I Cannot Code. I Built Anyway.

The first command I typed into Terminal was spelled correctly. This surprised me considerably more than it surprised the computer.

I should establish my credentials for ignorance before proceeding, because this story only works if you understand the starting position. Twenty years across Indian and Global

boardrooms, across industries whose relationship to software extended no further than the dashboards I demanded reports from. An angel investor in technology companies, which is a particular irony I had been carrying quite comfortably for years; someone who allocated capital to founders building things he could not, under any circumstances, have built himself. I write about business and politics under the brand UnCut. My tagline is Uncensored. Unencumbered. Unafraid. A statement that looks somewhat less courageous in retrospect when you consider that the person making it could not have set up a web server without calling someone significantly younger.

In the taxonomy that governs Indian ambition, there is an early and rather clean sorting. You are science-stream or you are not. I was not, and somewhere between my Class XII board examinations and the subsequent two decades, the engineering path and I looked at each other, reached a quiet mutual understanding, and went our separate ways. This is not, as I eventually discovered, the permanent sentence it was presented as at seventeen.

Early this year, a question began occupying the part of my mind that responds poorly to being ignored. Artificial intelligence had stopped being something I invested in and started being something I used daily, and the gap between those two relationships began to feel, on honest examination, like intellectual cowardice dressed as modesty. These tools were not impressive in the way a well-constructed Excel macro is impressive, which is to say technically credible but spiritually administrative. They were capable in the way that a talented person is capable; able to hold an idea, interrogate it, catch their own errors, and produce something real at the conclusion of the process. The question became impossible to shelve. With tools of this quality available to anyone with an internet connection and the sense to ask a well-formed question, what exactly was still stopping someone like me from building a real application?

The answer, I concluded, was almost entirely the story I had been telling myself about who gets to build things.

I own a website, lavkush.co.in, where I publish writing on business, culture, and strategy. It existed, it functioned, and it looked precisely like what it was, a stock WordPress installation selected from a grid of templates by someone who had more important things to attend to. My essays were presented in the aesthetic register of a government circular. The homepage offered no editorial hierarchy, no visual weight, no reason for a reader arriving for the first time to believe there was anything worth staying for. I had been publishing substantive writing about consequential subjects onto a digital notice board, and I had been doing it long enough that I had entirely stopped noticing.

That gap between the quality of the thinking and the mediocrity of its packaging seemed like exactly the right place to begin. So I opened Claude and typed my first real question.

What came back was a diagnosis I would not have produced myself in a decade of owning the site. The homepage was a catalogue with no architecture. The tag and category structure was actively confusing search engines. There was no About page, which meant new readers had no way of understanding whose work they were encountering or why that question should matter to them. None of this was visible from the inside. I had been, in the way one grows entirely blind to one’s own furniture, fully acclimatised to the inadequacy.

Claude mapped out a three-stage progression; beginning with what I could change in WordPress settings without touching code, moving through AI-assisted customisation, and arriving eventually at building something from scratch. We began the same afternoon.

Over the hours that followed, I rebuilt lavkush.co.in in ways I would not have believed at the start of the session. The old theme gave way to Kadence, a more structurally flexible framework. The design was rebuilt with deliberate reference to neuroscience principles around reading retention and typographic authority; a warm parchment and deep ink palette, Playfair Display as the headline typeface, Inter for body copy. Typefaces are not merely aesthetic choices but the first silent argument you make to a reader about whether you deserve their trust. When Kadence’s free tier proved too limited for the layout I wanted, we bypassed it entirely by injecting custom HTML and CSS directly into the site through a plugin called WPCode. I was deploying real code onto a live website with a readership I cared about, without being able, at that precise moment, to explain to anyone exactly what I had deployed or why. This is either evidence of courage or evidence of overconfidence; at the sharp end of any useful action, the two are genuinely difficult to distinguish.

The homepage became dynamic, pulling each new essay forward automatically as I publish. A Consulting page appeared with three pricing tiers connected directly to my calendar link, so a reader can move from discovering the writing to booking a conversation without any intermediate friction. The Angel Investing page carries my thesis across the sectors I back, with a structured intake form for founders. My photograph appeared in the header as a circular avatar. Social links threaded through the masthead and the footer. A complete editorial website, the kind a professional design firm would price meaningfully, built in a single session by someone who, six hours earlier, had been unable to articulate the difference between a theme and a plugin.

What the session taught me had very little to do with code.

I had not learnt to code. I want to hold this with some care, because it is where accounts of this kind tend to mislead, usually in the direction of making the narrator sound more technically capable than they became. What I had learnt was how to direct technical work with sufficient precision that the AI could execute it correctly, and how to read the output well enough to know when it was wrong. When a CSS snippet produced a mobile header that collapsed in a way that looked deeply off, I could not have diagnosed the underlying code. What I could do was describe what I was observing with enough specificity to generate a correct fix on the next iteration. This is not, on reflection, a trivial skill; it is the same skill that makes someone a good client for any talented professional. You do not need to know how the kitchen operates to recognise that the dish that arrives is not what you ordered.

I also understood, for the first time in a practical way, that tools compound. The WPCode plugin extended what Kadence could do. Kadence extended what WordPress could do. Claude extended what I could do. Each layer added range to the one beneath it, and I sat at the apex of that stack as the strategic layer, determining outcomes and setting standards. Technology had not replaced my judgment. It had removed the execution bottleneck that had, for twenty years, been the reason my judgment about things digital never needed to develop.

By May, with the site running well and my confidence recalibrated to something more accurate, I decided to ask a genuinely ambitious question. Could I build a native Android application for my website and publish it on the Google Play Store?

To give you a proper sense of the starting position; I entered that conversation by asking Claude for help building, and I paraphrase myself with full awareness of how this reads, a Google app. That was the complete brief. Four words, one of which was a company name being used as a category descriptor in the way people ask for a Xerox when they mean a photocopy. Claude, with the patience I have come to regard as one of its more commercially valuable qualities, clarified the question before building anything. What I actually wanted was a native Android application built in React Native using a framework called Expo, which compiles mobile applications on remote servers without requiring you to own or manage hardware. These terms meant nothing to me at the start of the conversation and something specific to me at its conclusion, which turns out to be entirely sufficient.

We built an app with a four-tab navigation structure covering essays, categories, and dedicated pages for my consulting and investing work. The home feed pulls articles from my WordPress site via RSS, each piece surfacing with its featured image and excerpt. Tapping an essay opens a full-width dark-mode reading experience with navigation at the foot of the screen, so a reader can move through the archive without ever breaking their rhythm. The visual language carried through from the website; deep blacks against warm amber tones, the same identity adapted for a device held in one hand by someone reading on a commute. When the build was complete, Expo’s servers compiled the code remotely and returned an AAB file, which is the bundle format Google Play requires. Before submitting it, I generated an APK of the same build, installed it directly on my Pixel, and walked through every screen with the methodical thoroughness of someone who had invested real hours in this and was not about to discover a flaw after the fact. Everything worked. The developer account is live.

The entire process, from first question to a confirmed working build on my own handset, took a single day.

It was not seamless, and I should not pretend otherwise. There were commands that failed and produced error messages that looked, to someone with my background in these matters, like the kind of text that appears on a screen in the final moments before a spacecraft disintegrates in a film. There were configuration files I could not locate in an unfamiliar folder structure. There was one moment when Terminal simply sat in silence and I did not know whether it was working, broken, or forming an opinion about my competence. My discipline in those moments, arrived at first through mild panic and then through deliberate principle, was always to diagnose before attempting a fix; to copy the error into the conversation, describe the preceding action, and ask what was happening before asking what to do about it. This prevented me from doing what one instinctively wants to do in unfamiliar territory, which is to start pressing things at random and see what shifts. In most fields I can think of, random intervention in a system you do not understand is precisely how a manageable problem becomes an unmanageable one.

Here is what I believe this experience demonstrates, offered as a specific operational claim rather than the kind of inspirational observation that sounds profound at ten in the evening and hollow by morning. The barrier that has historically separated those who can build software from those who can only commission it has been structurally altered. Not dissolved; craft and engineering depth still matter, and a working prototype is a genuinely different thing from a scalable system. But the distance between a clear idea and a functional application has compressed in a way that most non-technical professionals have not yet fully absorbed. What Claude and tools of its kind have done is remove execution as the binding constraint on ambition. The constraint that remains, and it deserves to be named clearly, is the quality of your thinking; what you want to build, who it is for, what problem it solves with what degree of honesty. Those are not technical questions. They are the questions that two decades of corporate and strategic work have made me rather good at answering.

Claude is, in this sense, not a tool that lowers the bar. It is a tool that raises the ceiling. It takes the quality of your thinking and gives it a means of becoming something real, something deployed, something that a stranger on the other side of the country can open on their phone and read. That capability, available to a humanities person from the wrong side of the Class XII science-arts divide, with no engineering vocabulary and a healthy scepticism about whether any of it would produce results worth having, is not a small thing. It is, I would argue, one of the more consequential shifts in professional capability that most working people have not yet taken seriously enough.

I came to this experiment with a mediocre website and a question I could not stop asking. I left it with a redesigned editorial platform that reflects the quality of the work I publish, and a mobile application heading for global distribution on the Play Store.

The first line of code I was ever part of writing is now on a live server.

For anyone sitting behind the same assumption I carried for twenty years, that building is a skill reserved for a different kind of person shaped by a different kind of education, I offer this as straightforward and specific evidence that the assumption has not kept pace with the times. The access is real. The tools are ready. Claude is waiting with considerably more patience than most collaborators I have encountered in two decades of professional life.

The only question that remains is whether you are willing to ask the first one.