Four Years of Three Things

Three Things #203: April 29, 2026

It’s been a great four years. (Nearly four and a half now!)

Note: like Issue #202, this article was also written back in December but didn’t get published due to changing workflows and priorities. I’m still prioritizing building over writing, but I’m continuing to catch up here. As before I’ve mostly left the original article intact, and added a few annotations.

--

I wrote reflection pieces at the end of years one and two of Three Things, and something resembling a reflection piece here a year ago [note: nearly a year and a half ago now!] at the end of year three. I didn’t intend to write a reflection piece this year, but, well, upon reflection I feel that I probably should. Reflection is, after all, extraordinarily important: if we don’t pause from time to time to consider where we’re coming from and how it’s going, we’re flying blind as we move forward while the world continues to change all around us.

Four years is a long time, even for an old man like me. There aren’t many other things I’ve done every day for as long as four years through thick and thin, good and bad, happy and sad times. Writing here has been my release, my celebration, my way of understanding things, and my place to come and reflect when things aren’t going so well. It’s cathartic, and it’s become a part of who I am. I’ve considered changing the format many times over the years, or stopping writing entirely. I still reserve the right to do that (especially the former), but it hasn’t yet come to that. [Actually, it’s begun to change a bit thanks to the influence of AI, but less than you might think.]

Given the length of time since I began writing here, I thought it would be fun to look at what’s changed, and what hasn’t, over these four years.

Thing #1: What’s Changed 💫

Four years is an interesting length of time. On the face of it, it feels like a long time. College takes four years and that felt like an eternity. It’s long enough for things to change in some meaningful ways. The most obvious examples, for me and for the world, respectively, are that I wasn’t a parent four years ago, and the proliferation and improvement of AI tools. Those are big changes. But in the grand scheme of things four years also isn’t a very long time, and not much big change tends to play out over just a few years, even if change is accelerating as it sometimes feels like it is.

Let’s start with the obvious: the rise of AI tools, which didn’t meaningfully exist at all four years ago when I began writing here. Three Things coincidentally happens to track the rise of AI tools almost perfectly these past few years. In the beginning, everything was done manually. Then in late 2022 I began to experiment with using image models to generate header images. I continued to experiment with these and other AI tools as they emerged, including briefly using ChatGPT to review all of my posts. Over the past few months, I experimented with AI drafting, co-writing, and inline commenting and reviewing.

I’ve done my best to stay at the cutting edge by experimenting with as many new tools as possible, and I’ve incorporated them into every corner of my life including here: as Ethan Mollick likes to say, no matter what I do, I invite the AI tools along.

They’ve had a huge impact on my work in general. I’m able to explore ideas, design, and build things much, much faster than I ever could before. On a good day, I feel like a one man army: I can accomplish more than a small team ever could before. It’s difficult to talk about the potential of AI without sounding hyperbolic and cliche, but even just looking at the ways AI has already had an impact on my life and the things it’s already enabled me to do, the change feels profound.

Maybe because I’m a software developer, maybe because my life is already quite digital, maybe because I’m an early adopter—whatever the reason, I can say with confidence that AI has already brought a lot of positive change to my life, although I realize this won’t be the same for everyone. Nevertheless, even ignoring everything else these models and tools are capable of, simply increasing the breadth and depth of human knowledge, and the efficiency of access to it, is going to change the world in the same way that older technologies like encyclopedias, telegraph and telephone, fax machines, the Internet, the web, and mobile phones did before.

But the impact on writing so far has been limited. After lots of on and off experimentation I did finally, fully switch to using AI tools to generate header images for these posts (I used to rely on Midjourney, but I’ve switched to API-driven models like GPT-image-2 and FLUX.2). It’s strictly better than using stock images, which I did for several years before that. (I feel bad for the stock image industry. I can’t imagine that it’s not already dead or quickly dying.) And the AI tools are of course very helpful for fact checking, generating references, etc. These are useful, but they’re not massive gains relative to the overall writing process.

Other than image generation, I pretty rapidly abandoned the other experiments I mentioned above, such as running all of my posts through ChatGPT. I tried asking the AI to emulate my writing style and make improvements to syntax, diction, structuring, etc. I did this for a month or two, then gave up. For one thing, it was too much work, and I found that it interrupted my flow: comparing the output side by side with the original and selectively copying over changes I like, checking for mistakes, etc. I’m not willing to copy-and-paste whole hog for several reasons, one being that, try as it might, these tools aren’t yet capable of capturing my tone of voice or writing style. The pieces that I used AI to edit just felt less authentic. [Recently I’ve continued the experiment, giving my agents access to comment on and suggest changes to my pieces, and while I’m amazed every day how good they are at coding and many other things, I’m still pretty unimpressed with their contribution to my writing so far.]

I’ve also tried using AI to help generate ideas for what to write about, but I’m generally pretty disappointed by the results, and in general this isn’t a big problem for me anyway. I’ll continue to experiment, and I remain curious how other writers are using AI tools to augment their writing.

Speaking of writing topics, what I’ve been writing about here has evolved gradually. In addition to the perennial topics—productivity, health, parenting, travel, geopolitics, tech, crypto—I’m obviously now writing a lot more about AI and about my experiments with it. I wrote a lot more about politics in the earlier years, but I find that I’m writing less about it lately.

In fact, I’m also thinking less about politics these days. While I used to follow world news and current events religiously, reading major newspapers every day and The Economist every week to stay on top of developments globally, I’ve gradually scaled this back over the past year or two. I find that, news being what it is today, it just increases anxiety and offers very little useful information on a day to day basis. I do think it’s important to know what’s going on on a high level, but I’ve found that this information tends to find me in various social ways even when I don’t try. [Note: this is part of my motivation for the Beacon project, which I wrote about last week. There are better, more efficient, less clickbaity ways to transmit this information.]

Thing #2: What Hasn’t Changed ⚖️

It feels like surprisingly little else has changed over the past few years. And what has changed is less than what I hoped would’ve changed by now.

I’ve written a lot over the years about the rise and the importance of blockchain and cryptocurrency, especially insofar as they enable decentralized, community-driven governance of networks of value creation and exchange. It’s true that these ideas and technologies are more mainstream than they were four years ago, and they’re more widely known and accepted, but they’ve only just begin to move the needle socially and politically outside our little niche. Unfortunately they’ve been drowning in bad press thanks to lots of bad actors.

To the extent that crypto has begun to have an impact on the way humans collaborate, work together, solve problems together, allocate resources, etc., those changes are largely still confined to narrow on chain use cases, primarily DeFi. That’s not to sell DeFi short: there are hundreds of millions of dollars captured on chain in these ecosystems. That’s a lot of value to a lot of people. Stablecoins, cross-border flows, crypto payment rails, and real-world assets are all promising and continue to grow even as the broader industry stagnates, but this is still just a drop in the ocean of all of the financial assets and resources in the world (and, to put things in perspective, it’s also a rounding error in AI terms).

To be frank, we should be further along than we are, and the fact that we aren’t demonstrates how immature these ideas and these technologies are, as well as the extremely negative impact impact of scammers and other bad actors over the years: the Do Kwons, SBFs, Hayden Davises and Donald Trumps.

Even on the micro, personal level, change has been slow. A few years ago I was regularly paying my friends back for lunch on chain in ETH and BTC. Today, despite the rise of stable coins, that rarely happens. As an industry, we seem to have collectively given up on the original idea of cryptocurrency as a form of “peer to peer electronic cash.” What’s more, our desire five or six years ago to dogfood cryptocurrency for small peer to peer payments spoke volumes about our optimism and sense of excitement for the new technology that we were helping build and make real. I sort of thought I’d be done with bank accounts by now; the reality is that today I unfortunately rely on them more than ever before.

A lot of that naive, early, OG-flavored optimism and excitement has faded, and it’s given way to something else that I can’t quite put my finger on: a combination of frustration (that things are taking so long, despite crypto being so much better than the existing financial system at so many things), exhaustion (from repeated, intense market cycles and being under attack for so many years), realism (accepting that the use cases are far more limited than we originally thought, but still important and powerful), and a healthy dose of remaining optimism mostly focused on institutional use cases and the areas I mentioned above.

Then, there’s software more generally. Another theme here over the years—because I see it as my life’s work—is making software more humanistic. In other words, building software that truly works on behalf of its users, not simply on behalf of big tech companies that advertise to us and monetize our data. That’s a long story arc that’s also going to take a generation to fully play out.

I wish I could tell you that, four years later, the software landscape has changed meaningfully. I wish I could tell you that I’ve finished my data hygiene review (something I’ve been working on for six years!) and found a suite of apps, tools, and infrastructure that respects my desire for sovereignty and privacy—more apps like Logseq, a fantastic open source, open data tool that I use every day for taking and organizing notes. I wish I could tell you that, even if we’re not there yet, we’re at least moving in the right direction with more open source, more open governance, more opportunity for community participation in value creation and capture, etc.

Unfortunately this, too, hasn’t changed much. Unlike some friends and colleagues I’m not a purist: I typically use the best tools for the job, and today the best tools for almost every job are closed source, closed data applications run by for-profit private companies. To some extent this is healthy, and unavoidable: for the same reason that DAOs are bad at products (you can’t build good products by committee), small companies are typically best at them. But this doesn’t mean that software can’t be operated in a more community-driven fashion, and that the global community of users can’t contribute to the governance of software, data, and networks, and participate in the upside. I’ve seen vanishingly few examples of this done well so far, but I haven’t completely given up on the idea. The power of AI to reshape software gives me hope that things will improve soon.

Thing #3: What Needs to Change 🆘

This is a difficult question to answer. I’ll approach it on three levels: personal, professional, and global.

Personally, my mantra for the new year is balance. I’m quite an extreme person, and I tend to take things to the extreme. I did many things this past year: parenting, working, and exercising, to name the big ones. The other big one was travel. Some of that travel, maybe a lot of it, was appropriate and necessary, but some of it probably wasn’t. I spent countless hours training for, traveling to, and running the New York Marathon in November. It was a great experience, one I’ll never forget, but it was an extreme one, on top of everything else. I was surprised at the end of the year to see how many hours—days and weeks, in fact—I spent on airplanes last year. Like everything taken to extremes, all that travel was interesting, valuable, and educational, but that chapter is now behind me and it’s time to start a new chapter: one that’s a bit more balanced and sustainable.

I really like to push things to the limits, in order to know where both the objective limits and my own limits are. One of the limits I discovered this year is how healthy you can be while constantly traveling: in my case, the answer is reasonably healthy, but not even close to my peak. [Note: I haven’t traveled in months and I’m doing much better on literally every objective and subjective metric.] I also can’t be a good parent when I’m constantly on the road away from my son, even if I visit home in between trips. And I definitely can’t be fully effective at work when I’m on the road a lot, even for work trips.

So the goal for the new year is to balance all of these things, without going to extremes: first and foremost, be a good parent and partner. Secondly, be a responsible leader to my team, and build great products that delight my customers. This definitely includes some degree of travel in order to have face time and build trust, as well as to participate in conferences, but most of the time it means doing the hard thing and staying put, remaining focused on shipping. Thirdly, stay fit and healthy. Improve my diet and sleep. Exercise every day. Continue trying to balance strength, endurance, and flexibility: running, strength training, and maybe yoga. Maybe run a race or two. But not necessarily on the other side of the planet.

Let’s zoom out a little. Professionally, I think our goal as an industry should be to hit pause and reset a bit. 2025 was a pivotal year for crypto: in spite of a sudden favorable regulatory regime, nearly every token in nearly every category was decimated. It seems like the average token is down around 97%. Lots of folks who have been in the industry as long as I have or longer are feeling burnt out and jaded lately. [Well, some things at least haven’t changed the past few months.]

That may sound depressing, but actually it’s healthy. It’s a perfect opportunity to take a step back and remind ourselves why we started. It’s an opportunity to look in the mirror and ask ourselves the difficult questions: what are we building, for whom, and why? To what extent do the things we’ve built, or are building today, reflect our values? How can we be better? These are the questions I’ll be asking myself and my team in the days and weeks to come.

Finally, on the global level as well, we need to have a similar conversation but with a different focus. We need to look at what we, as a society, are building—what tools, technologies, and institutions—and ask exactly the same set of questions. What are we building, for whom, and why? These questions are relevant to politics, which is extra messy these days by historical standards. Politics today seems to be more about violating norms and tearing down cherished institutions more than anything else. This isn’t necessarily a bad thing, but it’s essential that we reflect on why we need to do these things—and respect Chesterton’s fence: in other words, seek to understand the systems we replace before attempting to replace them.

These questions are especially pressing in an age of AI, as we contemplate what tools to build, how they’re going to impact society, and how to avoid the very worst outcomes. We’re racing to build bigger, better, faster, and more capable AI tools. This is a worthwhile pursuit because it will usher in a new era of prosperity that will, on the whole, benefit everyone. I believe that very strongly, and we’re beginning to see that play out today. But we need to have a more honest conversation about who the winners and losers in this system will be, and how to make the new world work well for as many people as possible. That’s another big question that’s on my mind this year.