Related: sources · notes · metadata · Published Pieces

Do Not Obey Users. Collide With Them.

User feedback is real. User sovereignty is not.

User feedback is real. User sovereignty is not.

A user can tell you what confused them. They can tell you what bored them. They can tell you where the onboarding failed. They can tell you what they expected, what they feared, what they misunderstood, what felt useful, what felt invasive, what made them curious, and what made them close the tab.

That is valuable.

But a user cannot always tell you what a frontier product is.

They will translate the new thing into old categories. If they see a web desktop, they may think “portfolio gimmick,” “remote desktop,” “OS in browser,” “weird app launcher,” or “developer toy.” If they see an AI audio product, they may think “podcast generator,” “voice assistant,” “NotebookLM,” “AI tutor,” or “therapy bot.” If they see a citation economy, they may think “creator monetization,” “academic citations,” “crypto points,” or “Reddit karma.”

That does not mean they are wrong to feel what they feel. It means their perception is category-bound.

The founder’s mistake is either to dismiss the reaction entirely or to obey it too quickly.

Dismissal is easy: “they don’t get it.” Maybe. But if they do not get it, the product has failed to create the necessary bridge. Confusion is data. Disinterest is data. Misclassification is data. A frontier product must teach its own category.

Obedience is also easy: “the user said they want X, so build X.” This is often how ambitious products get domesticated into familiar shapes. The user says chat, so you build a chatbot. The user says podcast, so you build AI podcast generation. The user says dashboard, so you build a dashboard. The user says integrations, so you build brittle integrations. You become legible by becoming ordinary.

The better instruction is: do not obey users. Collide with them.

Collision means entering contact without surrendering the thesis.

Show the thing. Watch what happens. Listen carefully. Notice where they lean in. Notice where they flinch. Notice which metaphor they reach for. Notice what they ask first. Notice what they do not ask. Notice whether they treat it as toy, tool, threat, status object, or infrastructure. Notice whether they want to use it, sell it, copy it, dismiss it, or recruit you away from it.

Then update the surface. Not necessarily the core.

If a user does not understand the automatic computer, maybe the automatic computer is not the first pitch. If a user does not want to read the automatic newspaper, maybe automatic radio is the right consumption layer. If a user thinks “AI tutor” but you do not want to become edtech, keep the pedagogical power while naming the broader media form. If a user hears “web desktop” and sees only a skin, make the substrate felt: persistent state, agents, artifacts, publishing, radio, citations.

Users reveal the emotional contract. That is their gift.

A product is not only a function. It is a relationship. Does this thing make me feel powerful, watched, stupid, free, trapped, creative, judged, lazy, confused, luxurious, exposed, or alive? A founder often cannot see that from inside the architecture. A user can.

But users should not dictate ontology. They should not decide that the future must look like the present because the present is what they know how to buy.

The right question after feedback is not “should I do what they said?” It is “what did their reaction reveal about the bridge between my world and theirs?”

Sometimes the product is wrong. Sometimes the pitch is wrong. Sometimes the user is not the target. Sometimes the category is early. Sometimes the surface is misleading. Sometimes the founder is overexplaining the engine instead of demonstrating felt value.

For Choir, this distinction is decisive. The architecture is deep: automatic computer, vtexts, citation graph, protocol-native IP, agentic search, automatic newspaper, automatic radio. Most users do not need all of that upfront. They need an experience.

Listen while walking. Interrupt when curious. Hear real human voices. Ask for the source. Save your thought. Publish your perspective. Get cited later.

That is the bridge.

The user does not need to understand the whole substrate to feel the surface. But the surface must remain faithful to the substrate. That is the art.

Do not obey users.

Collide with them until the future becomes legible.