The Personal App Store
Ship one-shot custom apps to your own phone instead of settling for someone else's.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 63%
The loop replaces the traditional build pipeline with a description-to-installed-app path. You start from a use case that mainstream software serves badly for you, then one-shot it: a single dense prompt that names the reference products to borrow functionality from, the design conventions to follow, the data schema in your own terms, and the integrations you want. The agent returns a working app. The second half is delivery: Naval built a personal app store, a web page that became an app on his own phone, so builds arrive somewhere he can install and upgrade with one click. From there iteration is conversational. The activation energy that used to sit in wiring services together disappears, which shifts the bottleneck entirely onto knowing what you actually want.
Origin
Extracted from Naval, where Naval describes rebuilding the app he had previously shipped with a team of eight or nine engineers, this time alone and without compromises.
Core principles
- 01Custom beats best-in-class only where your use case is genuinely narrow or private.
- 02Clarity of vision is the binding constraint, not coding ability.
- 03Software built alone carries no compromise tax from other people's preferences.
- 04An agent has no ego, so your idiosyncratic requests cost nothing socially.
- 05Delivery infrastructure matters as much as generation; an app you cannot install is a demo.
- 06Prototypes are the goal here, not scalable production systems.
How to run it
- 1
Pick a use case the market underserves for you
Choose something niche, private, or tuned to a habit only you have. Broad use cases already have hand-tuned best-of-breed apps you will not beat.
Watch out If a mainstream app already covers your need well, building your own is entertainment, not leverage.
- 2
Get the vision explicit before prompting
Write out exactly what you want it to do and how it should feel. Naval names clear direction as the hardest part of the whole exercise.
Pro tip Borrow specificity from products you already admire; naming them is cheaper than describing them.
Watch out Vague direction produces a generic app that is worse than what you could have downloaded.
- 3
One-shot it with a loaded prompt
Put reference apps, design guidelines, your existing data, the charts you want and the integrations into one substantial prompt, then ask for the whole app back.
Pro tip Include your real historical data in the prompt so the first build is immediately usable rather than empty.
- 4
Build the delivery path once
Create a personal app store: a page the agent ships builds into, installable on your own device with one click and upgradeable like a normal store.
Pro tip The same path lets you hand builds to friends and family, which is where distribution stops.
Watch out Device signing rules mean you cannot distribute these to the general public.
- 5
Iterate conversationally
Describe the change you want in plain language and let the agent implement it. Because there is no engineer to annoy, tiny aesthetic adjustments are free.
Pro tip Run the agent from your phone against a remote terminal so ideas can be built the moment you have them.
- 6
Know where the loop ends
This produces prototypes fast and true to your vision, but with weak architecture and possible security holes. Anything going to market at scale needs real engineers and probably a rewrite.
Watch out Do not confuse a working personal app with a production system.
In the wild
Naval wanted a workout app exactly to his own taste. In roughly one prompt he asked the agent to borrow functionality from Tonal and Ladder, follow Apple's human interface guidelines so it looked native, ingest a text log of his recent workouts, make re-entry and adjustment easy, generate progress graphs, calculate strength scores by reading scientific papers on how to score by body part, render a human-body diagram showing which muscles were larger, and connect to Apple Health for heart rate. A working app was delivered to his personal app store immediately.
→ A fully custom fitness tracker built in one pass and installed on his phone, with no engineering team and no compromises.
Naval had spent about a year on AirChat, a voice and video social messenger built with eight or nine engineers over nine to twelve months, which was eventually sold off. He is now rebuilding it from scratch alone. The difference he emphasises is not speed but fidelity: with a team, even a decisive founder accommodates other people and cannot keep asking for an icon to move left, right, then back again. With an agent there is no social cost to idiosyncrasy, so the result matches the original vision exactly.
→ A rebuild that is true to one person's vision, with the compromise tax of team development removed.
Common mistakes
Building what the App Store already nails
Broad use cases have products somebody hand-tuned and slaved over. Aiming your custom build at those wastes the advantage, which lies in niche and private needs.
Prompting before you know what you want
The agent will build whatever you describe, so an unclear vision yields a mediocre app fast. Clarity of direction is the actual scarce input.
Mistaking the prototype for a scalable product
This generation of output can carry weak architecture and security holes. Taking it to many users means recruiting real engineers and expecting to rewrite it.
Is it for you?
Best for
Creative, self-motivated, articulate people with a clear vision of a niche or private tool they want for themselves and a small circle.
Not ideal for
Broad, mainstream use cases already served by hand-tuned best-in-breed apps, or products that must scale to many users on day one.
From the transcript
“So I actually built my own little app store, which is an app store just for me.”
“But it really, really helps to have a clear direction. You have to know what you want. That's actually the hardest thing.”
“I give it a two line description, it builds me an app, it ships it to my App Store.”
From the episode
A Return to Code