How Streaming Shows That AI Is Making the Product a Commodity, Here’s How Startup Founders Survive

I find myself more oriented to media than most. I am a huge live music fan extending well beyond concerts and festivals into even musicals and Broadway. My career started in advertising by way of Yahoo! when “online” was little more than a concept. I was among the first with Netflix, familiar with The Pirate Bay, and a huge fan of movies. So, when startup ecosystems started waking up to the benefits of specialization, pushing MediaTech was the obvious direction in which I could lean on my experience to help cities stop thinking in broad terms such as being a “tech hub” and instead focus on being FinTech, AgTech, or MedTech, to draw from the market of talent, experience, and connections available there.

This isn’t about media. This is about what tech and founders might appreciate by better understanding it.

Who actually owns the pipe?

That question has been at the center of every media war since the first radio signal was transmitted, and it’s the same question that will determine whether your AI-forward startup is a real business or an expensive hobby. Streaming (both music and television) spent the last thirty years working out the answer the hard way, and software is about to learn the same lesson, at scale, on fast-forward, courtesy of AI. Founders. all, pay very close attention to the story of Spotify.

Radio Waves, Broadcast Towers, and the First Pipe War

When Guglielmo Marconi demonstrated radio transmission in 1895, nobody was thinking about business models; technology was the spectacle. But by the 1920s, commercial radio stations in the United States had figured out something that still seems to elude most entrepreneurs: whoever controls the transmission frequency controls the audience, and whoever controls the audience controls the money. The Federal Radio Commission, established in 1927 and later reorganized as the FCC in 1934, existed precisely because radio spectrum is a finite physical resource (a literal pipe with limited bandwidth) and someone got to decide who got to use it.

The medium was free to listeners. Advertising paid for everything; distribution was the actual product, because distribution was scarce. Listeners weren’t customers, they were users, they were the market, being funneled to ad dollars to fund the pipe. This is the first and most important lesson in the long history of media: the pipe is never just infrastructure; the pipe is the business.

Television inherited the same logic. Over-the-air broadcast television, funded by advertising, reached American living rooms for free (sort of, you bought the TV, the content cost nothing). The FCC allocated channels, the networks bought the airwaves in effect, and the system created ABC, NBC, and CBS as dominant cultural institutions. Scarcity was real and enforced by physics and regulation simultaneously. Content mattered, of course, but distribution was the moat.

Cable television cracked that open (well, more accurately, closed) in ways nobody fully anticipated. The technology was straightforward enough: instead of transmitting signals through the air, you ran coaxial cable into homes, which allowed for more channels and better signal quality. HBO launched in 1972. By the 1980s, cable was reshaping American television consumption entirely. The key shift wasn’t technical, it was economic and structural; cable operators weren’t using publicly owned airwaves, they were running their own wire. They could charge for access directly and suddenly, the pipe that had been “free” (subsidized by advertising and regulated as a public good) became a subscription product with steadily rising prices.

Between 1995 and 2020, cable and satellite television prices rose at roughly three times the rate of general inflation. The average American cable bill went from around $22 per month in 1995 to over $100 by the mid-2010s. Satellite offered competition but not a fundamentally different economic model; it was still a proprietary pipe with controlled access and escalating fees. What had been a public utility of the airwaves was replaced, channel by channel, with a controlled-access toll road.

Piracy Is Always a Business Model Failure, not a Moral One

When people stopped being willing to pay $120 a month for 500 channels of which they watched 12, they found another way. Napster launched in 1999 and immediately demonstrated what happens when you charge too much for a pipe and then make the content digitally distributable: people build their own pipe. BitTorrent, released in 2001 by Bram Cohen, gave pirates a decentralized distribution mechanism so efficient that the entertainment industry’s legal response couldn’t keep up with the technical reality.

Piracy is not primarily a morality problem, is a pricing and distribution problem. People did not pirate music because they hated artists, they pirated music because the CD was $18 for one or two songs they actually wanted. Evidence in the demand, people love the artists and the music. People didn’t torrent HBO shows because they wanted to steal from studios, they torrented because they didn’t want to pay $140 a month for cable to access the two shows worth watching. Every mass piracy event in the history of media has followed the same pattern: distribution became too expensive, too fragmented, or too controlled, and someone built an alternative pipe.

The entertainment industry’s response? Litigation, DRM (Digital Rights Management, which is the technical system that restricts how you can copy or share digital content), and lobbying for legislation, each, efforts that address the symptom rather than the disease. The RIAA sued Napster into oblivion and the MPAA went after individual file-sharers. None of it worked at scale, because the economic incentive for piracy remained intact. The pipe was just too expensive.

Netflix and the Broken Promise of One Affordable Pipe

Netflix launched its streaming service in 2007 and, for a glorious and brief window of time, actually solved the problem. For $7.99 a month, you got an enormous library of films and television shows, no commercials, no cable contract, no satellite dish. It was an obvious value proposition, and it worked spectacularly. Reed Hastings understood something the cable industry had forgotten: if you make the legal alternative appropriately priced and convenient enough, most people will use it rather than pirate. He was right.

The studios noticed. Of course, rather than leaning in on what worked, they made it complicated.

Disney, WarnerMedia, NBCUniversal, Paramount, and Apple all watched Netflix build a $100 billion business on top of their content and concluded, albeit correctly, that they were leaving money on the table. Beginning around 2019, the streaming wars began in earnest. Disney+ launched in November 2019 at $6.99 per month. HBO Max (now Max) followed while Peacock, Paramount+, Apple TV+, and Discovery+ arrived in rapid succession. The content that had made Netflix so compelling (the Disney films, the HBO shows, the Universal catalog) was pulled back into proprietary platforms.

By 2024, a consumer who wanted access to the full range of prestige television was paying for Netflix, Max, Disney+, Peacock, Paramount+, and Apple TV+ simultaneously. The combined monthly cost, depending on ad-supported tiers and bundling, approached or exceeded the cable bill that had driven people to cord-cut in the first place. And the user experience got worse: you now needed to remember which show was on which platform, manage six different apps and six different accounts, and deal with the peculiar rage of finding that something you watched last month has moved to a different service.

Admit it, streaming television sucks now.

Nobody is racing to subscribe to Paramount+ for Star Trek: Strange New Worlds and then cancel once they’ve watched it, then resubscribe when Star Trek: Academy drops. That’s not a streaming economy; that’s a cable bundle that requires more passwords and bills. The unitary promise of Netflix (one affordable pipe for everything) was broken by the studios the moment they understood that controlling their own pipe was more valuable than the licensing fees Netflix was paying. Peter Drucker’s maxim that the aim of marketing is to know and understand the customer so well that the product sells itself applies here in reverse: the studios understood their customers well enough to know they’d pay for fragmented access, because the alternative was pirating or simply not watching.

Music Made the Same Mistake, and It’s Terminal

The music industry ran the similar playbook, with familiar consequences, and Joel Gouveia’s recent Substack piece “The Death of Spotify: Why Streaming is Minutes Away From Being Obsolete” puts the structural problem better than almost anything else written on the subject recently. The piece is worth reading in full; here’s what matters for this argument.

Music distribution followed the same arc as television. Vinyl records required physical manufacturing and physical retail: the labels owned the format and the distribution. When RCA introduced the 45 RPM vinyl format and also manufactured the record players that played them, they were vertically integrated in a way that made the business completely legible: you controlled the product from recording session to living room turntable. Philips invented the compact cassette and owned PolyGram Records. Sony co-invented the compact disc with Philips and manufactured the Walkman.

As Gouveia writes, quoting a VP at Warner Music Canada: “For the first time in history, we (the majors) don’t control the entire means of consumption.” That is the entire story of every media industry disruption, in one sentence. It’s why you need to understand the implication in startups.

Napster destroyed the CD business model not because music became worthless but because the marginal cost of digital music reproduction dropped to zero while the price remained $18. iTunes, in 2003, offered a partial solution: $0.99 per song, legitimate, easy. It worked well enough to stop the bleeding without stopping the underlying structural shift. Spotify arrived in 2008 in Europe (2011 in the US) with the streaming model: all the music, one subscription, $9.99 per month. It was Netflixian in its logic, and it was enormously successful by the metric of subscriber growth.

But here’s where the music industry’s version of the streaming trap differs from television’s, and it’s where Jimmy Iovine’s observation is broadly devastating when understood. As Gouveia’s piece explains at length, Iovine (co-founder of Interscope Records, co-creator of Beats by Dre, architect of Apple Music) told David Senra on the Founders podcast (Episode 391) something that we have to talk about, “The streaming services, to me, are minutes away from being obsolete.”

Iovine’s reasoning is structurally relevant, not nostalgic. Unlike television, where Netflix differentiated itself by owning original content (you need Netflix to watch Stranger Things; you need HBO to watch The Last of Us – which we hate, granted, but that’s not the point here), music streaming services have no differentiation at all. As Iovine put it plainly on the podcast: “Right now, music streaming is a utility. All the services are exactly the same, they do the same trick. If one of them lowered their price the rest are toast, because there’s no unique offering.”

Spotify, Apple Music, Amazon Music, and Tidal all offer the same 100-million-song library. There is no music you can get on Spotify that you cannot also get on Apple Music. The product is identical. The only things that differ are UI design choices and algorithmic playlist curation; neither of which constitutes a durable competitive moat. What Iovine and Gouveia are describing, in marketing terms, is a commodity: a product so undifferentiated that competition occurs entirely on price, and whoever has the lowest cost structure wins.

The financial mechanics of this make it worse and are really what got me thinking about the implication to entrepreneurs. As Gouveia explains in his piece, music streaming operates with inverted economics compared to a normal SaaS business. Spotify pays out roughly 70% of every dollar of revenue back to rights holders. That cost scales linearly with subscribers (every additional stream costs money, unlike Netflix’s fixed-cost content model. Iovine “The streaming services have a bad situation, there’s no margins, they’re not making any money.” This is why Apple, Amazon, and Google can sustain music streaming losses indefinitely (they’re selling hardware and Prime memberships and cloud services) while Spotify, as a standalone streaming company, is in a structurally compromised position.

And it’s also why, as Gouveia argues, the only artists who survive the coming reckoning are the ones who build direct ownership of their audience; their own pipe to their own fans. A platform doesn’t want you to have a relationship with your fans; the platform wants your fans to have a relationship with the platform. Say that again this way, “A platform doesn’t want you to have a relationship with the solution; the platform wants you to have a relationship with the platform.”

Spotify guards listener data the way a cable company guards subscriber data: because that data is their moat, not the content.

Software’s Version of the Same Story

Software made the same journey, just on a slightly different timeline and with one significant structural wrinkle that made the parallel less obvious until recently. For most of the 1980s and 1990s, software was sold on physical media (floppy disks, then CD-ROMs) through retail channels. Microsoft’s dominance in that era was partly about the quality of Windows and Office, but substantially about the exclusive agreements with PC manufacturers (OEMs, Original Equipment Manufacturers) that put Microsoft software on every machine before anyone switched it on. The pipe was the retail shelf and the hardware bundling agreement; control the shelf, control the market.

The internet changed the distribution model for software just as profoundly as it changed it for music. When you could download a product directly, the physical retail channel became irrelevant. This enabled piracy of software at scale; warez sites and peer-to-peer networks distributed cracked versions of Photoshop and Microsoft Office to millions of users who would never have paid the retail price. Notice, now Google, actually selling something else just as in music, gives such solutions away for free. It also enabled something more important in the long run: the SaaS (Software as a Service) model, where instead of selling software as a product you installed, you sold access to software running on your servers, billed on a subscription basis.

SaaS solved the piracy problem; you can’t pirate a subscription to Salesforce. There’s nothing to copy. The software lives in the cloud; you’re buying access, not a product. It also created the most favorable business economics in the history of software: high gross margins, predictable recurring revenue, low incremental distribution costs. Andreessen Horowitz spent a decade funding SaaS companies on exactly this basis, because the economics were genuinely superior to any previous software business model.

But while SaaS was solving the piracy and distribution problem for software companies, it was simultaneously concentrating the internet’s value in fewer and fewer hands. Google owned search. Facebook owned social. Amazon owned e-commerce and, increasingly, cloud infrastructure. Apple and Google together owned mobile app distribution. The internet that had briefly been a genuinely open network (where anyone could put up a website and compete on roughly equal terms) became a series of platform oligopolies, each controlling their own pipe to hundreds of millions of users.

This is the context in which debates about Net Neutrality and Section 230 of the Communications Decency Act should be understood. Net Neutrality, the principle that internet service providers must treat all internet traffic equally, without throttling or prioritizing traffic from companies that pay extra, is fundamentally an argument about who controls the pipe at the infrastructure level. FCC Chairman Tom Wheeler, when implementing the 2015 Open Internet Order, stated that the policy would ensure “that no one (whether government or corporate) should control free open access to the Internet.” That order was repealed in 2018. The fight has never stopped because the stakes never changed: control of the pipe at the infrastructure layer is worth hundreds of billions of dollars.

Section 230, which provides internet platforms with immunity from liability for third-party content posted on their services, is a different kind of pipe argument. As Yale Law School’s Journal of Law & Technology published in a landmark paper by Adam Candeub, both Net Neutrality and Section 230 “reflect a historically typical ‘regulatory bargain’ first found in common carriage, the body of law that has regulated transportation and communications networks for centuries.” The argument about whether Facebook and X should be treated as neutral platforms (protected by 230) or as editorial publishers (liable for what they host) is, at its core, an argument about who controls the flow of information and on what terms.

The pipe, again.

Blockchain and Web3 emerged in the late 2010s as the tech industry’s answer to the concentration problem. The theoretical promise of decentralization (there is that word again) removing the intermediary platform (the pipe controller) and allowing peer-to-peer transactions verified by distributed consensus, is genuinely interesting as an idea. If you could rebuild social networks, financial services, and software distribution on decentralized protocols, no single entity could hold the chokepoint. The practice was messier; most blockchain applications replicated the same centralization dynamics in new wrappers. The platforms that launched “decentralized” applications still controlled the user experience. Crypto exchanges became the same kind of concentrated intermediaries as traditional financial institutions. The pipe problem proved more durable than the technology that claimed to solve it.

Vibe Coding, AI, and the Moment Software Becomes the Music

Which brings us to now. And to the thing that should make every founder who is building an “AI-powered” product think.

Vibe coding, a term coined by Andrej Karpathy, co-founder of OpenAI, in February last year, describes the practice of using AI to generate software through natural language prompts, without the developer needing to write or even fully understand the underlying code. As Karpathy wrote in the original viral post that spawned the concept: there’s a new kind of coding “where you fully give in to the vibes, embrace exponentials, and forget that the code even exists.” It was the Word of the Year for 2025. Merriam-Webster added it as trending slang in within weeks. It is reshaping who can build software and how fast.

According to TechCrunch reporting on Y Combinator’s Winter 2025 cohort, 25% of startups in that batch had codebases that were 95% AI-generated. A February 2026 analysis by Kristin Darrow found that 84% of developers reported using or planning to use AI tools in their workflow, with iOS app releases up roughly 60% year-over-year in late 2025. I’m meeting founders at networking events who have built deployed apps in a weekend, and they don’t know how to code. The cost and time required to prototype new products has dropped significantly. Tools like OpenClaw, Cursor, Lovable, Replit, and Bolt.new have made it possible for non-engineers to ship functional software in days rather than months.

On the surface, this is fantastic.

Democratization of software development is genuinely useful.

But why you’re reading with me is what I’d call the Spotify Test.

If every music streaming service offers the same 100 million songs and is therefore indistinguishable, what happens when every SaaS product can be built by anyone with a prompt and an afternoon?

If the music is the commodity, the platform that delivers it becomes the commodity too; except that the platform still has to charge you, while the music costs nothing to reproduce. Spotify’s structural trap is that the product it delivers (music) is available from every competitor at identical price and quality, while the cost of delivering it (licensing fees) is non-negotiable. Now apply that to software: if your AI-powered project management tool can be replicated by a competitor over a weekend, or built directly by a sophisticated user using Claude or GPT-4 or Gemini, what exactly is your product differentiation?

As I shared in my recent analysis of predictors of startup failure: “It’s like [X] but with AI!” isn’t a pitch; it’s a confession that your product can be recreated in ten minutes by anyone with access to ChatGPT. If your differentiation disappears when OpenAI ships a new feature, you don’t have a moat, you have a timer.” That was written before vibe coding went mainstream; the timer is now much shorter.

Krish Venkataraman wrote with Techstrong.ai that the AI commoditization problem is stark: “A majority of enterprises today are using OpenAI’s GPT-4 as their foundational model, mostly through Microsoft’s Azure AI service. Anthropic, Mistral and Cohere are also popular but less widely used. The performance of these foundational LLMs doesn’t differ widely, which basically means they are all using the same technology.” The analogy they use is Oracle versus IBM databases; both broadly similar, neither constituting significant differentiation on its own.

This is the musical streaming problem, translated into software: when the underlying technology (the AI model) is available to everyone at commodity pricing, and when the ability to deploy that technology in a product context (vibe coding) no longer requires a software engineering team, the product you build with it is not a moat.

It’s a demo. And demos, unlike products, are not businesses.

The Pipe, the Channel, the Platform, and the Code

Let’s start exploring the hook in my headline. What do we do about it? Marketing and business strategy have long recognized that distribution is not ancillary to value, it often is the value. A framework that makes this concrete distinguishes between the pipe, the channel, the platform, and the code.

The pipe is the fundamental infrastructure through which everything flows: the radio spectrum, the cable wire, the internet backbone, the smartphone operating system, the AI model API. Whoever owns the pipe extracts rent from everything that flows through it. The FCC knew this about radio. The cable companies learned it about coaxial wire. Apple learned it about the iOS App Store, which takes 30% of every digital transaction as a fee for access to the pipe. OpenAI is learning it about the API access to GPT-4.

The pipe generates the most durable, highest-margin revenue in any media or technology ecosystem, because it’s the prerequisite to everything else.

The channel is the distribution mechanism that sits in the pipe: the broadcast network, the cable system, the streaming platform, the social media algorithm, the app store category ranking. Channels aggregate audiences and direct attention. They are extremely valuable during the period when they have leverage over both content providers (who need the audience) and consumers (who need the content). They lose power the moment either side develops an alternative. Netflix was a channel until the studios took their content back. Spotify is a channel that loses power every time an artist builds a direct-to-fan income stream through Patreon or Bandcamp or Discord.

The platform is the product layer: the software, the interface, the experience, the specific application of the underlying pipe and channel infrastructure. This is where most startups operate, and where most startups mistakenly believe they have a durable business. Platforms derive power from network effects (the more users, the more valuable the platform to each user), from switching costs (how painful is it to move your data, your workflow, your relationships to a competitor?), and from brand trust. Platforms without genuine network effects or switching costs are perpetually vulnerable to better-funded or better-distributed competitors building the same thing.

The code is the specific implementation; the software that runs the platform. For most of the SaaS era, proprietary code was a meaningful competitive advantage. Writing a good algorithm was hard while building a complex database schema that worked at scale required years of engineering. Software engineering talent was expensive and scarce. All of these factors made the code itself something worth defending. Vibe coding is in the process of making that entirely untrue. When your code can be replicated by a competitor with a good prompt and a Cursor subscription, the code is not the moat.

I once wrote about startup distribution, “attention follows distribution. And distribution doesn’t come from hope. It comes from relationships.” The failure mode I described – startups failing in silence because they never built real distribution – is now going to become catastrophically more common as AI makes it trivially easy to build a product while leaving the harder problems of distribution and market ownership completely unsolved. The most important priority for any startup isn’t the product itself, it’s the marketing, which in its true definition means understanding the market, the customer, the distribution, and the competitive position before building anything.

Even if you disagree with me about marketing, appreciate that what is happening when you vibe code and release that app, is that thousands of others are looking at it, able to *freely* do it better. So, let’s not call it “marketing,” call it defensible distribution.

Is Being an AI-Forward Company a Commodity Position?

Apply the streaming lesson directly; in music streaming, the undifferentiated product (100 million songs available on every platform) means the platform itself becomes a commodity competing on price, UX, and algorithm quality, none of which generates the kind of durable competitive advantage that sustains a high-margin business. In software, AI is now making the code itself the undifferentiated product; every startup can access the same foundation models. Every startup can use vibe coding to build applications faster. Every user who is sophisticated enough will, in short order, be able to build their own customized version of your product for their own specific needs, at near-zero marginal cost.

This is not hypothetical. Kristin Darrow’s analysis of vibe coding’s state characterizes it in appreciating that vibe coding lowers the barrier to generating software. It does not eliminate the need for accountability. “The future norm is likely to be that organizations that traditionally would never have coded anything themselves are going to start creating purpose-built custom [software] given the ease of doing so.” Organizations that were your customers yesterday will be your competitors tomorrow, because they can now build what they were buying.

The pitch “we use AI to do X better than existing tools” has a half-life. OpenAI will add a feature, Anthropic will update a model, and Google will ship a product that does what you’re doing natively inside Google Workspace. If your value proposition is that you apply AI to a problem, and AI is now democratically accessible, you are Spotify with a library of 100 million songs that every other platform also has. The question is not whether this will happen, the question is how fast and whether your business model survives it.

Startup failure predictors: “If your differentiation disappears when OpenAI ships a new feature, you don’t have a moat, you have a timer.” The answer to the commodity problem is not to be better at using the commodity, the answer is to own something the commodity cannot replicate.

How Founders Actually Survive This

Return to what Gouveia concludes about music: the artists who survive are the ones building direct ownership of their audience. They are capturing phone numbers, building Discord communities, selling vinyl and merchandise directly to fans who care about them specifically, not about music as a utility. They are shifting their focus from the “ATM machine” model (feed the algorithm, hope for playlist placement, receive fractions of a cent) to ARPF: Average Revenue Per Fan. The question is not how many people stream your song; the question is how much revenue you extract from the people who actually care about you.

For startup founders, the translation is direct. The question is not how many users your AI product has. The question is what proprietary asset (what pipe or at least channel) you own that cannot be replicated by a competitor deploying the same AI tools you’re using.

There are four places where durable competitive advantage still exists in an AI-commoditized world, and they map directly onto the media history we’ve traced.

Own the data. Proprietary data is the closest thing software has to the pipe. If your product generates and compounds data about your specific customer context (their workflows, their organization, their historical patterns) that data cannot be replicated by a competitor starting from scratch. The model may be identical; the data trained on your customer’s specific reality is not. This is why Salesforce isn’t primarily threatened by vibe coded CRM alternatives (which I might remind, are free to make): the switching cost is fifteen years of customer relationship data, not the CRM software itself. If your AI product generates meaningful proprietary data as a byproduct of its use, you have an asset. If it doesn’t, you have a feature.

Own the distribution. Distribution is not a marketing afterthought, it is a structural asset. A startup that has built an engaged audience, through content, community, category-defining thought leadership, or channel partnerships, has something a well-funded competitor cannot simply purchase overnight. The media companies that survived streaming disruption did so because they had recognizable brand assets (HBO, Disney) that commanded consumer loyalty independent of the distribution pipe. If your company is identified with a specific problem, a specific community, or a specific approach to a market, that identity creates distribution pull that the next AI tool doesn’t automatically inherit.

Own the methodology. In a world where the code is a commodity, intellectual property in the traditional software sense means less. Methodological IP, the proprietary process, experience-based curriculum, framework, or workflow your product implements, means more. This is why consulting firms with strong methodologies still exist alongside automated versions of their services. McKinsey’s value is not primarily the PowerPoint slides or the spreadsheet models; it’s the diagnostic methodology and the client trust that methodology has built over decades. If your product is an implementation of a genuinely novel approach to solving a problem, a proprietary methodology that produces better outcomes than competitors’ implementations of the same, that is defensible. If your product is a generic prompt wrapper around an existing model, it is not.

Own the relationship. This is Gouveia’s core argument to artists and one I’m reinforcing to entrepreneurs because it applies to software companies whether you agree with me entirely or not. The platform does not want your customers to have a relationship with you, the platform wants your customers to have a relationship with the platform. If your customers’ primary loyalty is to an app store category rather than to your specific product, you are in the same position as a musician dependent on Spotify’s editorial playlists. You are subject to the algorithm, and the algorithm has no loyalty. Building direct customer relationships (real ones, with high engagement and personal accountability) through an exceptional TEAM that is also a series of irreplaceable relationships, is the most durable and least replicable competitive moat in a commoditized market. It’s also the one most startup founders systematically underinvest in, largely because you don’t know how and you refuse to prioritize marketing, because it’s slow, doesn’t look impressive in a deck, and doesn’t fit neatly into a CAC/LTV spreadsheet.

I made an argument you should read in Why Startups Fail to Get Enough Attention which is the distribution relationship challenge: startups should be “stitching yourself into the ecosystem… strategic alliances, co-marketing partnerships, integrations, distribution deals.” This is not tactical promotion or paid advertising; this is building a pipe. Every integration you complete with a platform your customers already use is a distribution agreement. Every Public Private Partnership you enter with other entities is a multi-layered channel agreement. Every partnership with a community your customers belong to is a networking relationship. Every co-authored piece of research that establishes your company as the authority on your category is a distribution asset.

None of this shows up in a product demo. All of it is what actually determines whether a company survives the transition from novel to commodity.

Content is Valuable but the Pipe is Invaluable

The streaming wars taught us that content is valuable, but the pipe is more valuable. That the platform is where the money is made, but only until someone builds a better pipe. That differentiation evaporates faster than anyone predicts, and that when it does, what remains is whoever owns the customer relationship and whoever controls the distribution.

AI is doing to software what Spotify did to music: democratizing the product to the point where the product itself no longer differentiates. If you are building a company in 2026 whose primary value proposition is that it uses AI, you are building the equivalent of a music streaming service with a library of 100 million songs. I’m sure your songs are amazing! The question is why someone would pay to access them through you rather than through the next service or simply build their own playlist.

Peter Drucker identified that only two things create value in a business: innovation and marketing. In the age of AI commoditization, “using AI” is no longer innovation; it is table stakes, the equivalent of “we use a relational database” or “we have a mobile app.” The innovation is what you do with the AI that nobody else is positioned to do: the proprietary data, the distribution relationship, the methodological advantage, the community ownership. The marketing is how you make that difference legible and compelling to the specific people for whom it matters.

And whoever owns the pipe still wins. That was true in 1927, it was true when Comcast laid cable, it was true when Apple launched the App Store, and it will be true when your well-funded competitor shows up with a vibe coded version of your product next quarter.

The only question worth asking as a founder right now is a simple one: what exactly do you own that the commodity cannot replicate? If you don’t have a clear answer to that, you don’t have a startup, you have a timer.