I started writing code when I was eleven. Born to Belarussian immigrants in the New York area who were both software engineers for a living, writing code wasn't considered particularly unique or meaningful. In fact, my parents all but discouraged me from doing it for a living. I can still hear the staccato in my father's voice: "Everything in the world is sales. You need to learn sales." Engineers were back office workhorses in 1990s New York - the glamorized image of the inventor engineer hailed all over Silicon Valley was nowhere to be found.
It would not be until my 20s that I would make software my profession, but the purist principles of software engineering and its promise to the world were instilled deeply in me as a young child. Like many, I was grappling with my first meaningfully existential thoughts as I grew - finding my moral compass and hashing out the rough universal ideology that would set the foundation for my perspective on the world. Software taught me systems thinking. Its scale and rapid ongoing permeance into our everyday lives taught me the impact of building good systems. The unfettered internet and the access it gave me to learn of world events at increasingly unprecedented speed taught me all the risks that came with building bad systems, and helped me develop my moral compass around defining, contributing, and protecting the good in a world of much bad. I developed my views on justice; software taught me how to think about breaking bad systems in favor of better ones. I was hacking before I knew the term.
It was through software that I formed my globalist view of the world. It was international collect calls that became the bridge for my family across borders and timezones before VoIP was widely accessible, and it was IRC chats and esoteric forums and Runescape lobbies that introduced me to entire worlds of people like me as I was chasing my youthful exploration from the confines of my family computer room.
It was software that brought about my fascination with the machines, and ultimately, my ideology around human-machine coexistence. Software felt like magic, and creating with it was like harnessing the wand. My college years brought me to machine learning - building models and watching them grow felt like birthing beings. I watched Ex Machina gripped by its imminent realism and stayed up late binge-reading about non-orientable surfaces and quantum theory with my friends. I grappled with marrying the temptation to chase theory and academia with the unfortunate reality that I needed stable income and a path to supporting myself as the oldest child from a family that wasn't graced with the privilege of affording me an endless pursuit in my passions. I chose the latter and went to Wall Street, putting aside my hacker identity and settling myself into the great Goldman Sachs machine. To my very much unexpected but nevertheless pleasant surprise, I learned a lot at GS that shaped my views on organizational execution and finance at great scale - a deep dive better saved for future writings.
San Francisco and the promise of a better internet
Eventually, it was software that brought me to San Francisco. It was in San Francisco that I grounded my modern philosophies on technology and started shaping my modern views on entrepreneurship. It was in San Francisco that I joined Uber and saw distributed operations on an international scale first-hand for the first time. It was at Uber that I really felt what it was like when software made an impact. It was at Uber when I was first introduced to crypto and its promise of building a better internet.
Bitcoin emerged out of the Great Recession, but much of today's modern blockchain infrastructure is built on principles that emerged out of watching Silicon Valley make concessions on user data and digital asset ownership in the 2010s. By the time I had meaningfully joined the conversation in 2017, the pitch for crypto was incredibly enamoring for a cyberpunk. Ethereum was driving the exploration of Turing-complete engineering on chain, and the pitch was simple:
The internet was one big black box, controlled in its access and in its truth by great powers that be with little accountability. What I had grown up envisioning as a portal to freedom was, in fact, only rented to us as such. What if we could fix that?
There were more ideas flying around than I could ever recount - every company in the Valley had a blockchain strategy. Most of it was smoke, of course, but at the time I was much too idealistic to internalize that. I joined Uber's new blockchain club, found my way to meetups and hackathons across the city, and found my way to coffee chats with baby startups you would recognize as behemoths today. I tore through Tarun's 5-article manifesto for launching Gauntlet and spent late nights at the office trying to get Truffle to work while brainstorming machine learning and blockchain intersections with coworkers (it was, after all, so much data).
I would ultimately decide that blockchain infrastructure wasn't yet performant enough to build on in 2017, but its ideologies stuck with me - here was a chance to rebuild the rails of the internet.
2021
If you were to imagine a better infrastructure than Solana to soothe my concerns about blockchain's performance at scale, you'd be hard pressed to do so. An old friend introduced me to Solana - the tagline was "fast and cheap". It was the gap I'd been missing. I found myself an invite to a Solana brunch at Bitcoin Miami and ate breakfast while listening to Austin Federa entertain our group with a fun (theoretical and entirely impractical, but economically intriguing) DeFi primitive. The rest of my summer was spent selling a startup I'd been building, as quickly as possible so that I could build on Solana.
In the fall of 2021, I started building mrgn. It was called Seaworthy first, then GM (Global Margin) Protocol (briefly; the "gm" trend had just kicked off and I was still new to catching trends), then marginfi. I like to believe we went with the best name in the end. When I was building my first startup, my biggest concern had been what I imagine most first-time entrepreneurs are concerned with - building something that works. It had worked well enough (that sale ended up going through), but I had done a poor job of solving for founder-market fit and picking a business that I actually enjoyed. After a lifetime of growing up under the immigrant survive-at-all-costs mentality, it was not until my late twenties that I really afforded myself the privilege of acting on something I was passionate about.
mrgn was a startup I built with my whole heart. Solana was the better internet I had been looking for, built on a better system, and an infrastructure to build better systems on top of. The proposition could not have felt more important. At its best, finance underpins our modern social existence - it sits next to few things at the top of the list of the most important primitives to get right and build in ways that protect the everyday person. We grew with the new internet emerging out of Solana, niche but powerfully promising in the unprecedented access, fairness, and opportunity it gave people all around the world. And then I left mrgn, and I still don't talk about that.
2025
The past year has been one of a consuming search for purpose. My time in crypto has afforded me more privilege than I could ever express – it's taken me around the world and I've had the opportunity to work with some really incredible people (even while searching for myself this past year). But my time in crypto – and perhaps just my time so deeply behind the veil in industry overall – has shown me heavy underbellies. I've met people whose relationship with money has fundamentally corrupted them. I've seen what greed looks like when it's fettered only by technological boundaries and at a scale beyond what most people get to see in their lifetime. I've seen short-term thinking overtake sentiment en masse and create case studies for a human society's capacity to self-destruct.
The juxtaposition is stark and maddening: a futuristic infrastructure with a very real and very stunning amount of potential to improve humanity, swimming in the oldest human shortcomings in the book. As an entrepreneur you're left with this dark question: what is this all for? In the self-help books we read as founders, we're taught go build products that solve people's problems. YC tells you to build what people want. Is this disheartening underbelly just the harsh reality of what people want? Was I just being idealistic (again)?
The Accidental Experiment
The less (directly) depressing way to frame the myopic behavior we've all seen way too much of in crypto is to argue that what we've been observing in our corner of the internet is an interest in entertainment above all else, which stems from an increasingly post-scarcity living experience. If the world is moving progressively towards post-scarcity, and your job as an entrepreneur is to build what people want, and what people want most is entertainment, then perhaps the entrepreneur's role in 2025 is that of an entertainer. I've spent a lot of my time this year on entertainment - Dark effectively came out of this. The accidental experiment that launched Dark goes like this:
I deployed a test coin on a token launchpad in December of last year, to play with a livestreaming feature you needed a token for . In April, it was discovered on the internet and unexpectedly went viral. I hadn't planned for it and I didn't have a strategy, but suddenly I had distribution - and in a world where entertainment is more important than anything, distribution reigns king. I decided there was room for an experiment in entertainment-driven product building and launched an effort underway (a month later, token launchpads started allowing token creators to earn trading fees off of tokens they created - this would be dubbed the "Internet Capital Markets" (ICM) meta and many other founders would start running a similar experiment). I paired my structural view on the ICM opportunity with an old love that I had come back to working deeply with again: AI (fka machine learning, fka statistics). I had been exploring a variety of problem spaces on that front - devops solutions for swarm systems at scale, model context protocol infrastructure - and aligning two experiments into one felt reasonable.
Crypto's myopic nature immediately hit me like a brick wall, but at least by this point I was prepared for the brick wall. Building in crypto with a token you didn't raise VC money on means facing a particular kind of cynicism: everyone - including your friends and all of your supporters - asks you daily if you're going to abandon the project. It's insane, actually - and incredibly unhealthy, even if you're used to it. The expectation of betrayal has become so normalized that it creates its own reality - everyone expects everything to be a rug, so they act as if everything's a rug, and then the likelihood of rugs increases. I considered leaving crypto, but not because of the abandonment trauma. Crypto's matured significantly and it's felt like the frontier has settled - you had your Bitcoin inflection in 2011, Ethereum in 2016, and then Solana in 2021. In early 2025, it felt like the only interesting work left was on the institutional side (which is important, don't get me wrong, but) - it was less clear where the opportunity was for early stage startups. The vibes in AI felt better - I started thinking about moving back to SF.
—
Crypto's tough to leave though. I've built my distribution here, and even more importantly, the userbase is high paying and very responsive - a generally good combination if you're looking to iterate towards product-market fit and build something that makes money (good startup goal). So we stayed and kept experimenting in crypto, working with users and orientating over iterations on a consumer-facing trading product that lived at a crypto and AI intersection. Like every other engineer, I have a soft spot for brand new tech - so while we kept our focus tight on building for our users, we also started experimenting with brand new solutions on both the crypto infrastructure side and the AI side of our product. Payments and wallet infrastructure seemed to pick up velocity over the summer, so we leaned into that. Raw LLM performance had been plateauing but scaffolding around LLMs in the modern AI experience was accelerating, so we leaned into that. As we built, a common thread emerged from our users' feedback: the unlocks (or the lack thereof) in our product experience was coming from the level of our investment in design and UX. This is something I was used to in crypto - crypto products are notoriously difficult to use, and there's a culture in the industry of deprioritizing user experiences that would appeal to the mainstream in service of experiences that appeal to the niche insider community (despite extensive conversation about onboarding the "next billion users"). In AI, conversation seemed a lot more focused on model performance than anything else, although a narrative thread was emerging about the lackluster adoption of AI products newly launched this year and the increasing focus on solving UX issues that live around these models.
And then I realized that crypto and AI are two industries with the same fundamental problem.
Dark: The human company
The less interesting likeness, in this case, that crypto and AI have as industries is that they're both massive technological breakthroughs and, I think undoubtedly, are going to change the way we all interact with the universe around us through our software. This part of their nature stands on its own merit, but it's irrelevant in the context of the value in building products at their intersection (I think this is why "crypto x AI" sounds like such a scam - it attracts people who just want to build vaporware with two cool words put together). The more interesting likeness is that they're both now at stable backend technological checkpoints - the blockchains run (mostly) and the AI models know things (sort of) - but both suffer limited adoption prospects from a lack of UX scaffolding around their core technologies. To be clear, the UX scaffolding problem are not without their engineering work either. Smaller focused models, memory solutions, efficient RL are important UX solutions as much as they are interesting engineering. Neobank infrastructure and on-chain policy engines that more robustly support safe on-chain agent executions are as much UX solutions as they are interesting engineering. But the core of my point is that the driving principle for adoption in the next leg of winning products is designing for the user experience, not for baseline functionality.
Crypto and AI are alike in another way: they're both industries with cultures that are heavily biased towards backend functionality-oriented work and not UX. They're made up of companies that are made up of people that are backend-minded and not UX-focused - the industries lack talent in line with the product quality lacking across them.
This is the opportunity: to pour world-class design into building products that leverage the transformational infrastructure already built in these industries, win users, and ultimately make something for people that feels impactful to them. To leverage crypto infrastructure for building efficient, accessible financial solutions that make it faster and cheaper for people to do everything they do with their money. To give people around the world comfort that their money is safe, protected. To build beautifully intuitive interaction experiences on top of foundational AI models (and imminently, on top of specialized models optimized for their roles in products) so that people more comfortable and more empowered to work with the products they're using.
I'd be lying if I said I don't think about credentials regularly. I generally think much of traditional academia is a waste and I have a whole philosophy I'll save for future writings, but it's hard to contemplate starting an AI company without noticing the lack of a Stanford PhD on your wall. Even outside of my AI work, I think about my lifelong bias towards industry and the carefree tending to my zen garden that I imagine going back into academia would feel like. Putting subjective views on formal credentials aside, I think you need a qualified team in some form if you're going to succeed at anything - and I think succeeding at (real) crypto x AI products in an era where consumer UX is, at a high level, the next breakthrough, requires a particularly unique combination of talent and culture. For their popularity, both crypto and AI are industries where the engineering is still relatively niche. You need engineering talent that's spent time in both before, understands the foundations, and can keep up with ongoing evolution. You need design talent that understands the traditional design challenges, the new opportunities these breakthrough technologies afford, and the limitations they come with. And above all, you need a strongly tight-knit culture that puts everyday users at the forefront of your work in industries that don't curate particularly favorable environments for thinking outside of their bubbles. I don't think there are a lot of teams that have a shot at fitting this criteria, but I think we're one of the few.
If we succeed, I think we can create something genuinely meaningful - something that makes a small dent in the universe in the right direction, even in the midst of entertainment-driven economies and post-scarcity lackluster living. We'll build products - starting with Scout, our interactive consumer finance app - and we'll build technology, and we'll open source what might help teams around us, and we'll share our learnings, and we'll see what happens. It's life's greatest privilege to chase your passions through your work, and I'm determined not to waste the opportunity.
That's why we're building Dark.