Jeff Park@dgt10011
This weekend I revisited an old problem-
When I was a hedge fund analyst in the 2010s, I hired an offshore developer to parse 13Fs and build a systematic trading strategy around what I considered proprietary signals. I knew which star managers were not only legit but also relatively unknown and under the radar- because they were... my friends. I knew how to "read" 13Fs based on matching strategies: special sits, RV, merger arb, long/short, weighting recency and duration with learnt optimization techniques. Without this context, 13Fs are lifeless data strung together for SEC disclosure that can be easily misinterpreted (esp. because derivatives and shorts are not reported entirely or correctly). But if you knew where the signal was, AND how to digest it, it became a priceless treasure mine, especially in sectors like biotech where alpha and dispersion run high.
The #1 constraint though, and why I ultimately stopped the project, was that the offshore dev was difficult to work with: communication, reliability, quality. The problem had little to do with my technical literacy (fun fact, I competed nationally in the American Computer Science League in high school, placing in both C++ and Java, so while I am by no means fluent, I can read and direct). The problem was entirely human. It became so frustrating to work with a partner whose incentives weren't aligned with my demand for urgency, accuracy, robustness. So I ultimately abandoned the project and that was over ten years ago.
This weekend, I revisited the same project using Claude. In that instant, I didn't just see the future- I touched it. I may have even grasped it. The best way I can describe the experience is as @karpathy described it a few years ago: "The hottest new programming language is English."
One of the primary reasons engineers have been so high in demand (and well compensated) is because they're essentially "translators of the modern business." Anytime you want to do anything, you need a "translator" to write your intent in a software-driven world. I can "speak Spanish to get by," but it's never going to be as good as someone fluent, so I always need a third party involved. In that process lies a game of translation that's arduous, frustrating, and time-consuming. And most often, engineers don't care as deeply about sales and product as sales and product care about their own deliverables to clients. The friction therefore is inevitable. Anyone who has ever tried to work with a technical implementation partner knows this inherent tension.
But Claude is a translator at will. I booted my pilot program in under four hours with dashboards, scripts, exactly with the context I wanted, with rudimentary agentic workflows that are still dynamically developing. What this revealed to me is the return of the Socratic method. Claude gets things wrong sometimes, so you have to ask critical questions to help it learn. The kinds of questions that used to take hours of education, often with a dash of hurt feelings or other bottlenecks that come with working with humans. Not Claude. Not AI. There is no principal-agent problem anymore.
Which led me to a great realization: It's not just that "programming is now English"- it's "programming is now unconstrained English." An Esperanto 2.0.
There are profound implications here for what it means to invest IN the future, and to invest FOR the future. In this particular project's case, the signals remain somewhat proprietary (since the world doesn't know which funds are run by my smart friends and 3c7 funds can't really be data scraped). I've long intuited that education will look different for my children- that they'll need to learn how to think critically to get ahead which is one reason I've experimented with progressive education models. I'm now convinced more than ever that the world will no longer demand domain expertise at the individual level for everything, but rather the ability to prompt, contextualize, and critically reason for adaptive outputs.
Contrary to many general assumptions, this "language" will actually specialize, and evolve far faster than any ivory tower curriculum can accommodate. Skill gaps will emerge at scale in ways we cannot predict. It's not dissimilar to elite schools teaching Latin while the rest of the world had already moved to English. We haven't invented the word yet for what this "language" is, but I would describe it as "symbolic systems to communicate with machines."
If you haven't played with agentic models yet, this is both an invitation and a warning: start now. An invitation because you'll feel like a superhuman, rediscover a childlike curiosity you might have once lost, and most of all, find your brain critically engaged with an incredible sparring partner, one that is as good as you make it and most importantly: yours to keep.
But also in the spirit of warning, there is a remote yet dark possibility that this tool may not remain as accessible in the future, especially if cost or regulation impedes it, or if populous society turns on it before a critical wealth effect is distributed. So drink the elixir while the fountain is madly rushing out of the firehose today. Because whatever the future may hold, you will never forget the taste for the rest of your lifetime to come.
