If you’ve been reading ARK Strategy for a while, you know that I studied political philosophy and history.
I want that on the record before I tell you the rest of this, because the rest of this involves an SQLite database, a Streamlit web app, Python, pandas, NumPy, Matplotlib, and a half-dozen other tools whose names I could not have defined three weeks ago and, in some cases, still could not define under penalty of perjury.
I have never learned to code. Let me be precise about that. I did not, over the course of one weekend, become a software engineer. I cannot whiteboard you a binary tree. If you asked me what NumPy is, in a sentence, I would do the thing I do at dinner parties when someone mentions a novelist I haven’t read: I would nod with great seriousness and change the subject. (Numpty Dumpty?)
And yet — at the end of a weekend of vibe-coding, I had a working app. A real one. With a database underneath it, a graphical user interface front end on top of it, and the bones of something I could imagine handing to a client on Monday morning.
That is the part I want to talk about. Not the code. The gap.
For most of my career, the minimum viable product was a thing other people built for you. You wrote the memo, drew the boxes, and found the engineer, scoped the sprint, fought for the headcount, negotiated the timeline, and waited. The MVP was an artifact of budget and politics as much as it was an artifact of code. In a hedge fund, in a startup, in a Fortune 500 transformation, it was the same story, but a different zip code. The wall between the person with the idea and the working version of the idea was real, and it was tall, and it was often guarded.
That wall fell over while I was making coffee this weekend.
I opened Claude Code on a Saturday morning with a vague idea and am even worse outline. I described what I wanted, in English — actual English, the kind I learned writing seminar papers on the Hegelian Dialectic, and the model started building. SQLite for the database. Python to glue it together. Pandas to push the data around. NumPy somewhere in the middle doing god-knows-what. Streamlit on top so I could actually see the thing.
Errors? Of course. Plenty. The first version didn’t run. I got big red errors in Streamlit when I tried to open it. The second version ran, but when I went to make changes to the data, it wouldn’t save my changes. The third version ran, didn’t mysteriously drop my edits, and then quietly broke when I added a time-series chart, with exportability in HTML and PDF. I am not telling you it was entirely clean. But I am telling you it was incredibly fast to iterate.
By Sunday night I had a working MVP. Not a vague sketch of a product. Not a wireframe report. A thing that opened in a browser, accepted inputs, queried a database, and returned a result.
A poli-sci major, a weekend, and a working product.
Let me try to say this without the breathless tech-bro register, because I think it matters:
The cost of building the first version of an idea has collapsed. Not the cost of building the good version. And not the cost of building the scalable version. Certainly not the cost of the version that holds up in front of a CISO. The cost of the first version. The version that lets you find out whether the idea is even worth having.
For twenty years, the first version was the bottleneck. You couldn’t test an idea until you’d found someone willing to build it, which meant you mostly tested ideas other people were already willing to build. The selection effect was brutal. Founders without engineering co-founders were stuck pitching decks. Operators with sharp instincts and no Python were stuck describing what they wanted. Consultants with brilliant frameworks were stuck slide-deck-ing.
That bottleneck is gone. Or, If I’m being honest about it, that bottleneck is gone for a specific kind of work, at a specific level of fidelity, for a specific kind of operator who is comfortable iterating in plain English. Which, as it turns out, is most of us…
The era of the MVP has never been faster. The wall between the idea and the working version of the idea is no longer guarded by a team of engineers. It is guarded, increasingly, by your willingness to sit down on a Saturday and describe the thing clearly to Claude Code.
A few things I am not saying, because I’d like to be precise about the asterisks here:
I am not saying the app I built will stand the test of time. I have no idea if it will run in six months. I suspect it will not, in its current form, survive contact with a real user load, a real security audit, or a real cofounder with strong opinions about architecture. And I’m not pretending otherwise…
I am not saying that I now understand the stack. I understand what each piece does in roughly the way I understand what a carburetor does: I could maybe point at it in a diagram and get the verb right. I could not fix one in a parking lot…
I am not saying engineers are obsolete. They aren’t. The version of this app that an actual engineer would build is materially better than mine, in ways I am not even sophisticated enough to enumerate. The gap between my MVP and a production system is still wide, and it is still bridged, mostly, by people who studied computer science instead of “Nationalism, Republicanism, and Cosmopolitanism.” (Yes, that was actually a class I took in college).
What I am saying is narrower, and I think more interesting: the floor of what a non-technical operator can build alone, in a few days, has moved materially, and permanently. In a way that changes the strategic calculus for almost every solo operator I know…
If you are running a portfolio career, or if you are a fractional anything, a consultant, a founder-in-residence, a chief of staff between gigs, an operator with three ideas and no engineering co-founder, this is your ultimate unlock.
You no longer have to pitch the idea. You can show the engineers, or the client, your idea.
You no longer have to describe the workflows. You can deploy the workflows yourself.
You no longer have to write the memo about what the dashboard should look like. You can spin up the dashboard, point at it, and say like this, but with your data. Or, like this, but make it more secure, and in the cloud.
The deliverable has changed. For a long time, the deliverable in my line of work was the deck, the memo, the framework. Those are still useful. But the highest-leverage deliverable in 2026 is increasingly the working prototype: the thing the client can click on, the thing the founder can demo, the thing the board can see. The prototype is the new memo. And the prototype, as of this weekend in my Brooklyn apartment, is now buildable by a humanities major with a Claude Code subscription and a cup of coffee.
This is what I have been circling for months in these memos: the judgment stack matters more, not less, in this economy. AI eats the rote task layer. It does not eat the taste, the framing, the what should we even build question.
I was in awe of describing a chart, or a particular cut of the data, a timeseries view along certain variables, or a particular breakout view, and watching Claude Code bring it to life on my screen.
But what AI now also does — and this is the new thing — is collapse the gap between deciding what to build and having a version of it to look at. That gap was where most ideas, historically, went to die. That gap was the product graveyard. But that graveyard is shrinking.
I am, fair warning, going to keep building things. Badly at first to be sure. Then less badly. Then, with any luck, well enough to put in front of a client. I will not become a software engineer, but I have been told by a programmer friend that I can call myself a developer now…. (I won’t).
But all of that is not the bet. The bet is that the operator who can describe a thing clearly in plain English and ship a rough version of it by Sunday night is going to eat very, very well over the next five years. And the operator who insists on waiting for the engineering team to free up a sprint is going to keep watching their backlog grow...
The MVP used to take a quarter. It’s now a weekend.
It will, soon enough, take an afternoon.
I studied political philosophy and history. I am telling you, with all the conviction of a man who could not define NumPy under oath: this is the part of the curve where the wall comes down. Build something this weekend. It doesn’t have to be good. It just has to exist. The existence is the unlock.
Errors and all.
— Alex
This is Strategy Memo №15 in the ARK Strategy series. If you’re building a portfolio career, navigating transformation, or trying to figure out what to actually do with these tools, subscribe below. One memo a week, careers as portfolios not ladders.
Alex Randall Kittredge writes the Substack ARK Strategy and is the founder and Managing Director of APR Strategic Consulting. He has advised 9+ CEOs, integrated workforces, and redesigned organizations across hedge funds, startups, and industrial companies. He is a graduate of Columbia University and the University of Cambridge, and mentors entrepreneurs & startup founders through Oxford Entrepreneurs Network, CamEntrepreneurs, and Plug and Play Tech Center. He is a Director of The Oxbridge Method Ltd, and is the author of the forthcoming book, How Your Side Hustles Will Save You: Creating a Durable Career that Transcends the Corporate Ladder.
Compliance Statement: All views and opinions expressed are solely my own and do not reflect those of any current or former employer, client, or organization with which I am or have been affiliated.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.