The last mile of AI app-building is sharing
AI made building a small app almost free. But the moment you want someone else to use it, all the old hard parts come back. The bottleneck has moved, and that's where the opportunity is.
A year ago, building a small custom app meant hiring someone or learning to code. Today you describe what you want to Claude or ChatGPT and a working app appears in the time it takes to write a paragraph. That part, the part that used to be the whole job, has collapsed to almost nothing.
Which means the bottleneck has moved. It’s no longer building the app. It’s everything that happens in the thirty seconds after it works.
The build got easy. The last mile didn’t.
Watch anyone build something with an AI assistant and you’ll see the same arc. Delight while it comes together. A working thing on the screen. And then a pause, followed by some version of the same question:
“Okay, how do I actually let other people use this?”
That pause is the last mile. The app runs perfectly, for exactly one person, on exactly one screen. To get it to a second person you suddenly need the things AI didn’t hand you: a place to host it, a way to let the right people in, and somewhere for the data to live. None of that got easier just because the building did.
So the cruel joke of the current moment is that we’ve made it trivial to build apps that are almost impossible to share. The value is all dressed up with nowhere to go.
Why the last mile is genuinely hard
It’s not that hosting is mysterious. It’s that “share this app” quietly contains three separate problems, each of which used to be a developer’s job:
- Hosting. The file has to live somewhere with an address other people can reach.
- Identity. Most useful apps aren’t for the whole internet. A family budget, a team roster, a client form: these need to know who’s allowed in. That means sign-in and an allowlist.
- Persistence. An app people use has to remember things. The moment data matters, you need storage that’s safe, isolated per user, and doesn’t disappear when the tab closes.
Any one of these is a small project. Together they’re the reason a ten-minute app turns into a weekend, or simply never gets shared at all. The person who built it isn’t a backend engineer. They’re a parent, a team lead, someone with a side hustle. And the last mile asks them to become one.
The tools we reach for weren’t built for this
The instinct is to grab a host. But the popular ones were built for a different person. Netlify and Vercel are build pipelines for developers; they’ll host your files and then expect you to bring your own auth and database. Supabase and Firebase will give you a backend, a whole one, that you design and maintain. GitHub Pages will serve a static file and nothing else.
None of them are wrong. They’re just aimed upstream of where the new bottleneck is. They assume you’re running a project. But the person finishing an app in a chat isn’t running a project. They have a finished thing and a simple need: let other people use it.
What “solving the last mile” looks like
If the building now happens in a conversation with an assistant, the sharing should too. The last mile is solved when the same assistant that wrote your app can also publish it, host it, add a login, attach storage, without you leaving the chat or learning a new tool.
That’s the bet behind Backlit. You connect it to Claude or ChatGPT once, and from then on you can just say “publish this and give my team access.” The assistant does the last mile for you: a live link, sign-in for the right people, saved data, in seconds. The three hard problems don’t go away. They get absorbed into the place the app lives, so the person sharing it never has to see them.
And the running app stays simple. There’s no AI inside it. The intelligence was spent on building it. What’s left is plain, fast, durable HTML with a backend it didn’t have to ask for. That restraint is the point: the last mile shouldn’t add weight; it should remove it.
The opportunity
Every shift in how software gets made creates a new bottleneck, and the bottleneck is where the next product lives. Cloud hosting made deployment easy and app stores made distribution easy. AI just made creation easy (for everyone, not just developers), and the new constraint is the gap between a working app and a shared one.
Close that gap and you don’t just save people a weekend. You change what’s worth building at all. When sharing is free, the calculus flips: the tiny, personal, almost-not-worth-it app (the one for your household, your five-person team, your handful of customers) suddenly is worth it, because getting it to the people who need it costs nothing.
The build is done. The last mile is the product.
Built an app with Claude or ChatGPT? Get early access to Backlit and share it in seconds.