Robert Y. Chen

🚨 THE BIGGEST NEGATIVE RESULT IN PSYCHIATRY Psychedelics look like miracle cures for depression when compared to placebos. But a new meta-analysis asks: what if we compare them to open-label antidepressants? The massive psychedelic advantage vanishes 🧵

What is hybrid attention @grok and why is it specifically useful for agentic AI? And what is orchestrated context? Explain this to me at three levels of understanding. At each level, give me three analogies or one analogy is orthogonal analogy is adjacent and one analogy is somewhere between. When you get into the details, make sure to share literature, references, and primary papers where I can dig into the details. Also explain to me one contrarian view about this if it exists, and the main debate in the field between experts on this topic.

MiMo-V2-Pro & Omni & TTS is out. Our first full-stack model family built truly for the Agent era. I call this a quiet ambush — not because we planned it, but because the shift from Chat to Agent paradigm happened so fast, even we barely believed it. Somewhere in between was a process that was thrilling, painful, and fascinating all at once. The 1T base model started training months ago. The original goal was long-context reasoning efficiency. Hybrid Attention carries real innovation, without overreaching — and it turns out to be exactly the right foundation for the Agent era. 1M context window. MTP inference for ultra-low latency and cost. These architectural decisions weren't trendy. They were a structural advantage we built before we needed it. What changed everything was experiencing a complex agentic scaffold — what I'd call orchestrated Context — for the first time. I was shocked on day one. I tried to convince the team to use it. That didn't work. So I gave a hard mandate: anyone on MiMo Team with fewer than 100 conversations tomorrow can quit. It worked. Once the team's imagination was ignited by what agentic systems could do, that imagination converted directly into research velocity. People ask why we move so fast. I saw it firsthand building DeepSeek R1. My honest summary: — Backbone and Infra research has long cycles. You need strategic conviction a year before it pays off. — Posttrain agility is a different muscle: product intuition driving evaluation, iteration cycles compressed, paradigm shifts caught early. — And the constant: curiosity, sharp technical instinct, decisive execution, full commitment — and something that's easy to underestimate: a genuine love for the world you're building for. We will open-source — when the models are stable enough to deserve it. From Beijing, very late, not quite awake.

@Yikes2024IN Very different side effects profile

@therealRYC Now do the same for side effects...

🚨 THE BIGGEST NEGATIVE RESULT IN PSYCHIATRY Psychedelics look like miracle cures for depression when compared to placebos. But a new meta-analysis asks: what if we compare them to open-label antidepressants? The massive psychedelic advantage vanishes 🧵

@deepgreendesign @DrAdamBorecky @RCarhartHarris Not sure about neurological “health” But people have done imaging studies and show they definitely affect different parts of the brain In large ways

@therealRYC @DrAdamBorecky @RCarhartHarris Do psychedelics have other non-depression related benefits to neurological health?

@TeamPostX Time will tell!

@therealRYC And if we compare side effects? Long term risks?

@NELCIgenetics @DrAdamBorecky @RCarhartHarris My friends tell me I am but I’ve never believed them

@therealRYC @DrAdamBorecky @RCarhartHarris You sound fun to be around at parties

@RCarhartHarris Haha I get it, and I appreciate sober takes, it’s what I love about this platform

@therealRYC Wrong side of the bed this morning. Thanks for laughing 🙂

But one set of trials (TAD) were open label and other set of trials (psychedelic) included impt elements of experimental control, procedures like randomization and blinding. Does functional unblinding mid-session justify treating these trials as procedurally equivalent? No.

My biggest worry is that clinicians will look at this result and say “why give a psychedelic when I could give a SRRI?” But anybody who has tried to psychedelic or taking care of folks who have tried psychedelics will tell you that they have rocket ship level impact on human consciousness, function and well-being

@RCarhartHarris @therealRYC Pharma is terrified of plant magic. Cannot control something that grows on its own all over the world.

@RCarhartHarris LOL

@therealRYC Clever as a fudge to sell a click bait story that suckers buy into.

@GeorgeKirov1 Yes, some bias. And it might bias toward worse response in psychedelics, to your point!

@therealRYC I am actually all for comparing them with antidepressants. But the results make no sense as we use them for people who don't respond to antidepressants. Some bias there?

I think the pushback on a negative result framing highlights my point - that a core message of the analysis is that we have been fooling ourselves about the effect size on traditional placebo controlled studies. If we ran an open label trial of psychedelics and a SSRI and found no difference in effect but massive differences in dosing frequency , side effects, etc, that would be a negative result based on the primary endpoint. But many positives can come out from a negative result. That’s I guess my point - people assume negative result is a bad thing. To the exact contrary, science is about falsifying hypotheses. Hope that helps! TLDR negative results arent bad but everyone thinks they are

@therealRYC So at worst they have equal impact on depression but with dramatically fewer adverse effects and substantially fewer dosings and that is a huge negative result? You can’t be serious.

@Rob_D_Wilkens True!

@therealRYC Not being, allegedly, as effective as another treatment is not a negative result. How about side effects comparison?

Some great discussion from people in the field: The PI on the study: x.com/psybalazs/stat… @DrAdamBorecky on importance of functional outcomes: x.com/DrAdamBorecky/… @RCarhartHarris on how much better functional outcomes really are: x.com/RCarhartHarris…

💡 Takeaways Talking to patients about this, people really WANT to try psychedelic-assisted therapy. That leads to the challenge of expectancy and "know-cebo". But is that really a bad thing? What are depression scales if not subjective ratings. What arguably matters more is the functional recovery patients experience - getting back to the life they want to live. This paper is a vital reality check. Psychedelics are active, effective treatments — but they aren't magic bullets vastly superior to what we already have. They just benefit from an incredibly strong expectancy and therapy wrapper. And that might be totally appropriate. Huge shoutout to @psyhbalazs and @QuantPsychiatry for this important work. Negative findings are often equally if not more important than hyped up positive ones. Lots to learn from the methodology here, which integrated Bayesian statistics, clever study design, and efficient utilization of existing data. Source: Williams et al. JAMA Psychiatry 2026 — doi.org/10.1001/jamaps…