(1)

I met Max Chiswick in June of 2024 because I wrote something wrong on the internet. He tracked me down at Manifest to tell me so—I had written about the shortcomings of poker as a teaching tool, but I had made several wrong assumptions about how one would play poker differently with the goal of learning instead of recreation.

First, you'd play with two players instead of eight. We nearly always play with eight-ish players because we want a relaxing game, where large gaps in the action are considered a feature. With two players, though, not only is it your turn four times as often, but more of your decisions are "live" because it's correct to play more (initial) hands and there are fewer that you should fold on sight.

Second, you'd play a game with fewer chips—meaning that you'd make bets in coarser increments and with a smaller ratio of largest-possible bet to smallest-possible bet. This makes the decision tree shallower and more amenable to explicit case-by-case reasoning, and it also has the nice effect of making the highest-stakes decisions less terrifying (while keeping the stakes of bread-and-butter decisions suitably meaningful).

But most importantly, if you're playing poker as an exercise for training your mind, you should focus less on becoming a theoretically-optimal robot. You can read books and blogs about "game theory optimal" play (e.g., Nash equilibrium) and drill yourself to get arbitrarily accurate at executing it, but that's simply not the useful part.

The part of poker that makes you smarter—Max explained to me—is observing the non-optimal patterns in your opponents' play, and playing the non-Nash strategies that exploit them. Instead of assuming that you are playing against a perfect player and aiming only to avoid imperfection yourself, you ask "What evidence can I see? What does that suggest is happening right now? How should I act in the world I find myself in? Can I get even more information about this? How?" You play the game, to win, as if it were an imperfect microcosm of an imperfect world. Because it is.

Put that way, I was (and am) convinced that my original post was arguing against a strawman. I had never actually played the kind of poker that would make a fair comparison! And, as the conversation went on, I—we—got excited about co-authoring a post exploring just where my first one had gone wrong.

We met the day after the conference; Max moved his flight to the evening so we could keep talking. And things turned from my original blog post, to our respective (mostly shared!) thoughts about the value of teaching through games in general, the richness of "find out what works" versus "learn what theory says", and what we were each up to next. And we hatched an idea.

We could—we realized in that living room meeting— we could and wanted to write an online course that presented a vision of 'games to learn through' the way we thought it should be presented.

There are moments in life where, if you're paying attention, your future forks and you get a choice. The project which wasn't even a suspicion twenty-four hours ago is now a possibility.

You can leave it as just that, an interesting possibility for an interesting afternoon.

Or you can decide that, whatever your plans were yesterday, you're now doing this.

Once upon a time, on a different continent and two years before, I had written games to teach trading skills. (I was working at the time at a New-York-based proprietary trading firm.) I loved that work. If ever we solve all of the world's problems and there's nothing that's important to do but each of us figure out how to well, then I'll be making games that teach. But now I was working on something different. I thought.

Screw that, I decided to write a course on games. We decided to write a course on games. Max had planned to expand his AI Poker Tutorial into an online textbook, but together we'd to go after the larger—and more satisfying—goal of entirely re-building it around interactives and challenges instead of readings. Then, after we'd taught the course to a few hundred students, we could pull it into a textbook.

It was June; could we be ready for when colleges start in September? Sure. Well, should we ship a Minimum Viable Product sooner than that and figure out what worked? That would be good. When would we have to…

Four weeks, we decided, was how long we had to write our first ten-class course. We'd teach through the lens of 'writing AIs for games', taking our students to the point where they could replicate the state of the art in 2019—when reinforcement-learning models beat human players in multi-player No-Limit Texas Hold 'em. Not with lectures, but with a series of weekly hands-on challenges. We were ambitious and a bit crazy—but hey, we were in San Francisco.

It being San Francisco, we did indeed launch our in-person beta on July 15—one week later than our initial goal, but just five weeks after Max and I met for the first time—at the SF Commons. (The amazing Ricki Heicklen also helped us prep and run that first class.) Just over forty students turned out for the first day! We had an interactive-first reading and challenge for them to work through in pairs.

"I'm exercising a part of my brain that hasn't been used since college!" said one of our students, "I feel like this is the kind of [stuff] that staves off dementia!"

And so, in a scrappy, startup fashion, we made it through our five weeks. Our students wrote AIs for toy-sized poker variants and imbalanced rock-paper-scissors games. They—we—discovered bugs in my reference implementation of counterfactual regret minimization and experimentally determined the conditions needed for it to converge that I had misunderstood. (Then we discovered and fixed the bugs second implementation attempt.)

Max and I built the airplane as we flew it—we met Sunday and Monday to prep for Monday's class, took Tuesday off to recover, Wednesday and Thursday to prep Thursday's, and…

In the end, we didn't get to AIs playing full games of poker—but we learned an incredible amount about what we needed to make to get our students there. We felt that we could do it with weekly classes over ten weeks instead of twice-weekly classes over five.

I had told Max—I had told myself—that this was only going to be a part-time thing for me. But he kept showing up to work on it. And I found that I kept showing up too.

In September, we tried something different with a single-day hackathon (at Fractal Tech Hub) that challenged students to write programs to play rock-paper-scissors against a series of flawed opponents (and each other). Our participants had a blast ("Felt like a mix between LeetCode and an escape room… I've never done anything like it!") but we weren't satisfied. We thought it could be so much better if we spent a few weeks improving it.

A few days later, Max asked whether we could do it again at Recurse Center in about a week. We could let that be an interesting possibility—or we could decide to do it. We did it.

Max found a seed-stage venture accelerator that was running a batch for games startups. Were we a VC-backable games startup? Well, what would it look like if we were? The deadline was in—oops what, less than a week? Are we really doing this? Well, we put together our best concept for what we'd do if we were building a metaphorical rocketship instead of an airplane.

We were rejected. ("…transparently, it’s still a little early even for [us].")

Well, what if we did it anyway? Max, bouncing between San Francisco and New York month-to-month, kept working. Could we turn our hackathon infrastructure into a daily challenge like an ultra-niche Wordle? Sure, we could make an MVP. Or, after a throwaway comment that he seized before it could float away, could we make a site that let you play poker against LLMs? He made enough of it that we couldn't not finish it, and so we did.

I think that a hackathon at Minifest is the first thing that we didn't say yes to—frankly, I just couldn't keep up—but we sketched out how to launch Hold’LLM (free to watch a few "featured" tables, or you could pay to run and customize your own). Max wrote a list of the tasks we needed to launch; I helped him finish it. I got distracted; he kept plugging away.

On Monday I learned that Max was dead.

He had felt fatigued traveling in Senegal, with something that felt like the a terrible flu. It got suddenly and dramatically worse as he made his way to Israel. It was actually a malaria infection that had reached his brain.

He was hospitalized on New Year's Day. And then he was gone.

(2)

When my grandfather died, I wanted to treat it as an encounter with Death, a meeting face-to-face with humanity's oldest and last enemy, one loss in the fight for a future where humans did not have to die when they weren't ready just because that was easier than staying alive. I wanted it to be part of a story that was bigger than one life ending forever because then at least it could make some sense.

Even then, though, I knew that that frame was for my benefit, because the actual Man-Hyong Yoo lived a story where death wasn't a thing we could end. To him, death was a thing that we postponed while we accomplished what we could with—and for—the people who mattered to us.

In that post, I quoted from Wilfred Owen's "The Next War":

Oh, Death was never enemy of ours!
We laughed at him, we leagued with him, old chum.
No soldier's paid to kick against His powers.
We laughed,—knowing that better men would come,
And greater wars: when each proud fighter brags
He wars on Death, for lives; not men, for flags.

Max Chiswick died of malaria.

Max wanted to end malaria. He worked on ending malaria.

No, he wasn't a doctor or a researcher, but he gave money to GiveWell because he believed that people not dying of malaria was a thing he wanted turn his resources toward—and wanted more than anything else he could buy with those dollars—and because as a player of games he thought that GiveWell's research was the best bet he could find to put his dollars on.

As I count these things, Max Chiswick warred on Death, for lives.

Max cared about doing it right. After he won the Manifest charity poker tournament and so could choose which charity would get $30,000, he asked me if there was anyone who I thought could turn the dollars into more lives than GiveWell.

Well, I love the work that GiveWell does, but I admitted that I have become even more excited about 1Day Sooner, and so we discussed my reasons. He went off to talk with their CEO to find out more, because if he could save more lives with those dollars, that was worth getting right.

On the same day that we met in my living room and decided we'd write a course on AI poker, we agreed to set the business up as a nonprofit, because he—we—were happy to take at most modest salaries and send any windfall profits on to the charities we thought could use it best.

Or rather, he had already made up his mind to do that with his share. He brought it up to see if I was on the same page and we could set up the company to work that way by design instead of as a for-profit.

(In the end, we never did try to make a profit in those first six months. There were a few points where we could have shipped something to sell, but each time we thought it would be better to work a bit more towards something much bigger.)

A friend, when they heard about Max's death, yelled in the kind of frustration that has nothing to point at but the whole world entire: "But I already wanted no one to die of malaria forever! I'm already doing the things it makes sense to do about it!"

Yeah, that's pretty much where I am, too.

The work that I put on hold every day that Max showed up with another Google Doc or feature idea or half a plan to put on an event was to make a world safe from infectious disease—as best I know how with the skills and funds and plans I have. And I don't have any better ideas for what to do about the problem except…what I was already doing before I met Max, that I put on hold for a day, a week, or a season at a time to work with him on our love of games. And so this can't exactly add to the amount that I want nobody to die of malaria, or what I'm going to do about it.

I don't know if this makes me feel better or worse.

(3)

You're not supposed to make sudden decisions at times like these, but I know that I won't be finishing the work we had planned to do together.

"I couldn't have done it without—" is a thing you're supposed to say to be gracious, but I know that I would not have stuck to the project week after week if Max didn't keep showing up.

I wasn't the best collaborator—I wasn't even a good collaborator—at the end. He would reach out three times to set up a coworking session, or a strategy call, or ask if I could fix a thing on the server that had taken everything down, before I would get back with a reply, because I wasn't managing my time well and because I was too stuck in a loop of guilt and shame and avoidance to open up a conversation about what to do about it.

Until two days after Christmas—five days before Max was taken to the hospital and didn't come out—when I screwed up the courage to write to him about it.

I had written it a few days earlier, but didn't want to send it on Christmas.

I don't know if he ever read it. Maybe I'll feel better if someone does:

Happy holidays! Sorry that I've been afk.

I've taken a bit of time to reflect on the projects that I have going on and, well, I have too many to be able to do them all well. I've been having a huge amount of fun with the time that I've spent on Overbet / Poker Camp / RPS things, and I'm really glad to have had that chance, but I think I just can't keep it up heading into 2025.

I do continue to think the work is very cool! If you decide it makes sense to keep working on things in the space, and want to keep me in the loop, I might be excited to make occasional contributions (especially on metagame-design / experience-design things). But in that world I think it makes sense to make a hard reset from an assumption of "two equal partners" to "Max, who leads the project, plus Ross, who might or might not sometimes contribute to things". […]

Obviously, this should also mean a complete reset on expectations of ownership (both economically and in terms of decision-making), starting from a default that all of it is yours and none of it is mine. I'm happy to do the work of making sure you have all of the stuff I've worked on, and catching you up so you can do with it what you want.

I feel bad to have done a kind of rush-forward-then-pull-back move here, but now that we're here, I think it would be even worse in the long term to keep trying to do things while pretending that I can be an equal partner. What I don't regret is, taking a chance on standing up the scrappy things we did—they've been energizing, fun, and I've learned a bunch from what we've tried (and shipped!).

I recognize this might leave you a bit in the lurch technically (to the extent that Claude isn't a sufficient partner for backend development!), and I also feel bad about that. Without implying an obligation to either side, I know that [a mutual friend] is working with part-time technical talent [...] and it could make sense for you to work with them in a partnership for your technical needs. I'd encourage you to have that conversation, if so!

Again, thanks for the experience of all we've done since June. It's been great, and I'm glad that when the opportunity came up we chose to jump on it instead of letting it pass us by.

There are times that you tell the world that you won't let this stop you, that you will pick up this banner and carry it forward in your friend's name—

—and there are times where you don't do that. Where it would be painful and lonely and hollow and just wouldn't make sense to go it alone. When all you can do is put it down, because it only ever made sense to do it together.

I am going to work on teaching through games again someday. But the projects that I was working on with Max made sense because we were working on them together, and I wouldn't do myself—or his memory—a service by carrying them on alone.

For now I've made sure that a few things that we made together are working, but I don't know that I'll keep them online forever.

(4)

What should I say about Max?

He was earnest and optimistic—unpretentious and generous—passionate and kind. He believed that we could just build things if we wanted to, and that it didn't need to be more complicated than that. I find it easy to get wrapped up in anxieties and doubts and paralysis when starting a new project, but when Max went first to break the ice, he just…started doing things. Programming. Organizing. Planning. He wrote first drafts, plans, and prototypes like breathing.

As far as I could tell, he ate a bag of carrots every day. If we took a break from working and a walk around the block together, we'd wander into the Trader Joe's; I wouldn't buy anything and he'd buy two bags of carrots. I mean, he ate plenty else, most of it intentionally planned, but you'd notice the carrots first.

I met him long after he had stopped playing poker professionally, and by then he didn't play it for the love of the game. But he cared about games, and he wanted people to understand games, like…like an art historian who rages at the impoverished experience of everyone who think that there's no point to representational art now that we have perfect photography.

He thought that games didn't become boring once they were "solved", because the world isn't made of abstract perfect opponents. And, for any game worth playing, you have to live and play in this world, not in the world of theory. Fortunately, the practice of playing games against imperfect opponents is so much more interesting and richer than the theory of Nash equilibria.

I believe these things about games, too. And it is said that friendship "is born at the moment when one man says to another, 'What! You too? I thought that no one but myself…'", so our friendship didn't take long at all to get off the ground.

He wanted the world to be better, and so in his un-anxious, Max-y way, went about doing what he could do about it. He had more money than he needed to make himself happy, and he gave it to charities that he thought could improve the lives of others. And he cared about his impact and not just his gestures, so GiveWell suited him well.

And, rare enough among those who care enough to take on the work of improving the world, he believed in taking on those little, in-effective projects that make your own heart sing. Max understood that hearts starve as well as bodies, and that living well meant doing what you love sometimes as well as what was important.

Even faster than he started things, he collected inspirations and ideas. He wanted to write "Bet Mitzvah, a set of materials that introduce betting and related concepts and the value that they can impart, especially as you become an adult." He wanted to fix certain shortcomings in the Pluribus paper (the first AI to beat serious human players at multiplayer poker), and wanted to write up his arguments and re-analysis. He wanted to make a site to track the quality of hotels' breakfast eggs. He wanted to ride a bicycle from LA to NYC eating supermarket rotisserie chicken every day, and made a site for tracking what Chabad houses he wanted to visit in the future (or for the countries you've visited, for the goyim among us).

He collected great URLs for his someday-projects with very little self-control. Just a few of them:

• poker.camp
• expectedvalue.org
• betmitzvah.com
• overbet.ai
• rps.bot
• bandits.band
• blotto.lol
• 1card.poker
• nerd.casino
• pokervschess.com
• kards4kids.co
• hoteleggs.com
• carrotmuseum.org
• chis.biz

I cannot stress enough that there are dozens more of these.

Max traveled. A friend of Max's writing at the substack Old Jewish Men:

In the month before he died, Max was in New York, San Francisco, Miami, Turkey, Senegal, Singapore and Israel. And, we can only assume, dozens of Whole Foods locations.

This, however, was typical. He was, somehow, in every country in the world while also in transit, chatting with you from 36,000 feet in the air. At one moment he was teaching probability and risk assessment to kids in Senegal, the next cruising a comped buffet in Vegas while also magically seeing you tomorrow for Shabbos dinner in Crown Heights.

The lifestyle left friends confused yet impressed by his ability to juggle time zones. Somehow Max always managed to be where you needed him. He lived everywhere and nowhere. But he wasn’t a ghost. He was the opposite: a constant presence in the lives of those he loved and who loved him. He could sleep on anything, and even preferred couches and floors...

He was always on the move, with a backpack full of organic produce, bike helmet in hand, a tennis racket on his back, zip-off cargo pants and laceless Salomon shoes; he was usually coming from or going to some obscure place for some bizarre reason that made perfect sense only to him.

Max never took a taxi or Uber if he could ride a bike there instead. He cycled from Cairo to Cape Town, walked the length of Israel, and hiked across Annapurna in Nepal.

Then there's things I've learned about Max in the past week. Back to OJM:

He knew nothing of art history, but loved art and collected it with a joyful passion. He became known in the New York art world, particularly through his connection to 56 Henry, a well-known Lower East Side gallery.

Friends who went to gallery openings with Max claim it was “like walking in with the king of England — everyone knew him.” But his tastes were his own; he didn’t pander. He was what one friend, an artist whose work Max collected, called “a deep diver who loved finding weird unknown artists like Paa Joe, Gaetano Pesce, lots of outsider artists and an obscure Hasidic artist named Pinny.”

He traveled endlessly, amassing a collection of nearly a thousand unique pieces, including works from Daniel Arnold, Al Freeman, Liorah Tchiprout, Ruby Bradford and Ohad Meromi...

Not at all surprising to me, however: "...His taste in art aligned with his taste in clothing and furniture — it was fun, and often carrot-themed."

Since we've gotten this far, I suppose we can take the time for a bit of standard obituary (again, thanks OJM):

Max Jordan Chiswick was born on December 1, 1985 to Peter and Ellen Chiswick in London, England. He grew up in Chicago and studied Electrical Engineering at Northwestern, graduating in 2008. He completed his masters [in algorithmic game theory] at the Technion in Haifa in 2017 before spending eighteen weeks at the Recurse Center, an independent educational institution for computer programmers, in 2019. It was during Chiswick’s days as an undergrad that he began his career as an online poker player. Not much is known about Max’s exact winnings, but he didn’t hold a regular job for at least a decade and a half following his retirement from online gambling in the mid 2000s. At the height of Max’s poker career, he was playing 100 hands a minute, and completed a challenge to play three million hands in a year.

Chiswick saw the writing on the wall when Artificial Intelligence entered online poker, and decided to exchange professional playing for teaching. He thought poker offered real-world tools that far exceeded the simple joys of casual gambling...

...and, well, that's where we started.

This is the picture of Max you'll find online most often, but I will forever be partial to this one instead, for obvious personal reasons:

From left to right, that's me, Max, and Nate Silver playing at the final table of the Manifest charity poker event. Nate and I were the second-and-third players to get knocked out (simultaneously); Max would go on to win.

(5)

There are moments in life where, if you’re paying attention, your future forks and you get a choice. Max and I sat in my living room in June and asked: "Are we going to try this crazy thing together?"

We had our respective well-laid plans; mine had no connection to teaching or games. We'd only just met each other. But there was the opportunity: to work on an audacious project that needed each of us to complement the other, on this subject that had both of us saying, "What! You too? I thought that no one but myself…"

I am so extraordinarily glad that we decided to go for it. For six-and-some months, I had the pleasure and privilege of working on one of my life's passions with an instant friend who was every bit as passionate about it as I am. I felt at the time that it was the kind of opportunity that doesn't come back around if you let it pass you by, but…I didn't expect that to be true like this.

Six months isn't a lot of time, on the scale of a life, but I am so grateful that I had the chance to share these months with Max Chiswick instead of letting the opportunity pass us by.

He was—more than most people are—important in the lives of those around him in many different ways. But in mine, the memories of our long walks and talks, our late nights spent hacking and planning for the future, and his quiet, daily example that making things you love doesn't have to be complicated, those memories will continue to be a blessing.