Manifest 2026
Manifest is a prediction markets conference held in Berkeley, every summer, at a quirky venue called Lighthaven. The Lighthaven campus was converted from the old Rose Garden Inn, it’s a few separate buildings with gardens in between:
This year, I decided to attend. I wasn’t really sure what I was aiming for, but I thought it might be a chance to meet some interesting people, and a good way to learn what smart forecasters think comes next with AI.
As soon as I got off BART at the Berkeley station, I knew it would be a good weekend for AI doomerism.
20 years ago, Berkeley had street preachers telling you that hell is coming. Now they tell you that AI is coming.
I showed up around dinner time and quickly faced the struggle of finding where to sit and who to talk to. This was my second trip to Lighthaven, after attending the LessOnline conference in 2024. The first time seemed easier to navigate — both socially and otherwise. There were about twice as many people at Manifest as at LessOnline, but talking to them also seemed twice as hard. Even just walking around the venue was more difficult, this time. You enter the conference from the street, walk through a house, and exit through one of two doors. The space just outside those doors was densely crowded with people who are standing and talking:
This is kind of awkward, because now everyone needs to squeeze through that clogged area to get to any other part of the conference.
I’m not sure why so many people choose to stand in the way here. Is it like… because this is a central spot, you’re more likely to run into someone you know, you start a conversation, and then you never move?
After that, maybe other people that know one of the participants latch on to that conversation? Much like with atherosclerosis, the small blockage has now become a fatal problem.
At LessOnline, they had a zone full of tables, everyone ate there sort of like an outdoor cafeteria, so you’d just pick a table and make some new friends.
This time around, socializing at Lighthaven didn’t seem as simple. Last year’s dinner area was now a bunch of chairs facing a stage. It was great for watching presentations, but harder to sit and talk to anyone.
After walking around the venue for a bit, I noticed there were was a table in one corner, which looked sort of like it might have space for one more person.
A kind man named Ari sensed my hesitation, asked if I was trying to sit down, and I squeezed in.
Ari had faintly purple hair and was wearing a purple cloak. Maybe a peacocking strategy? Maybe just his favorite color? He had a variety of different outfits that weekend, with purple always worked in somehow.
We talked a bit and then I chatted with Frederic, a French investment manager. We discussed how growing up differs in the US and France. After we discovered we’re both into trading, and both doing well at it, we tried to gain some alpha from each other.
I think I was surprised by how many people in finance there were at the event — I was expecting nothing but career programmers. I suppose that means that prediction markets have made it from niche techie interest into something more mainstream.
I met a lot of other people over the weekend, often at some corner table like that.
There were a lot of people working on startups. The ideas all sound like they’ve been created as ad libs, where you’re required to always pick words like AI, prediction markets, forecasting, calibration, and betting.
One product promised to measure a user’s own statements with AI and would inform them if they were making accurate predictions over time.
Another aimed to detect things which had happened in the world that AI did not expect to happen, and sell information on these surprising events.
One startup allowed people to sell their company insider knowledge, which it would resell to other people interested in making predictions.
One guy told me he was involved with a project to run adversarial psychological experiments on people without their consent.
A few minutes later he said that I was cool and wanted to get my phone number.
I gave him my Twitter account, instead. I already tend to think of Twitter as an adversarial psychological experiment that’s being run on me — how much worse could whatever this guy does turn out to be?
Other than programmers and financiers, there were a lot of unemployed gamblers and nomads.
People talked about the other conferences and events they were visiting this summer. There’s something called Vibecamp, maybe a blend between this kind of conference and camping. There’s a popup city in Healdsburg. There’s a gathering of 20,000 libertarians in New Hampshire. I asked what kind of guns and drugs were present, and also how well 20,000 libertarians get along with each other. Apparently the main conflict is between the nudists and the clothed libertarians. I forgot to ask how roads work at the conference.
Talking to strangers isn’t the full event — at any given time there are half a dozen talks you can choose to attend. Here’s one line up, for mid-day on Saturday:
Sometimes it was obvious what I wanted to attend. Sometimes it was a conflict between two interesting sessions:
Sometimes I was curious but not really sure. Like, what about Aella’s talk?
I didn’t end up attending this one because I wasn’t sure if “fucking” was a verb or an adjective in that sentence.
Sometimes I was certain I was not interested:
The Manifest people also built some AI into their webpage to give you personalized recommendations. That performed well, but it wasn’t much different than just reading all the descriptions — it did function as a helpful timesaver if you didn’t want to read the whole schedule, or if someone added a talk at the last minute.
Lectures about AI
After meeting a few people on Friday evening, I remembered I’m largely here to learn more about AI, so I was off to hear about that.
The Manifest organizers have invented a probabilistic way to reserve space at each talk — you can say that you’re 80% likely to attend, or 50% likely, or specify some other percentage. Their system multiplies the number of people by those percentages to guess how many people will be in the room.
In the end, it rarely matters how many people have signed up — most talks end up overly full with many people sitting on the floor and a few stragglers awkwardly standing just inside the door because they don’t want to climb over others to snag a floor spot.
I started to wonder how the rest of the world would work with Manifest’s reservation system. Maybe you’d reserve things like this:
And then the flight would look like this:
While you should be polite and reserve your space at each talk, that doesn’t guarantee you a spot — you quickly learn that it’s better to just show up 10 minutes early if you want a seat.
The quality of the talks is hit or miss. Or, perhaps, the relevance and importance to you is — most of the speakers put effort into their presentation, but I learned a lot more from some talks than others.
I started with Chad Jones’ talk on AI and economic growth. Chad is an economics professor at Stanford. The talk discussed the conflict between AI forecasts and economic history. Some people think AI will produce incredible economic growth, but history shows that real GDP tends to just grow at ~2% per year, regardless of what new technology we invent.
Chad presented a model for why GDP growth is limited, which mostly focused on the concept of “weak links”. If you had, say, a business building cars, it wouldn’t matter if you could suddenly get tires for a 99% discount — you’d still need to pay for the rest of the car, assemble it, add some computers, market it, and sell it. Likewise, it doesn’t matter if you can fully automate one small part of the US economy if other parts can’t keep up.
He used that model to present some counterintuitive conclusions — you might think that fully automating the software industry would produce huge economic gains, but one of Chad’s models says that would only raise GDP by 0.5%. If all of cognitive labor (currently ~1/3 of GDP) was automated, he thinks GDP would only go up by 11%.
Chad came to another counterintuitive conclusion, regarding who benefits from automation. At the moment, Labor gets paid about 60% of the value of the economy while capital gets 40%. In a world where 100% of tasks are automated, capital eventually gets all the benefit. But in a world where, say, 97% is automated and 3% of work is not, we would go in the opposite direction and all of the profit would eventually go to that small segment of labor that isn’t replaced:
This briefly gave me hope for a future where people work meaningful jobs like teacher, carpenter, cook, or artist, and all these people are also very materially rich thanks to automated creation of physical goods.
The more I thought about it, though, I wondered if the last few remaining job categories would be something more grim, like: CEO, prostitute, and unemployed.
Chad presented one model where AI doesn’t change the growth rate much from 2%, even until 100 years from now:
Chad presented another model which was faster, I think he described this as the case where AI is able to produce something like “Moore’s law in everything”, with exponentially improving efficiency in every single aspect of the economy, so that all weak links vanish over time. That model does produce explosive growth:
In that explosive growth scenario, people end up 3 times richer by 2040 and at least 16 times richer in 2050:
But you still need 100% automation of every single task for the graph to look like “the singularity”, and that still doesn’t hit for ~25 years.
I would have to study Chad’s models a lot more before I can claim to understand all of this, but I think the talk made me update a bit towards “GDP probably won’t go crazy because of AI”.
Of course, that doesn’t assuage my fears that I’ll be unemployable as a programmer in a few years. One guy at the conference, recently out of college for CS, was talking about how brutal his search for an entry level job has been.
It also doesn’t help with my concerns that things could go wrong with AI. Chad mentioned that his weak links theory can also be applied to disasters. It doesn’t matter how many rockets you can build, if the O-rings fail. It doesn’t matter how great our factories are, if you suddenly get a pandemic and no one can go outside and spend money. Chad presented some math on the response to Covid, and showed that our huge fiscal response to Covid was actually efficient if you placed the value of one human life around 10 million dollars.
The French investment manager I’d previously met raised his hand at this point and asked if 10 million dollars was too high of a value to place on a person. Chad responded that 10 million was derived from people’s own choices — i.e. how much incremental pay they request to do a riskier job. I didn’t check the math here, and I think maybe you’d get the opposite conclusion if you measured people’s recreational choices — many people pay money to smoke, drink, bungee jump, etc.
Chad likewise argued that we’re badly underinvested in AI safety, not just from a human life perspective, but even if there’s even a small percentage that some side effect of AI can derail economic growth.
I meant to go to the Eli Lifland’s talk to get (what I expected to be) a more explosive forecast of AI growth, but I didn’t end up making it to that one.
At some point in the weekend I went to a talk on using AI trading bots to produce huge gains on Polymarket. At the beginning of the talk, the presenters clarified that they wouldn’t be sharing code or any secret knowledge, and I indeed left that talk with no idea on how to make an AI trading bot.
I didn’t end up attending as many other AI talks as I hoped to, so I’m still left unsure as to what’s coming next. We’re either headed for AI utopia, human extinction, or a steady 2% GDP growth rate.
I have the same sense of unease that I did, going into the weekend.
Lectures about China
AI might be the most important trend in the world, but I tend to think that the competition between the US and China might be a close second. I tried to attend every talk that discussed China.
I hit Steve Hsu’s talk first. Steve is a physics professor who has founded several companies in software and genetics. He seems alarmingly normal for someone so intelligent. I have a notion that high IQ is invariably associated with some other personality flaws — many of the smartest people I’ve met tend to also be socially inept or dealing with some other mild mental illness (OCD, autism, ADHD, insomnia, bipolar, etc.) High Jewish intelligence is associated with neurotic stereotypes (think of Seinfeld or Curb Your Enthusiasm). Steve finished college at Caltech at the age of only 19, but he also seems social, athletic, and able to hold down normal conversations.
Steve’s talk showed a sequence of what he calls “skull graphs”, where China’s progress eclipses that of the US. Here, he’s showing electricity production in the US and China:
The same thing is true in other cases, like R&D spending:
Or nuclear power:
And it might soon by true for AI models — China is months behind the US in AI quality, but the models are cheaper to run. People doing competitive routing of AI are increasingly using Chinese models:
Perhaps the most important statistic here was just the ratio of engineers — China has about 4 times the US population and Chinese students are more than twice as likely to major in technical subjects. That gives China an almost 10 fold advantage in the number of engineers that will be building the next generation of computers, designing new medicines, building new fighter planes, and so on.
From that lens, the competition looks hopeless for America. Our best bet is developing AI so quickly that it gives some permanent edge over China. Even that bet depends on a few key pieces — the assumption that China won’t catch up with EUV lithography, that recursive self improvement in AI development is successful, and that China won’t take Taiwan.
The talk gave me a few other insights into the Chinese economy. One point I hadn’t appreciated is that Chinese capitalism is, in some ways, more competitive than the US — instead of Uber vs Lyft, China has dozens of competing ride-share operators, and the same is true of many industries.
Q&A was busy at this talk, so I didn’t get around to asking the question I had, which was: “why hasn’t Chinese GDP caught up to the US?”
I think maybe that’s not a simple question, because it depends on how you measure GDP. In one sense, China already overtook the US more than a decade ago:
That’s GDP measured by “purchasing power parity” — in the aggregate, Chinese consumers buy more stuff than the US (to be clear, this is total purchasing across each country, not per capita).
But in a nominal sense, the US GDP is still higher. That is, US consumers still buy more “dollars worth of stuff”.
Prior to the pandemic, that was predicted to change, and China’s nominal GDP would be another “skull graph” relative to the US:
We should be at the crossover point right now, based on that forecast, but the prediction did not turn out to be correct:
I don’t entirely understand what happened here. China just keeps improving, as measured by PPP, but their economy is barely growing in a nominal sense.
I don’t think this is just due to exchange rates — the dollar strengthened maybe 10% against the Yuan, but the exchange rate has been range bound relative to the dollar for the last 5 years, so that can’t be the entire explanation.
The rest I don’t entirely understand — I think the US experienced inflationary growth (mostly because of excess fiscal response to the pandemic) while China experienced deflationary growth due to price competition (excess manufacturing efficiency and production), a housing market crash, reduced domestic demand, and higher unemployment.
Skimming Steve’s slides, I found this graph showing the importance of China’s deflating housing bubble:
But, does that mean the US housing bubble is a good thing? Or are we forcing people to waste money on overpriced houses, when China has pivoted to industrialization and R&D?
Either way, China is producing more stuff, and exporting more stuff, so aren’t they winning?
I want someone more skilled at economics to explain all of this to me.
I went to another of Steve’s talks, on Saturday night, where Steve and Razib Khan mostly talked about genetics.
This went about as expected, but I was surprised by one comment — I’ve been disturbed by Steve’s gradual shift towards MAGA politics over the last few years, and I was happy to see that he at least criticized himself for thinking that Kamala would drag us into war with Iran, while Trump would be the peace president.
There were a few other good talks on China at Manifest. Cremieux gave a great talk on how China is outcompeting the US in biotechnology and drug development.
Cremieux showed a “skull graph” of his own, showing that the total number of drug trials in China was exceeding the number in the US. He also showed that China could do larger trials, do them cheaper, and approve and schedule them faster. He argued that Chinese companies were getting a better return on R&D, by working with the government, while US companies were struggling to get any return at all.
Ross Rheingans-Yoo gave a similar talk about biotech, based on his experiences on a recent trip he’d taken to China. I found this one harder to follow, so I don’t have many notes to share.
I also went to Bohan Lou’s talk about the venture capital system in China. I also don’t feel confident in writing much about this, but it sounds like an oddly different system. Among other things, you have national 5 year plans, local cities trying to fund goals in support of those plans, and then cities funding local companies. It sounds like there’s also a strange degree to which Chinese startup founders have to go public more quickly, and greater risks where they can be held liable if they fail.
I have no idea how well this system works, relative to US startups and venture capital, but I did leave with the assessment that Chinese capitalism is highly competitive at a local level.
Finally, I went to a talk about bringing prediction markets to Asia — such markets are currently not allowed in Japan, Korea, or China, and it might take different challenges to open such sites in each country. That talk turned out to be given by two high school students, who gave a surprisingly good presentation. I left happy in a “the kids are alright” sense, but also a bit confused by today’s fashion. I’m familiar with previous trends like bell bottoms, skinny jeans, and so on, but these kids wore baggy jeans that were like 3 times the width of their legs. I left thinking that fashion will be increasingly difficult to understand, the older I get.
Lectures about Conspiracy theories
I saw that Adam Piovarchy had scheduled two talks on conspiracy theories. One was more about categorizing what makes a conspiracy theory, the other was about using Bayesian reasoning to disprove conspiracy theories.
I write a lot about conspiracy theories, so I hoped to attend both of these, but I only ended up making it to the talk on Bayesian reasoning.
I have a long and conflicted history with Bayesian reasoning. When I debated against Rootclaim, I argued that the case Saar made for the lab leak theory was a table of entirely made up and incorrect numbers. Most of the time that I see Bayesian reasoning used online, it’s for some other conspiracy theorist to take a bad argument and make it look quantitative and scientific. As such, I’ve leaned towards thinking that the entire Bayesian approach is pseudoscientific nonsense.
Since then, I’ve wondered if this judgment was premature. For many complex questions, there’s no single argument which establishes the right answer, you just have to go through some process of weighing the evidence for and against some conclusion.
In a recent article on the identity of Satoshi Nakamoto, I summarized the case for two Satoshi candidates like this:
The problem here is that some of these arguments should have a higher or lower weight, so I should add a multiplier to each one, to show how good or bad the argument is. But once you’ve made a table with both evidence and weights, you’ve basically just reinvented Bayesian reasoning.
At this point in time, I guess I’d have to say that Bayesian reasoning is a valid tool, the problem is just that many people are bad at reasoning, whether or not the style of that reasoning is Bayesian.
Get Peter Miller’s stories in your inbox
Join Medium for free to get updates from this writer.
Adam’s talk made a few good points about conspiracy theories — he talked about how real events often involve weird evidence. For instance, after 9/11, one of the hijacker’s passports was found lying in the street outside the world trade center. Adam explained that doesn’t make a conspiracy more likely, because there’s no particular reason why finding the passport would be more or less likely if 9/11 was an inside job.
He then moved on to writing an equation related to this, and I kinda stopped following — I think he was arguing that including the weird evidence changes both the numerator and the denominator, but it didn’t seem like turning this into an equation added much to the argument.
When we got to Q&A, I asked Adam three questions. I told him that I’d met a lot of Bayesian conspiracy theorists online, and asked if he had any ideas for interacting with people like that. I asked how much writing out his own theories as equations helped, as opposed to just putting them in words. At the very end, I snuck in a quick question as to whether he believed any crazy or non-conventional things, himself. Adam briefly said something about MK ultra being real, but then he had to leave for another talk.
I put in that last question because I tend to think that the best conspiracy theories are never as simple or dumb as “9/11 was an inside job” or “the twin towers were brought down with explosives, not planes”. The last few years have brought us questions like “was Covid a lab leak?” or “does ivermectin treat Covid?”, which aren’t as easy — I initially had a lot of doubt about the right answer to both of those. In other cases, the answer may be obvious but the details aren’t — when I wrote about the tic tac UFO, I was pretty sure it wasn’t aliens, but I still don’t actually know what the pilots saw that day. Scott Alexander had to write several posts trying to figure out what the miracle of Fatima was, before coming to a mundane conclusion.
I also think that none of us agree on everything, if you really dig into a person’s beliefs — we probably all believe something that another person might call crazy. This is especially true for ideas that overlap with politics.
I met Claire after the conspiracy talk. Claire was a delicate blonde that didn’t look like she belonged at this conference for unkempt nerds, or even in California, for that matter — she described herself as allergic to the sun and demanded to sit in the shade while we talked. She was in finance, not in tech.
She’d asked another good question at Adam’s talk, which was along the lines of — “who holds the burden of proof, when proving or disproving a conspiracy theory?”
That sounds simple, but it isn’t always — in the Covid origins case, lab leak proponents think the problem is that no one has found the natural host animal yet, while lab leak skeptics think the problem is that no one has found proof of a lab leak yet. Ivermectin proponents ask why no one has disproven the specific results of their favorite drug trials, while skeptics ask why larger, more recent trials didn’t work. UFO believers want a definitive rebuttal for why certain sightings weren’t aliens, skeptics ask why all the sightings are so bad.
After the talk, I also asked Claire what her craziest belief was. This was actually a question we asked all the Rootclaim judges, while interviewing judge candidates, but I’ve since found it can sometimes make for a fun conversation starter. Claire said that she thought Trump was blocking the Strait of Hormuz to starve China of oil and ensure US dominance — the US can now survive off domestic supply as well as newly available oil that Trump liberated in Venezuela.
I happen to think that this idea is pure MAGA cope. You can recognize most MAGA support for the Iran war as cope by its inconsistency. If Trump opens the Strait of Hormuz, they’ll say it’s Trump winning the war. If Trump closes the Strait of Hormuz, they’ll say he’s defeating Iran’s exports or starving China or some other 3D chess theory. If the plan is to starve China of oil, then Trump supporters should complain if he ever ends the war or, worse, removes his blockade to let Iranian oil get to China while Iran doesn’t allow anyone else to transit the Strait.
The same held true with Trump’s tariffs — some Trump supporters said they were necessary to help the US onshore manufacturing, others said it was to replace the income tax. But those two points are mutually exclusive — if you onshored all manufacturing, they would stop raising any tax revenue. Trump supporters then floated 8 different narratives for what the tariffs are for. And then Trump supporters also celebrated whenever Trump canceled the tariffs (“something something Art of the Deal”).
I didn’t push back too hard on Claire’s idea, for a few reasons. The most important is that I’d asked her the question, and it’s rude to ask someone their craziest idea and then promptly tell them they’re wrong. It’s far more fun to just listen to people’s ideas and try to understand why they think the things they do. The one time in my life I met a Flat Earther, I just listened and asked questions for an hour. Political differences are nothing like Flat Earth, but it’s still better to try to listen and understand, and try to figure out how your own biases differ.
Claire had worked a lot of interesting facts into this theory. She brought up AIS spoofing and observations by Citrini analyst #3. She talked about the mismatch between heavy and light crude oil — US refineries are configured for heavy oil, US fracking produces light oil, US refineries end up processing oil from places like Canada but they could also get heavy crude from Venezuela while we propped up Asian allies by exporting light crude.
I don’t think that Trump has thought through all of this, nor that the plan would really work (it will take years for Venezuelan oil to come online, the gap in Middle Eastern output is too large to fill, we’re mostly just treading water by drawing down reserves). But should I really be 100% sure that my own theory for Trump’s actions is better? I have a strong belief that Trump is just an idiot — he thought defeating Iran would be as easy as kidnapping Maduro and he got played by Netanyahu. But assuming Trump’s an idiot doesn’t always help predict the world — you have to think he’s an idiot that keeps failing upwards.
I’ve also gotten increasingly conspiratorial because of the Iran war, just in different ways — I’m now way more open to the idea that Israel controls US politics.
Unexpected events lead to uncertainty about the world, conspiracy theories fill that uncertainty with meaning.
I think you can probably avoid most conspiracy theories by relying on good heuristics, trusting good institutions (scientists, some legacy media), and thinking about biases. But I’m not sure that process is easy to put in any simple equation.
Lectures on social anxiety
At some point in the weekend, I attended a talk on social anxiety.
It seemed appropriate — this whole event was awkward and I found myself thinking about that almost as much as the bigger questions I’d come here to understand.
I sometimes feel like there must have been a day at school where everyone learned how to socialize properly, to introduce themselves and make small talk and proper eye contact, and I just missed that day. Or, maybe these are just the factory default settings that neurotypical people come with and I came out calibrated wrong.
One nice part about Manifest is that these aren’t all neurotypical people — while I’m pretty far along the spectrum for average people, at Manifest there are many people more autistic than me. One guy looked and spoke a bit like autistic TV character the Good Doctor. He explained that he couldn’t read facial expressions but instead tried to understand conversations by how quickly their facial expressions were changing.
Still, Manifest was challenging, so I decided to hit Aaron Kaufman’s talk on social anxiety. Aaron claimed that psychologists misunderstand social anxiety and he proposed an alternate theory: he thinks that anxiety is mostly just OCD applied to social situations — some people think way too hard about whether they’re doing everything right, and this perfectionism makes them doubt they’re doing anything well.
I… have a lot of doubt that this is correct.
The solution to getting good at most things in life is simply to practice those things often, to improve your skills. Aaron’s message seemed kinda like “don’t ruminate”, but I think the more practical answer is “learn to socialize better in a variety of situations through deliberate practice”. I know how to navigate some situations well. I could learn how to handle more of them.
It became clear through the talk that Aaron had personally done something like this. He’d done exposure therapy type approaches. He’d done speed dating events (even going as far as creating a false persona for one, with a hilariously strong Russian accent). And then he got up on stage and told people to just not think too hard about it.
The other big disconnect here was that Aaron came across as an intensely confident person, while trying to tell a room full of anxious introverts that they just need to be themselves and worry less.
This reminded me a bit of an excellent essay about mental health, by blogger “Hotel Concierge”, who noted that male speakers can never really discuss their own mental health problems because they have to project confidence and can’t afford to weaken their reputation by honestly sharing real struggles.
I started to think about some of the other speakers through this lens.
At Cremieux’s talk on biotech, I kind of got the feeling he was doing a forced confidence bit. He was trying to look the part, wearing a suit that day, perhaps the only guy at the conference who dressed up (the next day Cremieux was sporting a normal shirt, a cowboy hat, and a sunburn). He called out names of other people in the audience at the beginning of the talk (i.e. “we’re about to talk about China, hi Steve Hsu”).
When it came to the Q&A section, he answered some questions with “you should Google that” rather than dwelling on the uncertainties.
But, I don’t know, I haven’t met Cremieux before, maybe this is just his personality.
I’m not trying to criticize — it was quite a good talk. I’m just musing about the human condition, here. Another speaker I watched that weekend fell flat due to good content but lack of confidence. Confidence reads as a positive thing, however real or forced. Sometimes I’m confident because I know I’m good at what I’m doing. With everything else, the choice is either “openly admit I have no clue” or “fake it until I make it”.
In that one essay, Hotel Concierge offered this suggestion for anxiety:
I don’t know if this is good advice, either. I’m really good at obsessing on one topic (i.e. a research interest) and bad at setting timelines. I think I could turn 45 minutes of worry into hours or days of intensely focused fear and doubt. But I think that the “with practice, human beings can get used to just about anything” part still rings true, and I’d bet that is better advice than “just be yourself” or “just quit worrying”.
Good advice can be harsh. I recall reading some suggestion to stop thinking about shyness as a personality trait and instead see it as a personality defect. I find that it helps, every time I do this. Speak up! Interject into conversations. No one will ever go out of their way to care what you think or what you need, it’s up to you to tell them.
Manifest did have some workshops that I think might work better than sitting in on a “social anxiety lecture”.
There was “speed friending”, where you talk to people for 2 minutes and then pick a new partner. There was “slow friending” where you talk for 15.
There was… whatever this is:
A guy named Patrick Langford had scheduled a lot of these social events.
One of his events, called “Hot Seat”, sounded interesting — one person takes the center for 5 minutes and takes challenging questions from the audience (something like this came up in “Wayward”, a Netflix series I liked).
Another late night session was for people to talk about inappropriate views:
That sounded interesting for a second, except that… this is the guy giving the workshop:
Some people in the rationalist community are really into “Race and IQ” theories, so it seemed like this talk might be a trainwreck where some clueless nerd comes in and tells the socially adept host that he is genetically inferior (maybe the audience begins to wonder if this whole IQ idea is all it’s cracked up to be).
I’m not sure how that session actually went. I also don’t need a session to vent about my inappropriate opinions — I already have a blog for that.
Anyways, props to Patrick for setting up that and so many other sessions.
I went to one of these “speed friending” events at LessOnline. I didn’t get around to any this time at Manifest, but I probably should have, it might be a good strategy to do that on the first day and meet a bunch of other attendees.
I do wonder if there are other shortcuts I’m missing, for meeting people more easily at these kinds of conferences.
Manifest’s AI recommends some people to you, but I have zero clue what you’re supposed to do with those recommendations — text them in the app and ask where they are at the conference? (“Drop everything, AI says we’d be great for each other”)
I wondered a bit about the peacocking approach. I mentioned Ari and his purple clothing. Razib has his cowboy hat. Eliezer goes further, with sparkly hats, shiny coats, and a methuselah beard.
I hesitate to try something like this. My experience growing up lead to using clothing more like camouflage — people are less likely to fuck with you if you don’t stand out. I went an entire year of middle school wearing a Chicago bulls sweatshirt, even though I don’t watch basketball.
I also wonder what the conference would look like, if we all tried to stand out like this. Imagine 800 people doing an Eliezer bit, all wearing sparkly hats and coats. The entire conversation consists only of the word “doom” — some speak it in hushed tones, others throw their hands in the air while shouting it.
I don’t know, it could be entertaining.
Maybe somebody should make Doomfest 2027.
That is, if there is a 2027.
Lectures on psychotherapy and life choices
Near the end of the weekend, I followed a new friend into a talk on psychotherapy, given by a guy named Damon. This was probably a mismatch to my interests, but a few people had high praise for the host: other than working in mental health, Damon is known for writing great Pokémon fanfic:
At his talk, Damon said that most problems can be solved in 3 steps, and offered to solve people’s individual problems.
I wasn’t really sure what to expect, I thought maybe the 3 steps would be:
Damon entered the room, sat on a table in lotus position, and delivered his wisdom. Over the next hour, he gave a variety of life advice, sometimes delivered with analogies to Magic The Gathering mana colors.
Damon’s 3 key steps were:
• Figure out what you want
• Get more knowledge
• Get more resources
Some of us were already lost by step one. Damon started sharing this list by asking the audience, “Do you know what you want in life?”
About two thirds of the audience raised their hands. I did not. I think that can be a tricky question.
When someone asked more about how to know what you want, Damon stated that your body usually knows what it wants, and you should listen.
I think that’s a bad answer.
Like, what about food?
My body always wants to eat, but eating too much is bad.
What about drugs?
Drugs can feel amazing. They’re also bad for you, but it’s not always easy to figure that out because the benefits are immediate and the costs are not. I used to be addicted to sleeping pills. It actually took over 10 years of taking them before I really felt the downsides — in many of the years in between, I felt like I’d discovered an amazing cure for insomnia, and a highly useful tool. A decade later, withdrawal turned my life into a nightmare.
Drinking might start as a great social life, turn into a habit that impairs your success in life, and decades later turn into death by cirrhosis.
What about sex?
There were a handful of girls at Manifest where my body reacted by saying, “I want that”. (I noticed 2 of those girls made it to the Razib Khan + Steve Hsu talk, which lead me to wonder — does eugenics cause good genes or do good genes cause eugenics?)
I didn’t pursue any of those girls for a simple reason: I’m happily married. But I still find other women attractive. And if I were married to a different girl, I know I would still find other women attractive.
Razib and Steve’s talk maybe gave a hint for what I’m actually seeking — they mentioned that prehistory is full of star shaped phylogenies for Y chromosomes. Steve asked Razib to give a very simple explanation for what this means, but I think Razib didn’t quite nail the answer. I would have started with something like, “when a Mongol warlord and a mommy, and a mommy, and a mommy, and a mommy…”
History is full of cases where one man (like Genghis Khan) had children with hundreds of women. The genetic effect is accentuated because Genghis Khan’s children also had higher social status — even though the kids didn’t go on to have hundreds of wives, they still had more kids than the average person, exaggerating the genetic effect of Genghis Khan’s initial success.
I worry we’re destined for a star shaped Elon Musk phylogeny in the future, but maybe we’ll luck out and Elon’s kids won’t also have high status or big families. Or maybe the woke mind virus will turn more of his sons into fabulous divas.
I sometimes wonder if we’re all just polygynous apes pretending to be monogamous. I’m attracted to a lot of girls because my DNA is telling me to be the warlord with a harem. That idea sounds fun for about 10 seconds, until I realize that if that were an option, I’d be dead and someone else would be the warlord. Then I go back to thinking monogamous social contracts are a good idea.
Anyways, my body has similar preferences to the average caveman, in an entirely different environment.
Your preferences may not fit the modern world. They work poorly in a world of plenty. They can get hijacked by superstimuli.
Even if you know exactly what you like, life is still about making tradeoffs — how much to delay gratification, how much of a good thing is too little or too much.
I like my life, but I think it’s hard to know when you’re just stuck at some local maximum, where life could be much better if you made some hard changes first. Those kinds of changes are usually not pursuing what your body wants — my body doesn’t want to get another degree or search for a higher paying job — but maybe I’d benefit if I made that effort.
Later in the talk, Damon mentioned people having unreasonable goals. That section did get kind of interesting.
Some of the careers that sound most interesting also sound totally unattainable. Would it be cool to be an astronaut? A secret agent? A billionaire?
Probably, yes, but most of us aren’t going to succeed at those goals.
I like climbing mountains. I’ve done some good climbs, but there are always bigger goals that I haven’t tried. I haven’t climbed K2 yet. Is that a good goal, or just something that would kill me? I could “solve the problems” that are holding me back from doing that climb. But maybe ambition is the actual problem there.
The actual rewards might not be there, for some ambitious goals. The day to day as a CEO might be just as boring as the day to day for an average job. Climbing the really hard mountain might be about as fun as climbing the average mountain. Meanwhile, some of the most pleasurable experiences you’ll actually have in life might not come from some ambitious goal. A seasoned older hiker once told me the #2 most pleasant experience in his life was trying heroin. His #1 most pleasant was when he got altitude sickness and then EMT’s gave him oxygen.
Damon’s talk made me think of Donnie Darko’s experience with a guru that came to his school, which Donnie Darko ended up heckling:
I also tried to heckle Damon’s talk a bit, but I quickly realized I was just pissing off the people who came there for help.
There’s a supply of gurus because there’s demand for gurus.
Wrapping up the event
I ended the weekend watching the UFC Freedom 250 event held at the white house:
This was partially because I do like to watch UFC, partially because I was curious which other rationalists did, and partially because I wanted to see just how ridiculous cage matches on the white house lawn would get.
The most interesting fight I saw was Hokit vs Lewis, a match weird enough to trigger its own conspiracy theories.
The first round ended with Hokit putting Lewis in an armbar, but then he somehow failed to complete it, and Lewis simply waited out the clock rather than tapping out.
Some fans concluded that Hokit lacked the skill for this technique. Others theorized that he really wanted to win by knockout, so he intentionally failed here.
Hokit went on to win in the second round, and then shared a conspiracy theory of his own with the crowd, declaring, “Michelle Obama is a man!”
I’m old enough to remember when Zelensky demeaned the white house by refusing to wear a suit, but this is dignified behavior?
I didn’t know Jacob Falkovich was a UFC fan, but I’m glad he hosted that.
Lighthaven held an after-party on Sunday night. You could sing Karaoke, get drunk, or listen to some music. Some of that sounded fun, but I’d already spent ~24 hours at the event, and I needed some exercise, so I skipped that.
After UFC, I went for a nice quiet walk in the Berkeley hills to unwind.
Rating the event
Manifest organizer were insistent that we fill out surveys on how much we liked the event. Those surveys are mostly just a bunch of “rate this aspect on a 1–10 scale” questions:
I didn’t get around to answering these.
Like, okay, some of these questions are easy — the food was really good. I got to eat Indian food, and Vietnamese, and Mexican, and Poke. I got to eat a plate of shrimp curry while listening to a talk from Bentham’s Bulldog, a big shrimp welfare advocate.
But I don’t know how to rate most of this stuff on a 1–10 scale.
Sometimes I wonder if hell will just be an eternity of filling out surveys about your life on a Likert scale.
I also don’t think the world needs to adapt to me. The Manifest organizers made an interesting event with lots of choices for what to do. I showed up, picked some stuff, and tried to get something out of it. However good or bad my experience was is more about my choices than theirs.
I said the same thing about Burning Man, when Sam Kriss criticized it — Burning Man is a diverse mix of 80,000 people trying out experiences both mundane and profound. If you can’t find something there to enjoy, there’s a problem with you, not the event.
I don’t regret spending 3 days at Manifest. I might go back. If I do go again, I’ll try to give some lecture of my own, and I hope at least one person heckles me or writes a critique.
