Subtle distinction between:
"building apps for AI to use"
and
"becoming data vendor/supplier to AI"
.... ChatGPT "hates" your app. If they could skip using it and directly provide information, they would.
The "chat to Apps" is a bandaid solution.
At some point the "app"will not exist.
You'll create an MCP (and API) that LLMs will use.
No app.
No interface.
No UI.
Nothing to download.
No installation.
For example, with e-commerce integration (Shopify/Stripe), soon, how many stores will see drop in visits without drop in revenue (hopefully)?
i love anish so i will respectfully disagree here.
when jobs launched the app store, there was no general reasoning underneath iphone. here, chatgpt already has world knowledge, context, & intelligence baked in. every long tail app is already baked into gpt or better yet it can simply generate it.
so with this in mind most valuable “apps” will just be thin wrappers around missing data. if you’re zillow, that’s housing data. if you’re canva, maybe that’s design templates (although openai can likely make posters already).
it’s not really an application play.. it’s a data play. & giving your data to openai is effectively deepening their moat, since every app built here trains their system to do what you do. users stay in chatgpt. your edge decays over time.
in essence devs are feeding openai context it doesn’t yet have, which then increases openai’s moat. the more useful their plugin, the faster their own differentiation evaporates.
we’ll see how this plays out but i think it’s imho it’s not yet equivalent to app store even with 100x distribution.