This conception makes sense iff you believe in ChatGPT as the universal user interface of the future. If anything the agentic wave is showing that the chat interfaces are better off hidden behind stricter user interface paradigms.
I suspect there are many, many things for which chat is a great interface. And by positioning ChatGPT as the distributor for all these things, they get to be the new Google. But you're also right that many domains for which a purpose-built interface is the right approach, and if the domain is valuable enough, it'll have someone coming after it to build that.
I have yet to see a chat agent deployed that is more popular than tailored browsing methods. The most charitable way to explain this is that the tailored browsing methods already in place are the results of years of careful design and battle testing and that the chat agent is providing most of the value that a tailored browsing method would but without any of the investment required to bring a traditional UX to fruition - that may be the case and if it is then allowing them the same time to be refined and improved would be fair. I am skeptical of that being the only difference though, I think that chatbots are a way to, essentially, outsource the difficult work of locating data within a corpus onto the user and that users will always have a disadvantage compared to the (hopefully) subject matter experts building the system.
So perhaps chatbots are an excellent method for building out a prototype in a new field while you collect usage statistics to build a more refined UX - but it is bizarre that so many businesses seem to be discarding battle tested UXes for chatbots.
agree.
Thing is, for those who paid attention to the last chatBot hype cycle, we already knew this. Look at how Google Assistant was portrayed back in 2016. People thought you'd be buying starbucks via the chat. Turns out the starbucks app has a better UX
Yea, I don't want to sit there at my computer, which can handle lots of different input methods, like keyboard, mouse, clicking, dragging, or my phone which can handle gestures, pinching, swiping... and try to articulate what I need it to do in English language conversation. This is actually a step backwards in human-computer interaction. To use an extreme example: imagine instead of a knob on my stereo for volume, I had a chat box where I had to type in "Volume up to 35". Most other "chatbot solved" HCI problems are just like this volume control example, but less extreme.
It's funny, because the chat bot designers seem to be continually attempting to recreate the voice computer interface from Star Trek: TNG. Yet if you watch the show carefully, the vast majority of the work done by all the Enterprise crew is done via touchscreens, not voice.
The only reason for the voice interface is to facilitate the production of a TV show. By having the characters speak their requests aloud to the computer as voice commands, the show bypasses all the issues of building visual effects for computer screens and making those visuals easy to interpret for the audience, regardless of their computing background. However, whenever the show wants to demonstrate a character with a high level of computer mastery, the demonstration is almost always via the touchscreen (this is most often seen with Data), not the voice interface.
TNG had issues like this figured out years ago, yet people continue to fall into the same trap because they repeatedly fail to learn the lessons the show had to teach.
Remember Alexa? Amazon kept wanting people to buy things with their voice via assorted echo devices, but it turns out people really want to actually be in charge of what their computers are doing, rather than talking out loud and hoping for the best.
“volume up to 35”
>changes bass to +4 because the unit doesn't do half increments
“No volume up to 35, do not touch the EQ”
>adjusts volume to 4 because the unit doesn’t do half increments
> I reach over, grab my remote, and do it myself
We have a grandparent that really depends on their Alexa and let me tell you repeatedly going “hey Alexa, volume down. Hey Alexa, volume down. Hey Alexa, volume down,” gets really old lol we just walk over and start using the touch interface
It's also a matter of incentives. Starbucks wants you in their app instead of as a widget in somebody else's - it lets them tell you about new products, cross-sell/up-sell, create habits, etc.
This general concept (embedding third parties as widgets in a larger product) has been tried many times before. Google themselves have done this - by my count - at least three separate times (Search, Maps, and Assistant).
None have been successful in large part because the third party being integrated benefits only marginally from such an integration. The amount of additional traffic these integrations drive generally isn't seen as being worth the loss of UX control and the intermediation in the customer relationship.
> the tailored browsing methods already in place are the results of years of careful design and battle testing
Have you ever worked in a corporation? Do you really think that Windows 8 UI was the fruit of years of careful design? What about Workday?
> but it is bizarre that so many businesses seem to be discarding battle tested UXes for chatbots
Not really. If the chatbot is smart enough then chatbot is the more natural interface. I've seen people who prefer to say "hey siri set alarm clock for 10 AM" rather than use the UI. Which makes sense, because language is the way people literally have evolved specialized organs for. If anything, language is the "battle tested UX", and the other stuff is temporary fad.
Of course the problem is that most chatbots aren't smart. But this is a purely technical problem that can be solved within foreseeable future.
> I've seen people who prefer to say "hey siri set alarm clock for 10 AM" rather than use the UI. Which makes sense, because language is the way people literally have evolved specialized organs for.
I don't think it's necessary to resort to evolutionary-biology explanations for that.
When I use voice to set my alarm, it's usually because my phone isn't in my hand. Maybe it's across the room from me. And speaking to it is more efficient than walking over to it, picking it up, and navigating to the alarm-setting UI. A voice command is a more streamlined UI for that specific task than a GUI is.
I don't think that example says much about chatbots, really, because the value is mostly the hands-free aspect, not the speak-it-in-English aspect.
> I've seen people who prefer to say "hey siri set alarm clock for 10 AM" rather than use the UI.
It's quicker that way. Other things, such as zooming in to an image, are quicker with a GUI. Bladerunner makes clear how the voice UI is poor for this compared to a GUI.
In an alarm, there is only one parameter to set. In more complex tasks, chat is a bad ui because it does not scale well and it does not offer good ways to arrange information. Eg if I want to buy something and I have a bunch of constraints, I would rather use a search-based UI where i can fast tweak these constraints and decide. Chathpt being smart or not here is irrelevant, it would just be bad ui for the task.
You're thinking in wrong categories. Suppose you want to buy a table. You could say "I'm looking for a €400 100x200cm table, black" and these are your search criteria. But that's not what you actually want. What you actually want is a table that fits your use case and looks nice and doesn't cost much, and "€400 100x200cm table, black" is a discrete approximation of your initial fuzzy search. A chatbot could talk to you about what you want, and suggest a relevant product.
Imagine going to a shop and browsing all the aisles vs talking to the store employee. Chatbot is like the latter, but for a webshop.
Not to mention that most webshops have their categories completely disorganized, making "search by constraints" impossible.
Funny, I almost always don't want to talk to store employees about what I want. I want to browse their stock and decide for myself. This is especially true for anything that I have even a bit of knowledge about.
The thing is that "€400 100x200cm table, black" is just much faster to input and validate versus a salesperson, be it a chatbot or an actual person.
Also, the chatbot is just not going to have enough context, at least not in it's current state. Why those measurements? Because that's how much room you have, you measured. Why black? Because your couch is black too (bad choice), and you're trying to do a theme.
That's kind of a lot to explain.
> Of course the problem is that most chatbots aren't smart. But this is a purely technical problem that can be solved within foreseeable future.
Ah yes, it's just a small detail. Don't worry about it.
I agree with you. I think chat interfaces are really good with voice interfaces while walking, asking for a foreign language lesson, effectively doing a web search while walking by speaking and listening to the answer.
Other app-like interfaces like NotebookLM can be useful, for me one or two real uses a week.
Then there is engineering small open models into larger systems to do structured data extraction, etc.
I am skeptical about the current utility of agentic systems, MCP, etc. - even though I like to experiment.
Someone else said that at least the didn’t go on and on about AGI today - a nice thing. FOMO chasing ASI and AGI will drive us bankrupt, and produce some useful results.
>If anything the agentic wave is showing that the chat interfaces are better off hidden behind stricter user interface paradigms.
I'm not sure that claim is justified. The primary agentic use case today is code generation, and the target demographic is used to IDEs/code editors.
While that's probably a good chunk of total token usage, it's not representative of the average user's needs or desires. I strongly doubt that the chat interface would have become so ubiquitous if it didn't have merit.
Even for more general agentic use, a chat interface allows the user the convenience of typing or dictating messages. And it's trivially bundled with audio-to-audio or video-to-video, the former already being common.
I expect that even in the future, if/when richer modalities become standard (and the models can produce video in real-time), most people will be consuming their outputs as text. It's simply more convenient for most use-cases.
Having already seen this explored late '24, what ends up happening is that the end user generates apps that have lots of jank, quirks, and logical errors that they lack the ability to troubleshoot or resolve. Like the fast forward button corrupting their settings config, the cloud sync feature causing 100% CPU load, icons gradually drifting away from their original positions on each window resize event, or the GUI tutorial activating every time they switch views in the app. Even worse, because their app is the only one of its kind, there is no other human to turn to for advice.
i can't imagine that users will be interested in asking chatGPT to ask zillow things, or ask chatGPT to ask canva to do things. that's a clunky interface. i can see users asking chatGPT to look up house prices, or to generate graphics, but they're not going to ask for zillow or canva specifically.
and if the apps are trusting ChatGPT to send them users based on those sort of queries, it's only a matter of time before ChatGPT brings the functionality first-party and cuts out the apps - any app who believes chat is the universal interface of the future and exposes their functionality as a ChatGPT app is signing their own death warrant.
This x1000. Are businesses short sided enough to help create and develop another wallet garden just like "Google" and "Amazon" are right now? Time will tell but I think businesses want to own their sales funnel, not just give the user a way to avoid interacting with them.
Every company should see OpenAi as a threat. They absolutely will come for you when the time comes.
It's just like Google and websites, but much more insidious. If they can get your data, they'll subsume your function (and revenue stream).
There's a lot of appropriate blowback against stupid AI hype and I'm all for it. But I do think in many respects it's a better interface than (1) bad search results, (2) cluttered websites, (3) freemium apps with upgrade nags, as well as the collective search cost of sorting through all those things.
I remember reading some not-Neuromancer book by William Gibson where one of his near-future predictions was print magazines but with custom printed articles curated to fit your interests. Which is cool! In a world where print magazines were still dominant, you could see it as a forward iteration from the magazine status quo, potentially predictive of a future to come. But what happened in reality was a wholesale leapfrogging of magazines.
So I think you sometimes get leapfrogging rather than iteration, which I suspect is in play as a possibility with AI driven apps. I don't think apps will ever literally be replaced but I think there's a real chance they get displaced by AI everything-interfaces. I think the mitigating factor is not some foundational limit to AI's usefulness but enshittification, which I don't think used to consume good services so voraciously in the 00s or 2010s as it does today. Something tells me we might look back at the current chat based interfaces as the good old days.
I think you need to be careful here because you shouldn't be comparing chat apps to the current state of search results. Instead you compare it to the ideal or to the state of them before companies decided that instead of providing what people are looking for it was more profitable to provide them with related content that they're paid to show.
We are at a moment where we're trying to figure out how to design good interfaces, but very soon after that the moment of "okay, now let's start selling with them" will come and that's really what we're going to be left with.
In that regard, things like adblockers which now a days can be used to mitigate some of these defects you talk about are probably going to be much more difficult to implement in a chat-app interface. What are we going to do when we ask an agent for something and it responds with an ad rather than the relevant information we're seeking? It seems to me like it's going to be even more difficult to be in control for the user.
at least with bad search results, you had to look at them to know they were bad or become used to certain domains that you could prejudge the result and move to the next one. LLMs confidently tell you false/made up information as fact. If you fail to follow up with any references and just accept result, you are very susceptible to getting fooled by the machine. Getting outside of the tech bubble echo chamber that is HN, a large number of GPT app users have never heard of hallucinations or any of the issues inherit with LLMs.
It's not just as ChatGPT as the interface. It's that Chat with AI will now be the universal interface and every tech company will have their version of it. Everything you want to do will happen in one place. Cards will provide predefined and interactive experience. Over time you'll see entirely dynamic content get generated on the fly. The user experience is going to be one where we've shrunk websites to apps and apps to cards or widgets. Effectively any action you need to take can be done like this and then agents can operate more complex workflow in the background. This is probably the interface for the next 10 years and what replaces the mobile app experience and stronghold that Apple or Google have. This lasts until fully immersive AR/VR become a more mainstream thing. At that point these cards are on a heads up display but we'll be looking at something totally different. Like agents roaming the earth...
This has been the pitched playbook for decades. (Metamates!) I'm increasingly convinced its driven by a specific generation of tech entrepreneurs who cut their teeth while reading ca. 1980s science fiction.
I could see chat apps becoming dominant in Slack-oriented workplaces. But, like, chatting with an AI to play a song is objectively worse than using Spotify. Dynamically-created music sounds nice until one considers the social context in which non-filler music is heard.
The thing it reminds me of is those old Silicon Graphics greybeards that were smug about how they were creating tools for people that created wealth when those other system providers "just" created tools for people tracking wealth.
There's a whole bizarre subculture in computing that fails to recognize what it is about computers that people actually find valuable.
Chatting with an AI to play a song whose title you know, sure.
Getting an AI to play "that song that goes hmm hmmm hmmm hmmm ... uh, it was in some commercials when I was a kid" tho
> Getting an AI to play "that song that goes hmm hmmm hmmm hmmm ... uh, it was in some commercials when I was a kid" tho
Absolutely. The point is this is a specialised and occasional use case. You don't want to have to go through a chat bot every time you want to play a particular song just because sometimes you might hum at it.
The closest we've come to a widely-adopted AR interface are AirPods. Critically, however, they work by mimicing how someone would speak to a real human by them.
more abstract than that, "I'm throwing a wedding/funeral/startup IPO/Halloween/birthday party for a whatever year old and need appropriate music". Or, without knowing specific bands, "I want to hear some 80's metal music". "more cowbell!"
You don't need AI for this, Spotify has, like, infinite playlists.
Also their playlists are made by real people (mostly...), so they don't completely suck ass.
Playlists aren't interactive though. I can't say "like this but with less guitar".
Also, following the Beatport top 100 tech house playlist, and hearing how many tracks aren't actually tech house makes me wonder about who makes that particular playlist.
It's because Zuck can't own a pane of glass. He's locked out of the smartphone duopoly.
Everyone wants the next device category. They covet it. Every other company tries to will it into existence.
I agree with what you are saying.
I’m building a tool that helps you solve any type of questionnaire (https://requestf.com) and I just can’t imagine how I could leverage Apps.
It would be awesome to get the distribution, but it has to also make sense from the UX perspective.
Your link is broken?
> conception makes sense iff you believe in ChatGPT as the universal user interface of the future
Out of curiosity, why iff?
Is it? Honestly, most agents and/or ai apps I interact with that are actually useful present some form of chat-like interface.
I’m not very bullish on people wanting to live in the ChatGPT UI, specifically, but the concept of dynamic apps embedded into a chat-experience I think is a reasonable direction.
I’m mostly curious about if and when we get an open standard for this, similar to MCP.
The whole value of an actual executive assistant is them solving problems and you not micromanaging them.
What users want, which various entities religiously avoid providing to us, is a fair price comparison and discovery mechanism for essentially everything. A huge part of the value of LLMs to date is in bypassing much of the obfuscation that exists to perpetuate this, and that's completely counteracted by much of what they're demonstrating here.
Yes, I certainly prefer "chatting with Claude Code" to "Copilot taking forever to hallucinate all over my IDE, displacing the much-more-useful previous-generation semantic autocomplete."
The former is like a Waymo, the latter is like my car suddenly and autonomously deciding that now is a good time to turn into a Dollar Tree to get a COVID vaccine when I'm on my way to drop my kid off at a playdate.
The apps can send any arbitrary HTML / interface back though.
e.g. Coursera can send back a video player
Once it's efficient enough, you will be able to just vocally talk to your computer to do all of this. Text chat is just the simplest form of a natural language interface, which is obviously the future of computing.
I don’t think natural language is efficient enough. Whether that be text or voice.
I imagine the Star Trek vision is pretty accurate. You occasionally talk to the computer when it makes sense, but more often than not you’re still interacting with a GUI of some kind.
The ChatGPT phone app has had voice conversation mode for a while now. it's more interactive than a podcast while driving. There are apps (Wispr, non-affiliated) to make talking to your computer easier. The future is definitely a hybrid of them. sometimes I want to talk, other times I want to type.
I think this is very interesting, but it is reminiscent of what we built with Phind 2 where the answer could include dynamic, pre-built widgets.
The problem with this approach is precisely that these apps/widgets have hard-coded input and output schema. They can work quite well when the user asks something within the widget's capabilities, but the brittleness of this approach starts showing quickly in real-world use. What if you want to use more advanced filters with Zillow? Or perhaps cross-reference with StreetEasy? If those features aren't supported by the widget's hard-coded schema, you're out of luck as a user.
What I think it much more exciting is the ability to completely create generative UI answers on the fly. We'll have more to say on this soon from Phind (I'm the founder).
That’s solved by MCP though. You can update your MCP’s servers schema dynamically without ever having to touch the app itself but the app will be aware of the new schema.
I'm not saying that the schema can't change from time to time, I'm saying that having any fixed schema at request time is not an ideal user experience because it may not be clear what is supported and what is not supported. From first principles, it's much better if the app schema can be created dynamically at request time so we can guarantee that we can fully serve the user's request exactly as they asked it.
Phind is awesome. I often forget to use it until legacy search engines fail to surface what I’m looking for after a dozen searches. Phind usually finds it.
That said, I used it a lot more a year ago. Lately I’ve been using regular LLMs since they’ve gotten better at searching.
Thanks for the feedback. I think that our main differentiator going forward will be this generative UI on the fly for answering questions as opposed to search alone.
In a similar boat, but have been increasingly returning to for its quick notebook/charting capabilities. Would be awesome to somehow be able to select between different UI modes offering search, ranking, graphing or else depending on user needs.
Given there is already a MCP-UI project, I’m not surprised it can be done. But even that I’m not very convinced that it’s the right approach. After all, it’s still far too slow for real usage…
Totally agree that it's too slow with conventional approaches, which is why we're training custom models for this that we can run fast
Ah, that’s interesting. I’m considering building something similar for our product, and my solution to the schema constraints you mentioned thus far is breaking my widgets into blocks as universal as possible, as to still be useful. All of this is just ideas yet mind you, but my thinking was—maybe I can get the model to pick from a range of composable widgets depending on the task that are interoperable?
For a concrete example, think a search result listing that can be broken down into a single result or a matrix to compare results, as well as a filter section. So you could ask for different facets of your current context, to iterate over a search session and interact with the results. Dunno, I’m still researching.
Have you written somewhere about your experience with Phind in this area?
Yes! We have a blog post here on how we designed these models and widgets: https://www.phind.com/blog/phind-2-model-creation.
Now that models have gotten much more capable, I'd suggest to give the executing model as much freedom with setting (and even determining) the schema as possible.
> If those features aren't supported by the widget's hard-coded schema, you're out of luck as a user.
Chat paired to the pre-built and on-demand widgets address this limitation.
For example, in the keynote demo, they showed how the chat interface lets you perform advanced filtering that pulls together information from multiple sources, like filtering only Zillow housers near a dog park.
Yes, because it seems that Zillow exposes those specific filters as a part of the input schema. As long as it's a part of the schema, then ChatGPT can generate a useful input to the widget. But my point is that is very brittle.
> Chat paired to the pre-built and on-demand widgets address this limitation
The only place I can see this working is if the LLM is generating a rich UI on the fly. Otherwise, you're arguing that a text-based UX is going to beat flashy, colourful things.
The problem is not the limitations of the capabilities per se but their discoverability (https://en.wikipedia.org/wiki/Discoverability). The user doesn't know what the capabilities are, as they are added and -- infuriatingly -- removed. Google Assistant is a perfect example of this.
Conservational user interfaces are opaque; they lack affordances. https://en.wikipedia.org/wiki/Affordance
They don't lack affordances, you can do stuff. They lack signifiers, ie it's not easy to discover the stuff you can do.
Affordance is not what it can do, it is what it signals that it can do. It needs to be perceptible, by the definition I use (Norman's). I see others go by different definitions that even admit hidden affordances. I do not.
From The Design of Everyday Things:
> Affordances represent the possibilities in the world for how an agent (a person, animal, or machine) can interact with something. Some affordances are perceivable, others are invisible. Signifiers are signals. Some signifiers are signs, labels, and drawings placed in the world, such as the signs labeled “push,” “pull,” or “exit” on doors, or arrows and diagrams indicating what is to be acted upon or in which direction to gesture, or other instructions. Some signifiers are simply the perceived affordances, such as the handle of a door or the physical structure of a switch. Note that some perceived affordances may not be real: they may look like doors or places to push, or an impediment to entry, when in fact they are not.
With Norman's definition, if a conversational interface can perform an action, it affords that action. The fact that you don't know that it affords that action means there's a lack of a signifier.
As you say, this is a matter of definition, I'm just commenting on Norman's specific definition from the book.
Yep, this is a big problem as well. If the user doesn't know what features will or won't work, they lose confidence overall.
Thank you for this word. I have felt it my whole life and never learned the exact word.
I immediately knew the last generation of voice assistants was dead garbage when there was no way to know what it could do, they just expected you to try 100 things, until it worked randomly
I know that AG-UI from copilot kit is in this space. But it hasn't worked well with the MCP model AFAIK
There was a recent post here about how deeply ingrained the chat interface is in OpenAIs organization. This really doubles down on that, but does anyone really like to interact with so much language instead of visual elements? Also feels horrible that you are supposed to remember a bunch of app names like "zillow" and punch them in the chat. And like an opportunity for them to slowly introduce ads for this apps or "preferential discovery", if you will, as monetization strategy.
Personally I don't hope thats the future.
Very much agreed. I think the dominance of the chat interface to LLMs has materially impaired the general usefulness of these tools — the sooner it goes away the better. It’s almost impossible to explain to a non-engineer how the illusion of a continuous conversation is crafted through context management and why past moments in a conversation might fall out of memory. My general advice to non-technical friends is to create a new conversation for each prompt so that they can get a more deterministic sense of how to formulate instructions and which are successful.
I was really hoping Apple would make some innovations on the UX side, but they certainly haven’t yet.
I feel like we're rehashing the debate around whether or not a GUI or terminal is more powerful.
For a large number of tasks that cleanly generalize into a stream of tokens, command line or chat is probably superior. We'll get some affordances like tab auto completion to help remember the name of certain bots or mCP endpoints that can be brought in as needed...
But for anything that involves discovery, graphical interaction feels more intuitive and we'll probably get bespoke interfaces relevant to that particular task at hand with some sort of partially hidden layers to abstract away the token stream?
counterpoint: a lot of people around me just type "zillow" in google to access it, so maybe it's not absurd to refer to it by name in a chat interface
Which post was that?
https://news.ycombinator.com/item?id=44573195 (in the article, search for:"Chat runs really deep")
Glad to see no AGI hubris in this presentation, but we also haven’t see anything groundbreaking: their own version of GUI plugins, their own version of a workflow builder, and an aspiration to take cut of every transaction on the web.
I hope their GUI integration will be eventually superseded by native UI integration. I remember such well thought out concepts dating back to 2018 (https://uxdesign.cc/redesigning-siri-and-adding-multitasking...).
AGI is so last year. Now it's all ASI which is great because it was achieved in like 1968 or something so nobody trying to achieve it can possibly fail
I see a lot of negative comments here but to me, it was obvious this is where OAI should land.
They want to be the platform in which you tell what you want, and OAI does it for you. It's gonna connect to your inbox, calendar, payment methods, and you'll just ask it to do something and it will, using those apps.
This means OAI won't need ads. Just rev share.
> This means OAI won't need ads. Just rev share
If OpenAI thinks there’s sweet, sweet revenue in email and calendar apps, just waiting to be shared, their investors are in for a big surprise.
> This means OAI won't need ads.
Ads are defenitely there. Just hidden so deeply in the black box which is generating the useful tips :)
If you ask it to build a headless frontend web app, it immediately starts generating code with Next.js. I’ve always wondered how it was trained to default to that choice, given the smorgasbord of web frameworks out there. Next.js is solid, but it’s also platform-ware, tightly coupled to commercial interests. I wish there were more bias toward genuinely open-source technologies.
There's probably different ways the LLM converged to it.
One could be for example: from people asking online which tools they should use to build something and being constantly recommended to do it with Next.js
Another could be: how many of the code that was used to train the LLM is done in Next.js
Generally, the answer is probably something along the lines of "next.js is kind of the most popular choice at the time of training".
To me it feels like the default choice in the industry, perhaps it's not and I'm wrong but if I could have that feeling I can see how the AI can as well.
I've never seen next.js in the wild. I have seen plain React plus dotnet, though, a million times.
It is a trap. But once you realise that you are already too deeply invested.
Because the AI labs are just hovering up all internet text that they can, I've been seeing more and more marketing pilots that deliberately seed marketing material in thousands of fake, AI-generated blogs and tutorials. The intention here is to get new LLMs to train on these huge numbers of associations between specific use cases and the company's product. All in a way that gets their marketing information into the final weights.
You may have started seeing this when LLMs seem to promote things based entirely on marketing claims and not on real-world functionality.
More or less, SEO spam V2.
> This means OAI won't need ads. Just rev share.
They obviously want both. In fact they are already building an ad team.
They have money they have to burn, so it makes sense to throw all the scalable business models in the history, eg app store, algo feed, etc, to the wall and see what stick.
A platform requires a user moat or unfair advantage. Having a better quality model is neither
Consumer LLM apps have moat. As it is, ChatGPT (the app) spends most of its compute on Personal Non work messages (approx 1.9B per day vs 716 for Work)[0]. First, from ongoing conversations that users would return to, then to the pushing of specific and past chat memories, these conversations have become increasingly personalized. Suddenly, there is a lot of personal data that you rely on it having, that make the product better. You cannot just plop over to Gemini and replicate this.
[0] https://www.nber.org/system/files/working_papers/w34255/w342...
How's having the best model not a most?
Because it changes all the time. A few weeks ago, it was Gemini 2.5 Pro, then Claude Opus 4.1, GPT-5 Thinking, now maybe Claude Sonnet 4.5, etc[1]. Having a good model isn't enough when they're basically interchangeable now. You need something else.
[1] This is an example. Which model was the best when is not important.
Because it depends on how much better “best” is. If it’s only incrementally better than open source models that have other advantages, why would you bother?
OpenAI’s moat will only come from the products they built on top. Theoretically their products will be better because they’ll be more vertically integrated with the underlying models. It’s not unlike Apple’s playbook with regard to hardwares and software integration.
This is an interesting branding exercise. Presenting MCP as 'Apps' makes it sound more accessible, while tools and MCP server sound very technical. Add a demo with Expedia and Spotify and you have an MCP that's end-user ready.
Ye, that's basically an MCP server, that can be used by ChatGPT.
It’s remarkable that will inevitably rush to build free apps that only reinforce OpenAI’s moat while cannibilizing their own opportunities.
When the iPhone came out, there were like 6 apps, and no app store.
In 2024, iOS App Store generated $1.3T in revenue, 85% of which went to developers.
Will this have a revenue share / marketplace built into it?
> Will this have a revenue share / marketplace built into it?
I'm genuinely surprised these companies went with usage-based versus royalty pricing.
That figure sounds way too high
Edit: yes I understand it is correct, but still it sounds like an insane amount
They're confusing "sales facilitates by the app store" with sales from the app store itself.
That 1T figure is real, but it includes things like if you buy a refrigerator using the Amazon iOS app.
Yeah, the article itself even lists the reality at about 20% of the 1.3T.
It's true, though.
It is now evident why Flash was murdered.
Because it was buggy, known for security holes and the single biggest source of application crashes in all software in the late 90's through early 00's.
you missed the "it drained battery like there was no tomorrow" argument.
I never really used it detached from a wall... mostly from work projects.
We get it, you drank the kool-aid.
Drank the kool-aid?!? I worked in the eLearning space, I was a prominent user and developer for Flash/Flex content... there was some interesting tooling for sure, I also completely disabled it on my home computers as a result of working with it.
I had a lot of hopes after the Adobe buyout that Flash would morph into something based around ActionScript (ES4) and SVG. That didn't happen. MS's Silverlight/XAML was close, but I wasn't going to even consider it without several cross-platform version releases.
>I was a prominent user and developer for Flash/Flex content
I was as well. It wasn't as bad as people describe it. It was an amazing platform, HTML5 just recently caught up.
In retrospective, Adobe should have open sourced it.
>MS's Silverlight/XAML was close
Hahahahahha, yeah sure! That tells me everything I need to know.
I agree it should have been open-sourced (at least the player portion)...
As for Silverlight, I mean the technology itself was closer to where I wanted to see Flash go. I'm not sure why you're laughing at that.
edit: as for not being as bad as people describe it... you could literally read any file on the filesystem... that's a pretty bad "sandbox" ... It was fixed later, but there were different holes along the way, multiple times.
> We now know why Flash was murdered
This is a stupid conspiracy given Apple decided not to support Flash on iPhone since before Jobs came around on third-party apps. (The iPhone was launched with a vision of Apple-only native apps and HTML5 web apps. The latter's performance forced Cupertino's hand into launching the App Store. Then they saw the golden goose.)
You ignore the state of things back then.
HTML5 was new and not widely supported, the web was WAY more fragmented back then, to put things in perspective, Internet Explorer still had the largest market share, by far. The only thing that could provide the user with a rich interactive experience was Flash, it was also ubiquitous.
Flash was the biggest threat to Apple's App Store; this wasn't a conspiracy, it was evident back then but I can see why it is not evident to you in 2025. Jobs open letter was just a formal declaration of war.
> HTML5 was new and not widely supported
Yes. It was a bad bet on the open web by Apple. But it was the one they took when they decided not to support Flash with the original iPhone's launch.
> Flash was the biggest threat to Apple's App Store
Flash was not supported since before there was an App Store. Since before Apple deigned to tolerate third-party native apps.
You can argue that following the App Store's launch, Apple's choice to not start supporting Flash was influenced by pecuinary interests. But it's ahistoric to suggest the reason for the original decision was based on interests Cupertino had ruled out at the time.
what's their moat that you refer to?
This is nonsense. Why would they destroy the incentive to get real-time, live data and MCP actions that help their users?
Connecting these apps will, at times, require authentication. Where it does not require payment, it's a fantastic distribution channel.
I don't understand, what could be built with this platform that wouldn't be made obsolete by conceivable updates to ChatGPT?
Another commenter suggested a hotel search function:
> Find me hotels in Capetown that have a pool by the beach .Should cost between 200 dollars to 800 dollars a night
ChatGPT can already do this. Similarly, their own pizza lookup example seems like it would exist or nearly exist with current functionality. I can't think of a single non-trivial app that could be built on this platform - and if there are any, I can't think of any that would be useful or not in immediate danger of being swallowed by advances to ChatGPT.
There are multiple branches they are exploring. This is a more structured one. But they also work on Agents that load the website and produce clicks to do the task. Also, this requires hand design, but they also work on generating the gui just-in-time, based on context.
They also have this new design gui for visual programming of agents, with boxes and arrows.
It's going to be a hybrid of all these. Obviously the more explicit work done for interoperability, the easier it is, but the gaps can be bridged with the common sense of the AI at the expense of more time and compute. It's like, a self driving car can detect red lights and speed limit signs via cameras but if there are structured signals in smart infrastructure, then it's simpler and better.
But it's always interesting to see this dance between unstructured and structured. Apparently any time one gets big, the other is needed. When theres tons of structured code, we want AI common sense to cut through it because even if it's structured, it's messy and too complicated. So we generate the code. Now if we have natural language code generators we want to impose structure onto how they work, which we express in markup languages, then small scripts, then large scripts that are too complex and have too much boilerplate so we need AI to generate it from natural language etc etc
I can't understand the documentation. How are the interactive elements embedded in the chat? Are they just iFrames?
The docs mention returning resources, and the example is returning a rust file as a resource, which is nonsensical.
This seems similar to MCP UI in result but it's not clear how it works internally.
If the connector is enabled by the prompt or via a UI interaction, it calls your MCP server. They have created some meta fields your tool can respond with, one of which is something about producing a widget along with a field for html.
In the current implementation, it makes an iframe (or webview on native) that loads a sandboxed environment which then gets another iframe with your html injected. Your html can include meta field whitelisted remote resources.
In 2018, I founded a startup specializing in chatbot for events. At the time the platforms were Alexa Skills, Actions on Google, and Messenger Platform (and LINE Bot, for people in Asia). I guess what's old is new again, but with fancier tech.
This time will be different?
We've actually got systems that can understand English now. Chatbots don't have to be glorified regular expression matches or based on inferior NLP. I've thought more than once that the true value of LLMs could well be that they essentially solve the language comprehension problem and that their ability to consume language is relatively underutilized compared to our attempts to get them to produce language. Under all the generative bling their language comprehension and ability to package that into something that conventional computing can understand is pretty impressive. They've even got a certain amount of common sense built in.
Yeah this seems accurate to me. All the talk of a bubble etc, but LLMs see genuinely useful at tasks like this and I'm sure we'll find more uses as time goes on.
Do people even want chatbots for events?
I personally prefer well curated information.
This is part of the fight regarding whether we will have utility apps inside the chat app or chatboxes inside the utility apps. Obviously OpenAI would prefer that they are in the driver seat and delegate to passive apps, while regular apps like Booking would prefer to be the app the user uses and to run an AI chatbox nested inside their own app UI, so they can swap it out etc.
Convenience-wise probably this model is more viable, and things will get centralized to the AI apps. And the nested utilities will be walled gardens on steroids. Using custom software and general computing (in the manner of the now discontinued sideloading on Android) will get even further away for the average person.
They also released ChatKit today for building in-app chat UI experiences, so it seems like OpenAI is trying to make sure they get a larger slice of the pie no matter which interaction model wins.
Seems wild to have an App SDK for a technology that's 1. supposed to free us from purpose-built APIs and interfaces, and 2. comprised entirely of a single textbox. Feels perhaps more like a MS-type strategy of standards and formal rules intended to lock down the extended ecosystem?
I think they want businesses to be more tightly integrated with ChatGPT to open up future opportunities for monetization.
remember when custom GPTs would just need an OpenAPI spec to be compatible with any existing API out there? we've been through this app store journey once before, maybe it's different this time since we now have agents and MCP
Do the examples work for any else?
It's poetic that Google attempted to pursue apps within Google Assistant years ago, but the vision of apps within an AI assistant is more feasible now with LLMs that (whether actually or not) understand arbitrary user intents and more flexible connectors to third party apps via MCP (and a viral platform with 700+ million weekly active users).
Custom GPTs (and Gemini gems) didn't really work because they didn't have any utility outside the chat window. They were really just bundled prompt workflows that relied on the inherent abilities of the model. But now with MCP, agent-based apps are way more useful.
I believe there's a fundamentally different shift going on here: in the endgame that OpenAI, Anthropic et al. are racing toward, there will be little need for developers for the kinds of consumer-facing apps that OpenAI appears to be targeting.
OpenAI hinted at this idea at the end of their Codex demo: the future will be built from software built on demand, tailored to each user's specific needs.
Even if one doesn't believe that AI will completely automate software development, it's not unreasonable to think that we can build deterministic tooling to wrap LLMs and provide functionality that's good enough for a wide range of consumer experiences. And when pumping out code and architecting software becomes easy to automate with little additional marginal cost, some of the only moats other companies have are user trust (e.g. knowing that Coursera's content is at least made by real humans grounded in reality), the ability to coordinate markets and transform capital (e.g. dealing with three-sided marketplaces on DoorDash), switching costs, or ability to handle regulatory burdens.
The cynic in me says that today's announcements are really just a stopgap measure to: - Further increase the utility of ChatGPT for users, turning it into the de facto way of accessing the internet for younger users à la how Facebook was (is?) in developing countries - Pave the way for by commoditizing OpenAI's complements (traditional SaaS apps) as ChatGPT becomes more capable as a platform with first-party experiences - Increase the value of the company to acquire more clout with enterprises and other business deals
But cynicism aside, this is pretty cool. I think there's a solid foundation here for the kind of intent-based, action-oriented computing that I think will benefit non-technical people immensely.
We have been building MCP servers and this looks very good directionally. Fills a bunch of holes in the protocol and gives meaning to something that were kind of like placeholders. Being able to return UI to the client is fantastic and will make lots of things possible. We have been working on these kinds of things assuming that the clients would improve to meet us.
This is honestly useful.
"Find me hotels in Capetown that have a pool by the beach .Should cost between 200 dollars to 800 dollars a night "
[injected with guerilla ads]
I don't see how this is a significant upgrade over the many existing hotel-finder tools. At best it slightly augments them as a first pass, but I would still rather look at an actual map of options than trust a stream of generated, ad-augmented text.
I think the end game is that rather than spitting out text back, the LLM transforms your plaintext request to something processable, and then chooses some relevant widgets to display the results.
The benefit I see is that it meets users where they presumable already are (GPT). As other comments allude to here, it's clear they see themselves as a staple of the user's online experience.
Developers, developers, developers!
Ads. They created ads. Now (or eventually) they can charge app developers to be featured first for a specific use case.
Ads was always the end goal, they have an opportunity to become a bigger player than Google in the ad space.
Instead of the user wasting time, ChatGpt can come up with the recommendations.
How else would the company sell their product? and keep people employed.
Of course ads will be there and this is good. A bad thing would be if they took a bunch of traffic from google and then gave no way to promote your products.
That would lead to companies closing and layoffs and economy decline.
Chat offers a far better experience than using Google—no more searching through spam-filled results, clicking between sponsored links, accepting endless cookie banners, and trying to read a tiny bit of useful content buried among ads and clutter.
It has the potential to bridge the gap between pure conversation and the functionality of a full website.
I’m just worried they we’ll go from very obvious advertising to advertising that’s a lot harder to spot.
I can block adds on a search engine. I cannot prevent an LMM from having hidden biases about what the best brand of vodka or car is.
Google could also have allowed invisible pay-for-placement without marking it as an ad. Presumably they didn't do that because undermining the perceived trustworthiness of their search results would have been a net loss. I wonder if chat will go in that same direction or not.
Pretty sure it's illegal to present advertisement and not label it as such in some form.
But as with everything, as new technologies emerge, you can devise legal loopholes that don't totally apply to you and probably need regulation before it's decided that "yeah, actually, that does apply to me".
I agree. But Google has gone in that direction long ago: ads are now harder to distinguish from genuine search results. In many cases, the organic results are buried so deep that they don’t even appear in the first visible section of the page anymore.
This feels like the death of the app, and the rise of the micro-app.
Remember "GPTs" and the thing before it which I don't even remember now. I think this will go the same route .. to nowhere
The GPT App Store (which is technically now obsolete with this SDK) was funny.
Are they still expecting us to get paid based on “revenue sharing”?
this seems kinda silly, especially given their previous app store flop. but I'm just happy there's some spark and competition in tech again. it's felt like the industry has been pretty stagnant since web 2.0 (more stagnant than any other time in the last 40-50 years, anyway). but this AI stuff feels like another "1977 Trinity" moment
so, best of luck to OAI. we'll see how this plays out
Honestly I see how somebody like kayak.com would build a "app" they work through commission, they don't care from where is the booking coming from. But they will sort the flight tickets based where do they earn the best commission. What's in there for me as a user ?. Also will openai let different providers pay for the top placement when somebody tries to buy ticket on chatgpt ?
Discovery, monetization. What is in it for developers?
deploying an app to 700M people?
Not only do you get to deploy your app to 700M users; you also get to provide responsive support for every single one of them!
Per the docs: 'Every app comes from a verified developer who stands behind their work and provides responsive support'
That's thinly veiled corporate speak for, Fortune 500 or GTFO
That's like saying making a website is like deploying an app for 7B people.
Sure, but deploying a website or app doesn't mean anyone's going to use it, does it?
I could make an iOS app, I could make a website, I could make a ChatGPT app... if no one uses it, it doesn't matter how big the userbase of iOS, the internet, or ChatGPT is...
Why does everyone think chat is better UX than traditional interfaces? I get the AI hype, but so many products are not a fit for chat interfaces.
Why would I use a chat to do what could be done quicker with a simple and intuitive button/input UX (e.g. Booking or Zillow search/filter)? Chat also has really poor discoverability of what I can actually do with it.
This x100. This is HCI 101. I'm glad I took that class during my master's program. It opened my eyes to a new world.
Does anyone think small players (like an independent developer) will be accepted ? Sounds like it will only for the big whales
That’s a great idea and Im wondering if Telegram can follow this path too, since they’re so advanced in mobile UX / UI, constantly updating their app and have some kind of crypto payments support.
What is the incentive for developers to build apps for this platform? I don't see any way of monetizing them at all.
fear of missing out, as always, be the first flappy bird in the store.
Your SaaS / Business is my Tool
chat is the best interface for information retrieval and REPL-like experiences. for all the rest, chat is horrible.
Just let the AI control my mouse and keyboard, let it use my device like a human. There's a huge swath of software already designed to be used by humans and anyone who uses ChatGPT knows that it's already been trained on every scrap of knowledge on how to use any existing complex software.
Tencent already has this with WeChat.Good to see it on chatgpt finally
They're looking like Facebook did with their phone project and later the metaverse - too big for their britches.
Lmfao..you've reminded me of the phone they made with HTC that had a Facebook button .
We've already sorta come full circle with the Meta glasses having a physical button to interact with the Facebook AI
If only this somehow resulted in fewer, better apps. <sigh>
This feels like a fever dream. As a developer everything changes every week. A new model, a new tool, a new sdk, paradigm we have to learn. I'm getting tired of all that shit.
As a JS developer for over 10 years who has seen multiple hype waves, here is my advice: You don't need to ride the first wave. You can wait until technology matures and see if it has staying power.
For example, React and TypeScript were hard to set up initially. I deferred learning them for years until the tooling improved and they were clearly here to stay. Likewise, I'm glad I didn't dive into tech like LangChain and CoffeeScript, which came and went.
LangChain has gone? I thought it was still around.
It's still around, but the hype has faded. Users discovered numerous issues with the project and began abandoning it. I remember one month when everyone was all, "LangChain is the future," and another month when the sentiment became: "LangChain is terrible."
You can see the hype cycle's timeline in HN's Algolia search: https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...
It is but I feel it's main value prop as a developer friendly abstraction layer has been very well solved for by the actual model providers themselves, while LangChain itself have become more bloated, clunky, and to under-opinionated.
This is how I feel about Rust.
The big hype wave has finished now (we still have the "how dare you criticise our technology bros" roaming around though), the tooling is maturing now. It's almost time for me to actually get my feet wet with it :)
Who says you have to learn this? You are free to ignore it if it's overwhelming.
I'd much rather see a thriving ecosystem full of competition and innovation than a more stagnant alternative.
With what exactly? They are desperately trying to create a "marketplace" and become gatekeepers on the backs of developers and businesses alike. There's no innovation here.
I guess what’s implied is that developers and businesses would innovate, not OpenAI directly.
Knowing OAI's history, only big whales could survive being copied by the platform's owner—case in point: Amazon Basics. They're so big that most of the time, SMBs can't escape them and don't have a choice but to cave to Amazon's demands. Is your product successful? Great, I'll copy you, add the "Amazon basics" label, and start bombarding users with my "product".
Amazon basics is a private label just like Costco and the Kirkland Brand. Same thing with Walmart, Target, Trader Joes, etc. And if these SMBs don't have to deal with Amazon, they will have to deal with a dozen copycats from China for anything that becomes a hit.
Please check how Amazon Basics works and what SMBs are saying.
For me the most annoying thing is APIs arbitrarily changing all the time. Completely change the entire Tailwind, ESLint, AWS SDK, etc APIs every 6 months? Why not! Heaven forbid you don't touch a project for a few months, blink and all your code is outdated.
You just point your AI agent at the docs and have it build the integration with your app for you :)
On a more serious note, it remains to be seen if this even sticks / is widely embraced.
Just get an LLM to do it for you.
The question is, whether having UI in chatgpt a game changer, fundamentally?
Specialize, escape, or accept.
Like "Abort, Retry, Fail"? And same as there, what's the difference between the first and the third? Is there a way of accepting a new sdk every week without specializing?
nothing really changed much here though. re llms nothing really has changed either, its mostly just scaling. there is really not much to learn as a consumer and app builder.
Welcome to technology
OpenAI launched an App Store in Nov 2023. A 23 month turnaround from major feature launch, to deprecation, to relaunch is a commitment to product longevity that’d put Google to shame.
I found it genuinely impressive how useless their "GPTs" were.
Of course, part of it was due to the fact that the out-of-the-box models became so competent that there was no need for a customized model, especially when customization boiled down to barely more than some kind of custom system prompt and hidden instructions. I get the impression that's the same reason their fine-tuning services never took off either, since it was easier to just load necessary information into the context window of a standard instance.
Edit: In all fairness, this was before most tool use, connectors or MCP. I am at least open to the idea that these might allow for a reasonable value add, but I'm still skeptical.
> I get the impression that's the same reason their fine-tuning services never took off either
> fine tuning because the work was repetitive enough that FT provided benefits in terms of speed and accuracy,
can you share anymore info on this. i am curious about what the usecase was and how it improved speed (of inference?) and accuracy.
Very typical e-commerce use cases processing scraped content: product categorization, review sentiment, etc. where the scope is very limited. We would process tens of thousands of these so faster inference with a cheaper model with FT was advantageous.
Disclaimer: this was in the 3.5 Turbo "era" so models like `nano` now might be cheap enough, good enough, fast enough to do this even without FT.
product roadmap was also ai generated
Is it just me, or does it seem odd that if you truly believed AGI would be achieved within a few years, you wouldn’t launch an app store for AI apps? I don’t think an app store makes any sense in a post-AGI world.
“Build our platform for us!”
So it’s take 2 for Open AI’s App Store moment. But this time surfing Anthropic’s MCP wave. Smart interop.. or just chasing the cool kids?
mcp was a dud
What is the superior way for an LLM to interact with your product?
llm can call my existing apis fine. curious what kind of problems you are running to with your existing apis?
This is them trying to build ChatGPT into platform, from which they will take some portion of revenue generated by these apps...hmm where have I seen this before.
They promised AGI and delivered SDKs. I think I'm gonna skip this one..
Hey don't sell them short, they also delivered a TikTok clone with vertically integrated slop generation. It's the 5D Chess path to AGI, they just need to rot the average human brain until the bar for super-human intelligence is reduced to an attainable level.
This was funny
Wow.
“CEO” Fidji Simo must really need something to do.
Maybe I’m cynical about all of this, but it feels like a whole lot of marketing spin for an MCP standard.
Yeah... no. I'm going to pass. The premise is bad from any angle. In the case of businesses, why "create" another "Amazon" and compete with other brands when the focus should be on getting customers through my sales funnel? For developers is much worse since they are going to copy Amazon's model with brands that found a niche: Amazon Basics. In this case, it'll be OpenAI "core" (or something like that), where you do all the work, and when your "app" is somewhat famous enough or getting traction, they'll copy it, rebrand it, and bombard all old and new customers to use it instead of yours.
I'mma call it now just for the fun of it: This will go the way of their "GPT" store.
Oh, I guess tomorrow when American HQs come online we will get some new shiny thing barely tested that needs to be deployed in production ASAP. Or maybe there is already something waiting for me in Slack...
The skepticism is understandable given the trajectory of GPTs and custom instructions, but there's a meaningful technical difference here: the Apps SDK is built on the Model Context Protocol (MCP), which is an open specification rather than a proprietary format.
MCP standardizes how LLM clients connect to external tools—defining wire formats, authentication flows, and metadata schemas. This means apps you build aren't inherently ChatGPT-specific; they're MCP servers that could work with any MCP-compatible client. The protocol is transport-agnostic and self-describing, with official Python and TypeScript SDKs already available.
That said, the "build our platform" criticism isn't entirely off base. While the protocol is open, practical adoption still depends heavily on ChatGPT's distribution and whether other LLM providers actually implement MCP clients. The real test will be whether this becomes a genuine cross-platform standard or just another way to contribute to OpenAI's ecosystem.
The technical primitives (tool discovery, structured content return, embedded UI resources) are solid and address real integration problems. Whether it succeeds likely depends more on ecosystem dynamics than technical merit.