I was chatting with a Type A, roll-the-sleeve CEO friend of mine who runs a large coffee chain here in Austin, TX, and he was so excited to talk about a vibe coded software that he built on a frontier model that outperforms the clunky ERP solution from the 2010s that his company has been using for years by 100x. It meets his requirements, and it took him about a month to deploy to his operations.

However, beneath his excitement, I could see a sense of fatigue and jadedness from long, caffeine-fueled nights spent getting this product out. On paper, everything works, but none of the modules are fully done. The tech debt is creeping, and it requires constant updates to keep it alive. He signed up to become a full time Developer, DevOps, and Program Manager of his vibe coded product at night, while performing his CEO duties at day.

The reality is, the thing you build to prove an idea and the thing you build to scale are almost never the same. AI shortened the on-ramp without moving the destination.

You can reach a working product in a weekend, but the distance from a few hundred users to ten thousand hasn’t changed, the tech debt is for real. You just hit that wall faster, usually without fully understanding a system you didn’t write yourself.

The people who confuse the two pay interest on the gap later.

The tools that made the weekend possible are available to everyone. Your competitor had the same weekend. Thanks to AI, the build is no longer the moat. So the real question arrives quietly, but surely, when you have to decide what you’re going to scale, and what you actually own.

Most people frame this as a Own versus Rent decision. That framing, in my opinion, is a decade out of date.

Buying has stopped meaning owning

For about thirty years the decision was simple to name. Build it yourself, or buy it off the shelf. Either way, what you ended up with was yours to keep. That last part is what quietly stopped being true.

Then, software became a subscription. Now, you don’t own your CRM. You don’t own your email, your analytics, your tools, your stack. What you own, in theory, is your data. But even that is getting harder to keep.

Your data no longer sits in a drawer you fully control. It lives inside someone else’s system, under terms you agreed to, and increasingly, those terms now let the provider learn from what you put in.

And it works right up until it doesn’t: a price increase, an acquisition, a feature you built your workflow around removed in a release note nobody read. Renting is the default now, and for most things that’s completely fine. The problem is we stopped noticing we were renting at all. So making the decision about any alternative becomes harder.

We’re already lived this somewhere more personal. Music, for instance. Most of us have thousands of songs in our streaming platform library and own none of them. We don’t even own the playlists we might have spent a decade building; they sit on a server we rent by the month. When Neil Young pulled his catalog off Spotify, it didn’t matter how often some of us played those records. They were gone overnight.

But despite that all, it was still a good trade. Realistically, we’d never have found half the music we love today just by going to a record store.

In this case, access beats ownership almost every time. Streaming isn’t a cautionary tale. It’s the right answer for what it solves. The trap is assuming the same logic carries over when the thing you’re renting is the thing you’re supposed to be better at than everyone else.

The floor is rising, which is great news, but also a trap

Jon Ballis, the chair of Kirkland & Ellis, put it better than I can. Explaining why his firm is spending $500 million to build its own AI, he said the widely available tools are raising the floor for everyone. But a firm like his, he said, doesn’t get hired for the floor. It gets hired for the judgment that sits above it.

Judgment vs. Intelligence

As I see it, intelligence is the currency AI works in. Judgment is the human layer that decides how that currency gets spent, how raw capability turns into outcomes that actually scale. Workflows, models, automations, they all hand out more or less the same intelligence to everyone who cares to ask. What separates you is the judgment sitting on top of it.

Rented or Frontier AI has an advantage that people underrate. It gets better while you sleep. The model you’re using today quietly improves, and you ride gains you didn’t pay for proportionately or build. The floor under every business rises at the same time. That is truly valuable, and it is also exactly why it can’t be your competitive advantage. A capability anyone can rent by Tuesday is a capability everyone has by Wednesday. Think of Frontier AI as the Utilities; everyone has access to the same.

That said, it’s worth being exact about what Kirkland is and isn’t doing, because the headline number invites a lesson that might not be right for everyone. They are not writing their own language model from scratch. They’re renting the floor like the rest of us.

What they’re building, and will own outright, is the owned intelligence layer on top. They’re taking the collective judgment of 250 of their lawyers, 100 of them partners, and encoding it into a system their competitors can’t buy, and their own technology partners aren’t allowed to resell.

They are taking a world-class Open Model and training it with their Data, workflows, and domain expertise. Against $10.6 billion in revenue they’ve committed $500 million, more than $100 million this year alone, with a 180-person technical team behind it. They’re not subscribing to the catalog. They’re pressing their own records, on equipment they happily rent.

Satya Nadella made a version of this point recently. Companies, he argued, need to turn their workflows, domain knowledge, and accumulated judgment into AI systems that improve with each use. Build private evals that show whether the model is actually getting better against the outcomes the business cares about (not just external benchmarks!). Let it learn from the real work happening inside the organization. Make institutional memory something the system can query. That loop, he said, becomes the new IP of the firm.

Ownership comes in layers

It’s important to recognize that ownership isn’t a yes-or-no question. It’s like an onion, and you get to choose each layer.

There’s the model at the bottom. The tools wrapped around it. The data you feed it. The workflows you build on top. And at the very top, the intelligence you train, the accumulated way your business decides things that your competitors can’t see. You can rent the bottom and own the top.

Rent the model. It will be commoditized within a year and replaced by a better one you also rent.

Own your AI and it compounds, and nobody can sell a copy of it to the firm across the street. Most companies, today, have this exactly inverted.

A decision framework I recommend

In one of her podcast episodes, Mel Robbins asks three questions to help people figure out what they actually want from their lives. I’ve found they work just as well for someone trying to grow their business. Here’s how I think about them:

1. If you stay exactly on the path you’re on, renting the whole stack, where does the business sit in five years?
Be honest about the answer. For most companies, it’s “indistinguishable,” competing on price against everyone who rented the same tools the same week.

2. If the path you’re on disappeared tomorrow, if the tool you depend on tripled its price, changed its terms, or got acquired and killed the roadmap, what would be left?
Whatever survives that question is the part you actually own. Everything else was always a rental.

3. If nothing were holding you back, not budget, not headcount, not time, what is the one thing you’d want to own outright?
That’s almost always the thing your customers already come to you for. Start there, and rent the rest.

So strip away the subscriptions, the tooling, the demo that got the room to clap. Five years from now, what do you own that a competitor can’t rent by Friday? If the honest answer is nothing, the head start you feel today is already evaporating.

As for my friend, the Type A CEO who built his own software in a couple of weeks: I told him what I’d tell anyone standing where he was. The build proved the idea, and that was worth doing. But the thing he actually owns is the business he spent years building and the judgment that runs it. That’s his top layer.

Rent the rest, and put the nights back where they belong.

If you’re on a similar path and need a sounding board, happy to chat.

‍
Originally posted on "Wide Angle by Pradeep Nalluri" on June 18, 2026.