Right now, AI companies profit much more from wrappers, which pay API prices, instead of regular users, who pay loss-leader subscriptions. There’s little incentive for them to bother competing with wrappers today.
Over time, however, I do think most wrappers are doomed. Prices will inevitably equalize, and AI companies can rapidly capture a lot of market share with little effort — via custom bots, agents and the like. If a wrapper is nothing more than an optimized prompt and a beautiful UI, it will be crushed by the competition one day.
That being said, wrappers that implement proprietary algorithms can still thrive in the long run. Of course, if an app is that good, AI companies might eventually go after it anyway, regardless of how advanced the app is.
>wrappers that implement proprietary algorithms can still thrive in the long run.
This is true, but they need to defend from multiple fronts. Do you believe in Perplexity long-term, or a) OpenAI moving into search, or b) Google moving their search into AI.
It's an interesting hybrid product where I suppose the winner is unclear, because it seems Perplexity has executed fast and strong thus far.
The bitter lesson will probably happen and the wrapper won't need to exist. Just ask the agent for what you want. More like a concierge than a shopping mall.
> regular users, who pay loss-leader subscriptions
who on earth told you about me???? :)
I asked ChatGPT something, and it told me it needs 7 days to create that thing. I asked it to make small incremental dumps every 6 hours, just so I don't get a bunch of BS/hallucinations on Day7. I usually got 4-5 chats running in parallel, so I am definitely over-use.
On the other hand, there are weeks that I don't even log in at all, so..
No B2B customer really cares what technology underpins a business. They never have, and they never will. They care about two things: does it help them save money, or does it help them make money? And does it do either of those things better than the alternatives? Many GPT wrappers do it better than raw dogging GPT, or for cheaper -- especially when factoring in cost of time -- so it's easy to see the value, albeit technology-wise it's also easy to get caught up in the simplicity; but most buyers don't get caught up in such things.
There is existential danger for wrappers. OpenAI is running ads for wrapper-like use cases you can run for free right now.
The AIs will be eventually trainable to be a wrapper by ordinary people. I.e. they ask for the wrapped thing. "Make my photo look good for LinkedIn". It will do a better job.
I am thinking that some years ago everyone with a Windows PC had to buy an antivirus SW and then MS added Defender.
And then (in reverse) Windows had Solitaire and Pinball and now they don't.
Right now ChatGPT does 'ten things'. I believe it will be like Amazon. Allow you to build your shop-in-a-shop, and then once they see that they can jump this number from 'ten' to 'twenty', they will start offering the extra services/products within the same price, and cut the wrappers 'organically'. Unless those wrappers innovate and improve forever. Or until ChatGPT will M&A those small wrappers.
Cursor is not merely a wrapper. There's a lot of optimization going on under the hood. At the very least, it indexes the whole codebase and can search it. That's not including the agent stuff, MCP, context management, etc.
Perplexity might have been a wrapper some time ago, but today, it's far more. I tried asking it for the meaning of the lyrics of a certain song in my head but only gave it the band name, not the song. It picked the wrong band-song at first, but noted that similar questions were asked for the exact song in my head. If I ask GPT/Gemini/etc, it will say, "I know the band, but which song?"
I think this sort of invalidates the rest of the question - most of the things that people just assume are wrappers are not. The things that are only wrappers die out pretty quick because people would just ask ChatGPT instead of Essay Writer 2025.
I don't agree with this word trickery of 'everything is just a wrapper'. The internet, browsers, and rest APIs are much more like scaffolding or 'infrastructure'.
Internet (and/or browser) :: website or webapp
is not the same as
OpenAI/Antropic/model company :: GPT wrapper
There is a substantial qualitative difference, at least at this point.
Right now, AI companies profit much more from wrappers, which pay API prices, instead of regular users, who pay loss-leader subscriptions. There’s little incentive for them to bother competing with wrappers today.
Over time, however, I do think most wrappers are doomed. Prices will inevitably equalize, and AI companies can rapidly capture a lot of market share with little effort — via custom bots, agents and the like. If a wrapper is nothing more than an optimized prompt and a beautiful UI, it will be crushed by the competition one day.
That being said, wrappers that implement proprietary algorithms can still thrive in the long run. Of course, if an app is that good, AI companies might eventually go after it anyway, regardless of how advanced the app is.
Good comment, agree for the most part.
>wrappers that implement proprietary algorithms can still thrive in the long run.
This is true, but they need to defend from multiple fronts. Do you believe in Perplexity long-term, or a) OpenAI moving into search, or b) Google moving their search into AI.
It's an interesting hybrid product where I suppose the winner is unclear, because it seems Perplexity has executed fast and strong thus far.
The bitter lesson will probably happen and the wrapper won't need to exist. Just ask the agent for what you want. More like a concierge than a shopping mall.
> regular users, who pay loss-leader subscriptions
who on earth told you about me???? :)
I asked ChatGPT something, and it told me it needs 7 days to create that thing. I asked it to make small incremental dumps every 6 hours, just so I don't get a bunch of BS/hallucinations on Day7. I usually got 4-5 chats running in parallel, so I am definitely over-use.
On the other hand, there are weeks that I don't even log in at all, so..
No B2B customer really cares what technology underpins a business. They never have, and they never will. They care about two things: does it help them save money, or does it help them make money? And does it do either of those things better than the alternatives? Many GPT wrappers do it better than raw dogging GPT, or for cheaper -- especially when factoring in cost of time -- so it's easy to see the value, albeit technology-wise it's also easy to get caught up in the simplicity; but most buyers don't get caught up in such things.
There is existential danger for wrappers. OpenAI is running ads for wrapper-like use cases you can run for free right now.
The AIs will be eventually trainable to be a wrapper by ordinary people. I.e. they ask for the wrapped thing. "Make my photo look good for LinkedIn". It will do a better job.
Very valid points, agreed.
I suppose my interest is more in the long-term durability in such companies.
I am thinking that some years ago everyone with a Windows PC had to buy an antivirus SW and then MS added Defender.
And then (in reverse) Windows had Solitaire and Pinball and now they don't.
Right now ChatGPT does 'ten things'. I believe it will be like Amazon. Allow you to build your shop-in-a-shop, and then once they see that they can jump this number from 'ten' to 'twenty', they will start offering the extra services/products within the same price, and cut the wrappers 'organically'. Unless those wrappers innovate and improve forever. Or until ChatGPT will M&A those small wrappers.
Cursor is not merely a wrapper. There's a lot of optimization going on under the hood. At the very least, it indexes the whole codebase and can search it. That's not including the agent stuff, MCP, context management, etc.
Perplexity might have been a wrapper some time ago, but today, it's far more. I tried asking it for the meaning of the lyrics of a certain song in my head but only gave it the band name, not the song. It picked the wrong band-song at first, but noted that similar questions were asked for the exact song in my head. If I ask GPT/Gemini/etc, it will say, "I know the band, but which song?"
I think this sort of invalidates the rest of the question - most of the things that people just assume are wrappers are not. The things that are only wrappers die out pretty quick because people would just ask ChatGPT instead of Essay Writer 2025.
Marketing > Product
if you can get a gpt wrapper and get thousands of customer, who is going to say it is not a genuine tech company?
I don't know. Would we have called prior software "internet wrappers" or "rest api wrappers"?
I don't agree with this word trickery of 'everything is just a wrapper'. The internet, browsers, and rest APIs are much more like scaffolding or 'infrastructure'.
Internet (and/or browser) :: website or webapp
is not the same as
OpenAI/Antropic/model company :: GPT wrapper
There is a substantial qualitative difference, at least at this point.
Someone else above/below commented "Marketing --> Product".
I see it that I have a problem, someone solves that problem. I pay $$ per month. Everyone is happy.
I'd say "solution" or "product" and not get stuck on the technical details (for/to which I agree with you but 90% of the people don't care).
When you buy milk you don't buy liquid+container+shipping, you buy food. You consume the food. Belly feels nice. See you tomorrow for another bottle.