argument, and maybe the two angles above show how one might escape such a conclusion.

I hope this helps clarify where I'm at!

Indeed, I updated some model parameters for the simulations in the paper, which resulted in a lower DLES burden calculation. I believe the main difference was the estimated fraction of severe attacks relative to milder attacks.

Given the large error bars, I also included a sensitivity analysis in the paper, i.e., I calculated the DLES burden assuming lower and upper bounds for various key variables. This resulted in lower and upper bounds of 700k and 11m DLES.

addressing large, persistent gaps in care or prevention that lead to avoidable suffering or poor life outcomes

I'm so grateful you're looking into this! I think there's a lot of low-hanging fruit in this space (I'll soon share other ideas I have for folks to look into). I'll be following with excitement. :)

Thanks for all this info! Keep it coming; I'm keeping track and plan to make good use of it. Get in touch if you'd like to discuss more. :)

Wishing you all the best! :)

Thanks for sharing your thoughts! While I'm obviously sad that you've updated negatively, this exchange has been very helpful for me to reflect on how to communicate my level of confidence in these treatments, especially given the little data we have from RCTs specifically. The total evidence still looks overall highly compelling to me (I wouldn't be working on this ~full time otherwise), but I'll work on improving my communication (and generating more scientific evidence).

The main thing I'd like to say is that I'm really not committed to psychedelics as an intervention (since you brought up the lobbyist language), and I'm particularly excited and hopeful about non-hallucinogenic analogues (such as BOL-148 and Conjugated Psilocin), as well as about non-pharmaceutical interventions, which we're also exploring. Ultimately, I just want patients to have universal access to treatments that safely and effectively prevent or abort their attacks. Currently, I sincerely believe those happen to be indoleamines.

If the effect is really as strong as claimed, you wouldn't even need 100 patients for an RCT. Perhaps clusterfree could even make this happen faster

Absolutely agree, and we're actively thinking about how to do this!

Clusterbusters, because torturous pain belongs to the past (especially if it can be easily prevented).

I understand it seems hard to believe! And yes, I'm motivated to support / advocate for RCTs and I'm developing various ideas with some collaborators (not just involving psychedelics). I agree that more data published in peer-reviewed journals is very much needed.

I still genuinely believe that the funding and legal barriers are very difficult to overcome (plus other barriers). A good resource on this topic is the book Psychedelic Outlaws (describing the history of psychedelic use and research for CH). This deep research by Claude may also provide useful context: Why psychedelic research for cluster headaches has stalled despite decades of patient evidence. It concludes:

The documented barriers to psychedelic cluster headache research reveal systemic failures across every institution that might otherwise advance treatment. Schedule I classification adds years to research timelines. NIH has invested almost nothing in cluster headache specifically. Pharmaceutical companies see no return in treatments requiring only three doses annually. Academic institutions impose unique scrutiny on psychedelic protocols. Patient recruitment struggles with rarity, episodic presentation, and placebo reluctance.

The BOL-148 story crystallizes the tragedy: a non-hallucinogenic compound showed unprecedented efficacy in 2010, yet required 13 years to enter Phase 1 trials because it fell between orphan drug and commercial viability. As Bob Wold observed: "You can't introduce transformative medicine into a broken healthcare system." The gap between patient discovery in 1998 and the current state of research—with perhaps one dedicated US researcher and trials measured in dozens of participants—represents what Halpern called "a process that has begun 40 years too late."

Why suspicious? The fact that DMT aborts cluster headaches is a pretty recent discovery. The word is just beginning to spread among sufferers. And there are very few researchers working on this topic, and very little funding generally, and it can be pretty difficult to get all the permissions needed to run trials using psychedelics. (I wouldn't be surprised if Schindler's group at Yale was running the DMT survey as a preliminary step to justify running a proper trial. Note that Schindler "is believed to be the only researcher in the United States studying psychedelics in headache disorders.")