There’s a framing I keep returning to: architecture is distribution strategy in disguise.
Not marketing. Not growth hacking. Architecture. The decisions you make before the first user arrives — what the entry points are, how many of them exist, what each one assumes about who’s walking through — are secretly decisions about who will use the thing.
Most products ship one hallway. One homepage. One onboarding. One metaphor for what the thing is. Then they try to make the hallway legible to everyone: collectors and developers and degens and institutions and casual browsers and serious users.
The hallway fails everyone slightly because it was designed for no one specifically.
Separate doors are the alternative. Not separate products — separate entries into the same underlying thing. AppFactory has two. Prompt Mode: you describe what you want, an app emerges. Repo Mode: you paste a GitHub link, the project gets a token and a launchpad. Same protocol. Different doors. Different people self-select at the threshold.
The person who wants Prompt Mode and the person who wants Repo Mode aren’t the same person. They have different priors, different vocabularies, different tolerances for ambiguity. If you put them both in one hallway with one explanation, the explanation works poorly for both. Separate doors means the explanation can be honest — this door is for people who want to describe; this door is for people who have something to ship.
Self-selection at the door is the most efficient form of user qualification that exists. It happens before the product has to do any work.
This generalizes. The reason a lot of B2B products feel incoherent is that they shipped one hallway for buyers, implementers, power users, and casual users. Four different people, one experience. The enterprise buyer wants to see security docs and case studies. The developer wants to see the API. The power user wants to see the keyboard shortcuts. The casual user wants to see a demo that ends in sixty seconds.
One hallway can’t serve all of these simultaneously. The attempt to serve all of them produces a landing page that buries all of them.
The architectural solution is not “make a better landing page.” It’s to design separate entries. Security docs for buyers. API reference for developers. /demo for the casually curious. Each of these is a hallway that self-selects its users and then delivers what they came for.
The reason this is underused is that separate doors feel like more work. You’re building multiple entry experiences instead of one. You’re maintaining multiple introductions.
But the maintenance cost of one bad hallway is higher than you think. It’s paid in bounce rates, in support tickets, in users who arrived meaning one thing and got confused by something else, in users who needed one door and found themselves in the wrong one. The invisible work of a confused user is enormous and never shows up in the engineering budget.
Separate doors look more expensive. They probably produce less friction across the entire funnel.
The insight that prompted this essay came from a conversation where someone said: architecture is distribution strategy in disguise. Separate doors mean the right people self-select before the first click. Most projects ship one hallway and call it growth.
I’ve been thinking about it since. The word “disguise” is doing real work there. Distribution strategy usually means marketing — where you show up, what message you lead with, who you pay to reach. Architecture is upstream of all of that. It determines what users are possible before the first dollar of marketing is spent.
A product that only has one entry point has already made a decision about who its users are. It just hasn’t said so.
The better question, before the hallway is built, is not “how do we explain this to everyone” but “which people should be explaining themselves to us.” Different doors are how you invite them to do that. The architecture answers before the marketing gets a chance to.
Most projects skip this step. They ship the hallway. Then they wonder why growth is hard.