I don't think it's an overstatement to say that with the right guardrails, this is the killer app for generative UI and perhaps general purpose AI for the masses. From a platform perspective, it's probably the most reasonable endgame for browsers and bespoke software for most people. It solves the problem of relying too much on chatbot UX for interacting with AI in a way that's obvious and understandable: just make the app make apps for specific tasks.
Yes, companies will rely on SaaS tools because they don't want to deal with the maintenance burden. But assume the tools improve over time, and agents can handle increasingly intricate work. It's not far-fetched to think a tool like this could be what most people interact with in the future except for the most low-level or niche of tasks.
I'm sure that in an ideal world, local models and hardware become so powerful that something like this can run entirely on-device. Out in the absence of the broader industry, I think this would be really cool. It's basically a holodeck for the web.
But if my goal were to obtain control of the web, replacing the need for literally any website and tool on it Using a single end-to-end stack would be a pretty good way of doing so. And I couldn't begin to think about the security and privacy implications of this app.
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.
A little hyperbole, but as an American, the idea that the average person in my country would rather drive somewhere rather than feel inconvenienced by a short walk is very accurate.
This is the Jevons paradox [1] in full display here. It's much easier to take advantage of hardware to run software at 120 FPS, so why not?
And I agree about liquid glass being successful iff they make the developer tooling for this as easy as additional modifiers to components, or even the default for SwiftUI.
AFAIK, Google stopped making Timeline accessible on the web in favor of local-first storage to avoid having to give location data when subpoenaed by law enforcement (since they can't give away data they literally don't have access to). And they didn't want to deal with the headache of user privacy-related lawsuits, so they defaulted Google accounts into auto-deleting location history (which was already opt-in for years).
I don't think the issue is that the West is going backwards in capability; rather, it's that although it has the capability to produce great products (software, media, etc.), it deliberately chooses not to because it's not as cost effective, because the people with expertise are overworked and understaffed, or because management had other priorities (see AAA game development).
AAA games are eye-wateringly expensive though, management aren't imagining it; my point is things becoming more expensive is a symptom of decline. I'm sure the late romans consoled themselves they could build another Pantheon they just cared more about efficiency now.
Where I work in government we've stopped paying for important data from vendors (think sensors around traffic etc.) because the quotes are eye-wateringly expensive. But I've worked in data long enough to know the quotes probably reflect genuine costs, because data engineers are so incompetent (and if it's a form of pricing gouging it's not working because gov isn't paying up). So it looks like we're choosing to be in the dark about important data, but it's not entirely a choice.
Saying we can do stuff but it's unaffordable is imo just another way of saying we can't do stuff.
They are, but they don't have to be. This is definitely an example of the West faltering where the East is really flourishing. They aren't trying to make the next Fortnite nor GTA 6 or whatever billion dollar day one hit. They pick a more modest scope and budget, reuse assets smartly, and get reliable releases out.
Google's recent launches have been technically impressive (especially Veo 2), but given the company's past track record on creating new products, I'm not very bullish that they can turn those launches into products with the same excitement and sense of direction as OpenAI at least appears to have. Google has the benefit of having platforms that span billions of devices and people, but with the looming threat of antitrust regulation, I'm not so sure they'll have the benefit of the last thing for long. Granted, I doubt that 1-800-ChatGPT will be a significant source of users for the product, but it does signal some of the creativity from the company that seems to be escaping Google regularly (see: NotebookLM's leads leaving to form their own startup).
Yes, companies will rely on SaaS tools because they don't want to deal with the maintenance burden. But assume the tools improve over time, and agents can handle increasingly intricate work. It's not far-fetched to think a tool like this could be what most people interact with in the future except for the most low-level or niche of tasks.
I'm sure that in an ideal world, local models and hardware become so powerful that something like this can run entirely on-device. Out in the absence of the broader industry, I think this would be really cool. It's basically a holodeck for the web.
But if my goal were to obtain control of the web, replacing the need for literally any website and tool on it Using a single end-to-end stack would be a pretty good way of doing so. And I couldn't begin to think about the security and privacy implications of this app.