I recently read an interesting piece on VC Market Fit. The author, Jack Cantillon, co-founder of the VC fund Green Room, argues that as AI drives the cost of software execution toward zero, a company’s differentiation will converge on the founder’s or CEO’s taste. Using fashion designer Jonathan Anderson as his central example, he suggests tech companies are heading toward becoming multi-brand, portfolio-style organizations like LVMH.
It’s a compelling argument, but reading it gave me a strong sense of déjà vu. What’s fundamentally different from the D2C brand playbook that peaked around the pandemic? Here’s where that unease led me.
The D2C déjà vu
Most D2C brands sold products with almost no functional or price differentiation, and competed instead on worldview and storytelling. “Once execution is commoditized, taste is all that’s left” has the exact same shape as that logic. If this is only true for companies that have already found product-market fit and are scaling, and if the differentiator really does come down to something like worldview — honestly, that felt boring.
LVMH doesn’t sell convenience
But as I thought it through, I noticed something important. LVMH isn’t selling convenience. A bag can’t become more convenient, and the very idea of functional differentiation doesn’t apply to an accessory. Luxury value sits on a different axis entirely — scarcity, identity, status signaling — not convenience.
Software’s execution costs falling toward zero, on the other hand, is an event on the convenience axis. It’s worth questioning whether these two axes belong in the same conversation. The standard outcome of commoditization is a race toward zero margin on features and price — luxury-ification is actually a fairly unusual escape hatch.
That said, this isn’t limited to visible, consumer-facing goods. In B2B, if two vendors are functionally and financially identical, a deal can still fall apart simply because a buyer doesn’t like the CEO’s vibe. Push functionality and price alone to their limit, and you get a race to the bottom that looks like shopping for workwear — cheapest option that does the job. Differentiating on taste, tone, or trust is one of the few ways out of that race.
The D2C lesson: taste alone wasn’t a moat
Which brings up something worth remembering: most D2C brands ultimately couldn’t defend that exit. Casper, once the poster child of D2C mattress brands, saw its valuation collapse shortly after going public and was eventually taken private. Rising ad-driven customer acquisition costs and a flood of copycat brands are widely cited as the main causes. Building a worldview, on its own, never became a real barrier to competitors.
LVMH’s real strength isn’t cross-selling — it’s time
So what actually makes LVMH different? My first instinct was to say a scaled-up D2C brand couldn’t replicate this structure even after finding success — but that needed correcting.
Digging in, I found that LVMH deliberately avoids integrating customer data or cross-selling across its houses, and preserves strong autonomy for each brand. What the group centralizes isn’t customer data — it’s vertically integrated infrastructure spanning capital allocation, real estate negotiation, and raw-material sourcing, plus talent development and mobility across brands. It’s also characterized by patient capital: decades-long investment horizons with no demand for immediate returns.
Amazon FBA brand “roll-up” companies like Thrasio tried the same multi-brand playbook without those underlying assets — accumulated capital, a magnet for talent, patient investors — and it didn’t work. The structure of “going multi-brand” is something anyone can conceptually copy; the assets that make it function can’t be built overnight.
Apple: the exception where the aesthetic economy actually worked
Here’s a case where taste-based differentiation actually did work: the iPhone. Strictly speaking it’s hardware, not software, but it’s a suggestive case nonetheless.
As I noted earlier, taste-based differentiation works on products you can hold, wear, or be seen with. The iPhone meets exactly that condition. So Apple’s success is better read not as validation of this article’s thesis, but as a rare case that happened to satisfy the narrow conditions under which taste actually works.
But Apple’s real strength was never taste alone. It’s industrial design combined with the switching costs of leaving its ecosystem (the “flow-type fit” I get to below), plus the organizational continuity of having one extraordinary individual at the center of the company for decades. Just as a Jonathan Anderson only comes along rarely, so does a Steve Jobs. Apple isn’t proof that anyone can win the taste game — it’s closer to proof that this only works at lottery-ticket odds.
Which also means it’s inaccurate to split the smartphone market cleanly into “Apple = taste, Android = commodity race.” Some Android challengers — a UK-based brand known for its transparent hardware design, for instance — have pushed taste-driven differentiation hard enough to bring on well-known artists as brand ambassadors and shareholders. Meanwhile, the cheapest price tiers still compete mainly on battery capacity, screen brightness, and storage. The real dividing line isn’t Android versus Apple — it’s price tier and brand strategy, and that line runs across operating systems, not along them.
The thread running through all of this: time as the scarce resource
Line these points up and they converge on a single axis. Trust, the ability to attract talent, patient capital — every durable differentiator depends on assets that compound over time. And time is exactly what AI is compressing. The real question isn’t whether you differentiate on taste or on execution speed — it’s whether you can still get time on your side.
Data has a stock side and a flow side
That same lens applies to data. There are two kinds. One is “stock” data — static records like purchase history or account details — and as AI-driven portability improves, the time needed to accumulate this kind of data could shrink. The other is “flow” data — the ongoing adaptation of a system to a specific user’s or team’s quirks through continued use — which can only be created by actual elapsed usage time.
If that’s right, then ironically, the scarcest and most defensible asset in the AI era may be neither taste (this article’s thesis) nor code itself (the old SaaS moat argument), but the accumulation of real usage that AI simply cannot accelerate. Though it only becomes an asset once it’s translated into product experience — hoarding data by itself means nothing.
Visibility is a matter of reach
Going back to the Apple point — the condition that taste-based differentiation works on things you can hold, wear, or be seen with — this is better reframed not as a question of whether visibility exists, but of how far that visibility reaches.
This connects back to the B2B example I raised earlier: a deal falling apart because a buyer doesn’t like the CEO’s vibe. For a buyer to know anything about a CEO’s vibe in the first place, it has to be visible somehow — through a sales meeting, a conference talk, something posted on social media. So visibility does exist in B2B. It’s just that its reach is narrow, closed within a peer group of specialists. Consumer goods like an LVMH bag or an iPhone, by contrast, reach a stranger walking down the street. What separates B2B from B2C isn’t the presence of visibility — it’s how far that visibility reaches.
Publicly announcing usage on social media, or having continued use be externally observable, can extend that reach to something close to what a consumer good has. NFT profile pictures already anticipated something close to this — the mechanism of externally verifiable, flaunted ownership was doing exactly that job. Right now this is limited to partial visibility on social media and experiments like NFTs, but if usage inside VR-style virtual worlds ever becomes visible and mainstream, choosing a piece of software could become a status signal with the same reach as a bag seen on the street.
Conclusion: is the aesthetic economy inevitable?
Having worked through all this, I think my initial “this is boring” reaction needs revising. Differentiating on taste or worldview is, on its own, a straight continuation of D2C — nothing new there. But if AI keeps compressing both execution costs and the time needed to accumulate stock-type data, and the one thing left uncompressed — flow-type fit — is where the next competitive edge lives, the picture changes.
The thrill that has driven the software industry has long come from building something no one else could, using pure technical skill to outmaneuver everyone else. If the world shifts entirely to a taste-centered game, that thrill might disappear. But if designing “fit” — something that can only be refined through real accumulated use — becomes the next competitive axis, there’s a different kind of interest left in that. Not the thrill of out-building someone with unique code, but something genuinely technical, and worth building over time.
