There is a pattern we have watched play out across 25 years of working with companies big and small, and it shows up in places you wouldn't expect to look at together. A Google marketer who used the company's own rhetoric as a weapon against it. A novelist standing in front of 200 AI engineers telling them their product was silencing a generation. A prediction market blocked by a government that didn't like what the crowd was saying. A tablet getting its first real upgrade in five years. And a long meditation on what architecture in software actually is. These look like five different stories. They are one story.
The story is about who controls the voice, and what happens when that voice is not yours.
The former Google insider's account in The New Yorker is the cleanest version of it. She got good at amplifying the company's language, its stated mission of organizing the world's information for the benefit of humanity, and then she used that exact language to push back internally when the company drifted from it. The institution's voice became her credential, and then her leverage. But notice what that required: she had to subordinate her own voice first, for years, before she had enough standing to use the borrowed one. And when she finally did use it, the institution didn't reform. She left.
That's the trap. You adopt the voice to get in the room. The voice becomes your identity. And then the room closes.
Dave Eggers didn't play that game. Sam Altman invited him to speak to OpenAI staff, probably expecting something diplomatically challenging but ultimately manageable. Eggers told them directly that ChatGPT was silencing an entire generation of writers. He didn't soften it for the room. He didn't adopt the frame that AI is neutral infrastructure. He showed up with his own voice, said the uncomfortable thing, and left them with it. He had nothing to gain by performing for them. That's the difference. He wasn't trying to keep his seat at the table.
We've worked with founders who have been through this exact compression. Not at Google scale. At the scale of a $2M business that starts chasing a bigger client, or a grant, or a distribution partnership, and slowly starts writing emails that sound like that client. Starts pitching in that client's language. Starts measuring success by that client's metrics. And then one day they look up and their actual customers, the ones who found them because they were different, are gone. They traded the voice that built the business for the voice they thought would scale it.
The architectural framing from Uwe Friedrichsen's piece is useful here. The argument is that real architectural work isn't about drawing boxes and arrows. It's about holding decisions under pressure. Knowing which constraints are load-bearing and which are just habits. Most teams can't tell the difference until something breaks. The same is true for brand voice. Founders almost always know, in their gut, which parts of how they communicate are load-bearing and which are just noise. But they abandon the load-bearing parts first, because those are the parts that feel risky in a pitch.
The voice that made you strange is the voice that made you valuable. The moment you sand it down to fit the room, you become a commodity.
France's crackdown on Polymarket is the government version of this same impulse. A prediction market lets crowds say things institutions don't want said, prices things institutions don't want priced. The response is to classify it as gambling and block access at the ISP level. The mechanism doesn't matter. What matters is the pattern: when an independent signal gets too loud and too accurate, the institution finds a category to put it in and then restricts that category. It's not about gambling. It's about who gets to be the authoritative voice on what will happen next.
You, as a founder, are a prediction market. Your positioning is a bet. Your pricing is a bet. Every service you offer is a claim about what your customers actually need versus what the mainstream thinks they need. The moment you let an institution, a platform, a big client, a VC, reclassify what you do into their language and their category, they have won the same way France won. You're still technically operating, but inside a frame that isn't yours.
The iPad mini is a smaller version of this. It spent five years barely updated, languishing in Apple's lineup while the company focused on the iPad Pro. It survived because a specific group of users refused to accept that bigger was better. They kept buying the old one. They made enough noise on forums and in reviews that Apple couldn't quietly kill it. It persisted because its constituency was loud and specific about what they wanted. Now it's getting OLED and a real processor upgrade. The niche held.
Your niche will hold if you hold it. But you have to keep saying the thing that makes you specific. Not the thing that makes you broadly appealing. The thing that makes your actual people feel seen, even if it makes everyone else scroll past.
Here's the practical version of all of this. When we work with founders on their positioning and their web presence, we watch for the same thing every time: the point in the copy where they stopped sounding like themselves and started sounding like their category. It's almost always visible. It's the sentence that's technically accurate but has no heat. The paragraph that could have been written by a competitor. The headline that's "safe."
That's the architectural decision. What do you protect? Not from a style guide perspective. From a survival perspective. If you strip your voice down to fit a room, you may get in the room. But you will not get the contract. The room will have fifteen versions of the safe thing. They won't need another one.
The founders who close deals are the ones who still sound like themselves in the pitch. The ones who say something that makes the room slightly uncomfortable, and then say it better. That's the Eggers move. That's the only move that actually works at scale, because scale without distinctiveness is just overhead.
Know which parts of your voice are load-bearing. Don't let anyone convince you to move that wall.