It’s 2010 and I’m reinventing the Matthew effect from first principles, using random sampling techniques, because in Monte Carlo I trust.
It’s 2017 and I’m working for Polymath as chief architect. I write in a Medium blog post, “What twitter did for celebrities who wanted a direct channel of communication with their fans, Polymath could do for startups who need a direct connection to a broad group of investors.” I still want that model to succeed.
It’s winter 1999 and my wife and I are living in Jessamyn West’s Vermont barn. Jessamyn lets us stay there for free, but we have to buy her old Saab for $500. It slips in and out of gear on the way to Barre for groceries. At the end of our stay, we leave it with a neighbor, Jack, who adds it to his collection of beat-up cars in a gravel-lined section of his yard. I assume it’s long since been parted off, or rusted into dust, but I have no idea.
It’s 2008 and I’m lying in bed, alone, at a resort in Roatan, Honduras. This is my first all-inclusive vacation. I spend hours at a time listening on loop to a Kirtan version of Shiva Shambho, piped right into my ears, while I alternate between trying not to think about the implosion of my first marriage, and thinking about nothing but that.
It’s 1999 and we’re driving a rented Mitsubishi Montero through the hills northeast of Cochabamba, Bolivia. Just outside the city a cholita (Quechua-speaking, indigenous woman) with a huge bag of onions flags us down along the main road, and gets into our back seat without a word. She asks to “bajar” a dozen km later. On the way out, she hands us a peso, worth about 15 US cents.
It’s 2018 and users are getting “deplatformed” from social media websites right and (occasionally) left for having unacceptable views. By then I’ve already given years of thought to how you build a system of decentralized, censorship-resistant identity, where your network of friends and followers can be separated from the walled garden in which it lives. I come out with a white paper explaining how that might work, and call the project called “BeforeTheBan,” which to this day I think had one of the coolest logo ideas ever (though execution was sketchy).
It’s 2006 and I’m deep in thought about how sometimes, the longer you wait for something, the longer you are likely to have to wait. I compare this to exponentially distributed wait times, in which, no matter how long you wait, you never get any smarter about when the event will happen. But you don’t get any dumber, either. Would it be possible, I wonder, to build a distribution where the longer you wait, the less you know? “In other words,” I write, “can we take our level of uncertainty up to 11?” The Unreliable Friend probability distribution is born. After that, I go on to create an even more diabolical distribution, the FML.
It’s 1993 and I’m standing in a double line of boxers who are paired off, waiting for their matches. This is my first fight, after a couple years of training off-and-on at Windy City Boxing, where I got to watch Polish heavyweight Andrew Golota spar, the man who could have been champion if only he was able to keep his punches above the belt. In my own match, I enter the ring to the sound of cheers. For my opponent. He’s a golden boy with custom-made shirt and shorts and a six-and-oh record. He beats me soundly, but I stay upright and get in a few good punches.
It’s 2005 and I’m reading an article in Outside Magazine online after someone sent me the link. It tells the story of an attempt to bring back the body of a scuba diver from the bottom of Bushman’s Hole, the deepest underwater cave in South Africa. I end up reading it several times over the years, and in 2021, while living in the Florida Keys, I interview author Tim Zimmermann about it for my radio show.
It’s 2022, and I write an obituary of sorts for Rush Limbaugh, noting that he was “the Purdue Pharma of copium, the undisputed king of making losses go down easier.” Years earlier, in 2013, I write about how an argument about Rush, the band, and its drummer, turned sour.
It’s 1997 and I’m wandering around Port Townsend, Washington in a daze while my mother struggles with cancer 2000 miles away.
It’s 1996, or maybe 1997, and I’ve just very self-consciously picked a bunch of wavy-capped P. cyanescens mushrooms from a public park near Seattle. The trip I go on is deeper than anything I’d experience before, or since. I realize the extent to which our perception of the world is run through linguistic filters we ourselves construct, an observation that is at once banal, but also continually forgotten. Language, of course, is one of the filters. Strangely, or perhaps not, artificial intelligence has now come in the form of models that operate entirely on the level of sequences of mathematically transformed words. The filter is eating our world.
It’s 2012 and we’re at the zenith of “global warming will kill us all” messaging. I decide to do my own, from scratch, analysis using NASA’s data. The resulting article, which compares temperature changes to a random walk, results in red hot firestorm of comments. My take today, in 2026: The temperature data we use is basically worthless, and whatever the underlying trend is, and the extent to which humans are behind it, we won’t be able to tell based on that data. Garbage in, garbage out.
It’s 2021, and I’ve just finished filming the final episode The Mattasher Show, a long-form interview program that aired on PBS. I jump into our pool with my clothes on, exhausted, completely satisfied with the season we’d just wrapped up. The marketplace, sadly, gives the show a lukewarm reception. Who knew discussions with graphic novelists, philosophers of science, ex-congresscritters, and NSA agents who went undercover in the bigfoot scene wouldn’t be a hit?
It’s 1996 and my girlfriend and I are on a road trip down the Pacific Coast Highway on a 1981 Honda CM400 with an oil leak that’s solved temporarily by welding the entire chamber shut. As part of a magazine article I’m working on related to an anniversary of Easy Rider, everywhere I go, I ask people how free they feel. I write, “Around noon, I spotted two landscapers taking a break. I introduced myself to the young men. One was light brown with a curly beard, the other pitch black.” When I ask them how free they felt, they responded emphatically, “Not free.” Turns out they were wards of the state, doing supervised work detail.
It’s 2019 and I’m holding the hand of a stepson who has already died, but whose body is being artificially kept alive to use for transplants.
It’s 1998 and I’ve begun to play around with building websites. As a fun little experiment, I create lyca.com to complement the experimental work of fiction I’ve already been working on for 7 years. The site remains almost exactly as it was up until this day. The current status of the cat is unknown.
It’s late 2020 and my wife and I are in Stockholm, taking a break from Ontario’s ongoing lockdown madness. The food is amazing. While there, I interview Patrick French about his biography of legendary, but largely forgotten, explorer Francis E. Younghusband. I call Younghusband’s adventures “military tourism,” a description French largely agrees with.
It’s 2026 and I’m cleaning up after statistician Nate Silver, who I once, entirely accurately, described as the Tim Tebow of political prognosticators. Silver is trying, and failing, to explain why small countries can field highly competitive World Cup soccer teams. I write, in a Hacker News comment about his article, “The answer is simple once you understand that for thin-tailed distributions, the mean is way more important than population size for getting extreme results. In concrete terms, suppose that to win the Olympics you need 5-sigma players (ones who are 5-standard-deviations better than the global average). Five-sigma players are extremely rare: a population of 100 million gets you about 25 to 30 of them. But now suppose you could bump up the quality of your soccer players until the average among them was raised just one standard deviation above the global mean. Now you only need a population of 1 million to generate the same number of five-sigma players. The end result: a tiny country of fanatics can compete against a huge country with tons of casual players, like the US.”
It’s 2020 and my podcast, The Filter, is off to a good start. My audience is growing. My guests are happy. Russ Roberts tells me “Great Job!,” Michael Shermer says “This was a good conversation,” Katie Herzog thinks I say something that is “kind of genius,” and Sandra Tsing Loh, OG podcast goddess, who helped make This American Life, This American Life, says “I love your show and everyone should listen.” Meanwhile, super high-functioning stats autist Andrew Gelman tells me, “I don’t know what to think about that.”
It’s 2013 and my brain’s ability to focus hasn’t yet come under assault from TikTok and I’m still reading lots of full books, and posting reviews to then-independent website Goodreads. These include my full-throated defense of Twilight, in which I write, “Meyer is a first-rate storyteller, as we see with her story-within-the-story about the legendary birth of werewolves, and in her ability to capture the reader’s attention despite how irritating, unpleasant, and downright tiresome the main characters can be.”
It’s 2026 and I’m finding the LLM era amazing, but the tools for agentic communication are still primitive, so like many others I’m rolling my own. Unlike others, mine is quite good, as it allows you to do threaded messaging with agent instances across multiple harnesses and models, including LLMs, in an intuitive way. I won’t go into all the details here, as this may or may not be an internal tool I spin out, but to see threaded conversations done right you can check out a demo I did for Hacker News a decade ago, which I still use every day to browse that site.