Whoami
I'm Chris Caruso. I've been interested in computing for longer than I can remember. There's a photo of me at a Macintosh, maybe four years old. I had my first laptop before most people had a cell phone, probably before I had one.
I'm a builder at heart. I like to want things that don't exist, decide to build them, and then build them. Software is just the medium with the highest return I've found: the barrier to entry is low and the possibility space is enormous. I've been surfing that wave since I was thirteen, when I taught myself C# in middle school. I don't think I'll stop. Less than two hours after major heart surgery I was back on my laptop in the ICU, building. I have a picture of that too.
I'm self-taught, and looking back I think the real thing I learned was how to learn. I never followed a book. I had something I wanted to build and I went depth-first: variables, then functions, then classes, then programs, and finally the shape of the experience I was after. The wanting came first. The skill was just whatever the wanting required.
Programming turned out to be a means to an end. For years I thought the craft was the point. It was a detour, a beautiful one, but a detour. The ideas were always the distance. That has never been more obvious than now, when the code itself is becoming optional.
Today I'm a Principal Architect at Microsoft, where I work on Rayfin, our backend-as-a-service. Most of the industry would call my role a staff or principal engineer. The shorthand I like best is that I manage the tech, not the people. I came up through Power BI, most recently Copilot in Power BI.
The AI thread
I've worked with AI for most of my life, long before it was a wave. It started in college with character-level RNNs, including one I trained on Trump and Hillary tweets during the 2016 race and called Trilary. What hooked me was that the same idea could learn language, vision, or control, whatever you pointed it at. My favorite project from those years was GCNet, which watched GIFs and streamed back captions one word at a time, an autoregressive multimodal model years before that was a normal phrase. Between then and now there were models for nearly every hackathon I could find, including a cafeteria garbage sorter! I set my GitHub bio somewhere in there and never changed it: "I'm fascinated with Neural Networks. They are the future of everything."
I've tried to stay ahead of it ever since. I was part of the first generation of AI engineers, the ones building before there was a bandwagon, before GPT, before the Transformer. I independently built function calling, agentic use, and orchestration before they had names or standards, inside constraints that look quaint now. I thought about joining OpenAI in 2016, but got entangled with life instead, and still wonder what that timeline looks like if I had. Either way, I'm a builder who sees code as the means and AI as the ultimate builder: humanity's last invention, and our greatest achievement in a very old process we are lucky to be part of. The longer version of that origin story lives in The Fortunate Fall.
The architect
I started as a junior engineer and became an architect without meaning to. In an early interview I asked someone what an architect actually does, because the title sounded mysterious. He said it's something you just become. I didn't realize I was already on the path.
I'm a visual thinker. Once the details of code became second nature, I moved up to systems, and I've had a whiteboard following me since high school. The people I work with know me as the diagram person. I live in FigJam as a shared whiteboard, distilling complicated systems into simple pictures that build a common understanding better than any wall of text. A good diagram is a campfire. Everyone looks at the same object, reaches in, and reshapes it. The problem becomes fungible from every perspective.
I have strong opinions about the job, and they mostly come down to one rule: don't be the oracle. If you're the oracle, you become the bottleneck and the place where the buck stops, and you never leave room for anyone else to grow. So I usually set the structure and let the people doing the work do the solving. I lead them there. They go further because the solution becomes theirs. My code review comments tend to be questions rather than answers: have you considered this, what happens if that.
The technical core is table stakes. The real craft lives in the outer rings: problem solving, perspective, and communication. An architect is fundamentally a translator. I work with PMs as much as engineers as much as managers, bridging the technical and the human, pointing each in a good direction without drowning anyone in detail. I try to hold the whole picture at once, because that is what lets me see the possibility space, where customer desire meets technical reality meets the people who can actually deliver, and reach across that surface to pull possibilities into being.
Here is something I rarely say out loud: for years before coding agents existed, I was vibe coding through people. I would rarely write the code myself. I would set a design and let others fill in the detail, just enough on my end to keep the outcome on track. I assembled teams of up to thirty engineers for multi-month projects, cultivated leads, handed each the right surface area, and designed the organization as deliberately as the system. Now I build organizations of software agents the same way. The art turned out to be the same art.
I'm still practicing the role more than I've mastered it. The surface area of building keeps expanding, and there is always more to learn in order to build what I want. That is what keeps me excited.
The other half
I'm not only an engineer. I studied computer science and took enough philosophy to nearly double-major in it. The two complete each other: one is how I build, the other is how I decide what is worth building. I'm a photographer. I'm a lifelong reader of science fiction and a perpetual student of the sciences, astronomy and physics most of all.
All of it comes together in The Fortunate Fall, my most recent and most ambitious writing. It's a systems-level look at the intersection of physics, biology, intelligence, and meaning, and at where each of those is quietly heading. For most of my life I have built castles in the sky out of software. This was a different medium, language at a level I had not really practiced, and it became a true window into who I am and what I think. It's the best thing I've written. It's also not finished.
Dreaming and building are my yin and yang.
Bookshelf
A few of the books that shaped me: The Singularity Is Near, Superintelligence, anything by Carlo Rovelli, Rendezvous with Rama and the rest of Arthur C. Clarke, all of Asimov, Project Hail Mary and the rest of Andy Weir, Children of Time and the rest of Adrian Tchaikovsky, The Player of Games and the rest of Iain M. Banks, Ubik and the rest of Philip K. Dick, Diaspora and the rest of Greg Egan, and the more academic ones, SICP and GEB.
A few of the films: 2001: A Space Odyssey, Interstellar, Arrival, Dune, Blade Runner and 2049, The Matrix, Star Wars, The Lord of the Rings, Ad Astra, Predestination, Ex Machina, Her, Inception, Sunshine, Moon. You can probably see the theme. The list looks narrow, but I'll happily watch anything made well.
Still building
I'm doing the same thing now that I did at that Macintosh, just with bigger tools: wanting something that doesn't exist yet, and building it until it does. The medium keeps changing. The wanting never has.