I have been using Sites a few months now, really underrated feature IMO. Any time I get a quick app idea I just hand it off to Sites and usually get a working prototype within the hour.
Polished and easy to use website builders have been around for a long time, from Wordpress all the way to Wix.
The fact is, most people aren't builders and never will be. Building software is about formalizing and automating processes, and 95% of folks aren't interested in putting even the most basic effort into improving their business by approching it at a strategic level. Too focused on keeping the lights on and/or more interested in field operations.
For another example look at the many businesses using @gmail addresses to run their business while Google Workplace has streamlined custom domains for a while.
So in practice, ChatGPT Sites will mainly enable software agencies to build websites and offload maintenance to the customer : just manage it from your ChatGPT account. The equivalent of Bring your own cloud for Mom and Pop businesses, and other small companies. Not even sure ChatGPT Sites will create meaningful downward pressure on agency rates.
> Polished and easy to use website builders have been around for a long time, from Wordpress all the way to Wix.
The issue with the Wordpress and Wix is that the site stay polished for approximately 1 second after it has been initially published. In 6mo~1yr it's unmaintainable pile of dogshit that is easier to redo from scratch.
Oh come on. Sure, it might not be a production-capable system. But comparing an Astra 6/Sol 6.1 powered sites to wordpress and wix is apples and oranges.
An Astra powered site is your representation of a polished and capable business website. For the average joe, a Wix website fits the bill, their representation is not the same as yours. They focus on items such as whether or not the theme is beautiful.
I would bet there’s a (very) nontrivial number of people who have attempted to build something in Wix etc and got bogged down and frustrated enough to quit.
Lots.
I’m a pretty sophisticated user (both a designer and software engineer) and I’ve ragequit several of these tools before.
I hate Wix. I tried it to see if it could streamline things for me about a year and a half ago, and I quickly realized I could set up an Eleventy site, customize it and host it way faster. Now that agents are better and Cloudflare's API is fully exposed, I can have the agent search for a domain, buy it, set it up, set up the site and get it working on pages and probably be done before Wix onboarding was even complete.
Yes, I've exported and self-hosted in the past. I'll sometimes use Sites to mockup something and then I'll download and pass it onto another agent for use in something bigger.
Really cool idea! Beat level 6! I'm still not sure how to explain how it is showing more than 3 dimensions on a 2D screen though! Would be nice to be able to choose which dimensions are represented at the various N-dimensions, but the UI might be tricky to get nice.
Yeah you have actually picked up on something that the game intentionally does to help the user: If the cheese can be reached by any of the dimensions you are currently on, then it will ensure that dimension is in the subset of dimensions you receive when you go down a dimension.
Otherwise I'd have to introduce some kind of dimension picker. Maybe a fun hardmode.
Oh I see! Thanks for explaining that - I did notice that sometimes a subset that I didn't expect was chosen.
Did the LLM explain its reasoning for how it is showing the higher dimensions? Or perhaps this is a standard way of attempting to display higher dimensional space?
I didn't ask about standard representations, but I think the graph representation works well when paired with the discrete moves this maze allows. If we wanted continuous spaces it would probably be harder to play and harder to represent, but maybe more interesting.
As a developer I almost got defensive, but I think I feel what you're feeling. Dealing with change is hard. I am a developer who once felt value for knowing these things too. Today I wouldn't even bother opening the developer pane to test CSS changes. Tools that held so much potential to me are now headed for a museum, and it sucks.
Same. I went to a music festival and their website was hard to use, and the schedule was difficult to understand. And the area had bad connectivity. A few prompts and I was able to build a much better offline site.
Yeah if I were to make it harder I would introduce more sparse connections requiring the user to actually understand the layout, right now you can kind of "feel" your way towards the cheese. Most brutal maze would be one with only one possible path and lots of backtracking required.
Maybe I'll do that and post to HN later, didn't expect this much response!
Got curious and don't seem to be able to reproduce what you mean? In 2D Y+ corresponds to top of the screen and Y- to the bottom. X- is left and X+ is right. In 3D and above you can rotate the shape around so there is no real up or down? But maybe I'm missing some edge case where you go up into some high dimension position then back down?
People seem calm in here, but I really now think is replacing a lot of industries. Like Website Designer. I totally see people asking chatgpt to build them a website instead of paying them 2,000 USD.
Soon it will connect to Ads and it will runs for it as well.
Honestly since Opus 5.5, Opus is my interface for 95% off my work. I don’t open emails anymore i let it summarize draft and handle most of the work and I only talk to Claude and not read/write the emails myself.
I also barely go in analytics dashboards anymore, Most I do is creating artifacts to explain if text is not enough for me to understand.
I’m going to bed. I have analytics running for every cancellation on my software to see what goes on. I have a/b tests running that are getting interpreted automatically. My bugs are getting fixed automatically.
I can monitor my competitors and copy every feature a/b test resolution they’ve done instantly for (3$ of my budget)
I’m an early adaptor of the latest models. The labs will have better tools already, but it really feels like a breaking point. What’s normal for me today is normal for more tomorrow and the friction has never been lower...
The killer app behind Azure and GCP in a lot of cases, the thing that really drove multi-billion dollar decisions made by the largest cloud customers in the world, was that they were not Amazon.
Why would Walmart want to let Amazon hold their business hostage, tell them all their secrets - literally pay to hand it over to them - and generate profits that would surely be re-invested in competing against them in retail? Even if AWS handled Walmart's account in 100% good faith and helped Walmart tremendously, they would still ultimately be paying to make themselves less competitive in a world where their single competitive edge (low-margin retail) would be eroded.
Would the consumer - or really anybody that likes having cheap, widely available products they don't personally make for themselves - ultimately benefit in a world where Walmart's only choice is AWS (startup-branding) and AWS(ea-branding) and AWS(IBM branding) and AWS(Microsoft-branding, if they were also evil chinese communists who maybe trained on the traces they paid for)?
It's a fascinating case of use our product to build websites & apps until we can get our product to entirely replace websites & apps.
AI is this strange double-threat for folks in software development: attacking both the demand and supply at the same time.
Me, I don't think they are super eager to be a hosting platform even though this is a very logical extension for them. I think they're hoping this is a catalyst to make login with OpenAI as ubiquitous as login with Google.
Most people host their data on cloud anyway. Once you turn off training there is no difference to storing data with OpenAI/Anthropic vs Google or Meta.
Look at the "beneath the surface" example. Click on a creature close-up. Click the "rotate creature" button. The creature's rectangular jpeg, complete with black background, rotates on screen. No 3d or anything. This is their demo! It has that Potemkin village quality, much like AI apps, code, and answers.
Unfortunately right now most AI models get into the weeds of asking the user to sign up to netlify or firebase or buying a domain etc - none of which most users are interested in doing.
I wonder if they'll combine this with the currently-in-closed-beta Sign In with ChatGPT to let people make sites that can make inference calls that get charged to the user's ChatGPT subscription rather than the site owner's API billing account, like Claude Artifacts already lets you do. That would synergize well with ChatGPT Sites being generally more robust and feature-complete than Claude Artifacts.
They’ve both been working on this for a while, I’m surprised they still haven’t launched.
Domains, auth, and databases are probably the stickiest software products of all time. I suspect it’s a combo of security concerns and the risk of eroding goodwill among developers (their most important customer base rn) if they move up the stack too quickly. Once you know that the goal is to become cloud vendors, train on/compete with their own customers, and own the entire software stack e2e it’s hard to really feel grateful that they’re dragging it out but planning on doing it anyway.
For those of us working in infra/SaaS outside these companies it’s pretty clear that the only viable path that doesn’t involve getting cannibalized is training your own agents/models. The new coding agent-infra-data business model is a path towards full commoditization and undifferentiated prompting in 3-4 vertically integrated walled gardens. If you ever start making real money on pure software infra they’ll just be able to eat you alive by implementing something similar, training on their own tools rather than yours, and integrating it into their stack (which all of your customers are on already).
Also, you should only ever use $20-200/mo subscriptions on work you want them to train against or you don’t mind automate yourself out of. Think a little bit about what you’re teaching them to do when you use their products, especially if they’re the primary interface you’re working in. People are going to start caring about this a lot when the AI companies feel safe enough to begin the “extinguish” phase of AI coding. Train your own models!
> I do not see enough people discussing this and trying to position their businesses in a way to avoid destruction at the hands of the labs.
What can most even really do about it? Assuming they pull off what they're saying they will: it won't matter what you do, what vertical, what moat you think you have. It will be a concentration of capital and power we've never seen before, and frankly that's terrible for the world.
We need to build collaborative post-training pipelines to augment models with new custom capabilities without catastrophic forgetting/excessive loss of capability in other domains.
Currently what makes this too difficult for anybody but frontier labs is the lack of access to the full distribution of workloads/data they use to RL multiple separate envs/evals without regressing more than they advance in general capabilities. Because they are the primary buyers/builders of that stuff, they have no incentive or reason to allow anybody to replay or resample it but themselves. And it's all so very expensive to do in aggregate so there's not really demand for other product shapes except more openly available, collaborative/bundled RL that you could plug your specific workloads into (who the frontier labs would obviously not want to support with their business).
However, if someone were to build a collaborative rollout platform that you could use to train private workloads (ie create useful IP that doesn't just become profit for other labs/come from what they already can do), the upfront investment would be amortized over the very large number of potential buyers once it gets into the 5-8 figure range, who essentially have no other choice if they want to remain competitive in the technology industry.
Until openai/anthropic ipo the concentration of capital/spending and ndas/loss of employability is too concentrated for the best researchers to really do this without rocking the boat. And a lot of also-ran ai/saas have the same risk due to the lack of capitalized acquisition opportunities or AI vendors to partner with.
If the plan is ultimately to drive you out of business if you ever build anything profitable with their products, and monetize your knowledge without fairly rewarding or explaining their intent to do so, you might as well defect early and build what you inevitably would need anyway.
It’s really only an existential problem for pure-software businesses/developers.
There’s a good chance that the next step after open weight models is something that makes it easier to post-train existing models on new workloads without running an entire pipeline just for yourself. Somebody has to actually build it though.
On one hand, I am incredibly tired of seeing the absolute train wreck of waste that is suave founders taking millions and blowing it on hundreds of personnel who do practically nothing— especially having seen what a single person can do with these tools.
On the other hand, death of a livelihood in a social structure that doesn’t care about you.
We’ve always had pump-and-dump trash. The terrible stuff loses and the good stuff wins. I don’t see why that would change.
The companies that ditch the management structures of yore and successfully build better software faster with the new tools will come out on top. I don’t think these companies will replace ALL saas subscriptions with in-house solutions, because one huge benefit of paying someone else to hold that bag is liability and regulatory overhead.
I think this has always been the end result - when the user can enter a business problem and it spins up a site, and they don't have to think about the hosting, database, app, etc.
> And it will be good enough for 90% of use cases.
I think it’ll just raise the automation bar that much higher. In other words, it will be good enough for 90% of yesterday’s use-cases. But tomorrow’s? I’m bearish.
I imagine the amount of onerous bureaucracy when interacting with any entity is going to explode if they expect you to do it in an automated fashion, for both private companies and the government. Especially as roles are cut because they can be “automated”. Except only 80% of the workload can be reliably automated, the other 20% requires a long tail of effort.
Its not for those uses cases. though. THink about squarespace/wix hosting all the company landing pages, restaurant menus, daycares. This will replace it.
Honest question for devs out there... is anyone worried? I don't see enough frank conversation happening around this subject. Timelines. How long until salaries fall? Is it too early to stalk thinking about back up plans? Am I crazy?
No, you should be thinking about backup plans, because you should assume you’re not going to walk into another career (I am not suggesting that you are).
- People in other industries actually have specialist knowledge (this is something IT people forget)
- AI is coming for those jobs too
- and so are lots of other people, not just displaced techies but from any number of other industries.
Thinking about it in advance should include training yourself up for other sectors.
If AI is successful, it will be a very different world of jobhunting. If it’s unsuccessful there will still be job losses for a while because of overspending, and because of the wider economic issues.
You're not crazy. I think there's still room for the top 5% of devs working on genuinely hard technical problems.. but I don't see a bright future for the rest.
I think technically building this stuff is now easier for the masses. But knowing what to build, how to describe it, how to test it, figuring out how to understand the users -- that's the hard part, always was and I'll be employed doing that for a couple more years at least. But after that, I see this all going towards building for agents, building for users will be a thing of the past. Then I think its not as easy for people to relate to that, probably want a machine designing for other machines. UX and taste don't really matter much at that point.
There’s going to be a lot of chaos at some point but I think that “hybrid builders” (scientists, sales people, product developers, etc). true specialists, and new opportunities in hw/robotics/training/ will soak up a lot of talent, and be rewarding to work in. Feels very sadly k-shaped.
Software engineering on one end is going to be more like finance or medicine where the cost of poor performance is actually higher than the risk that they wont be worth what you pay them (ie there isn’t any amount of pay where it’s worth it to hire them). Then on the other it’ll be like knowing how to use PowerPoint. All the new tech categories have too many bottlenecks and challenges beyond just producing code to consolidate like that.
No. Saas hedge is to offer solid api + mcp for agents to use whilst still maintaining their own products alongside. That's it, that's the whole thing.
Are you going to trust a vendor who has been running for years with your business critical operations or brad_69's awesome saas skills and agents? Do you think these businesses are going to shift their whole product to be in Altman's walled garden when they don't need to?
Salaries won't fall, they'll go up. The number of roles is what'll fall. SWE is going to become a giant game of musical chairs, and people who can't find a seat will get pushed into human facing roles ("product" / "technical sales" / "customer support" ).
I've said it before but I'll repeat it here until someone changes my mind: jobs in IT are safe until using AI gets its equivalent of "no one ever got fired for buying IBM". Until there can be financial and legal responsibility for when things go bad, people are going to be needed.
Once it's so good AI labs will be happy to accept liability, it's game over. It was never about code quality.
Slapping up a slop website is no threat to professional tech folk.
When they have full DevOps, with multiple environments for testing, compliance with security best practices, can do database updates and data migrations, move the app smoothly between different platforms, streamline costs, and scale to at least hundreds of thousands (if not millions) of users without crashing and burning... then go ahead and worry. We aren't anywhere near that.
On the other hand, if you are purely a coder, and cannot run a project that does all of the above... maybe start your worrying earlier.
We aren't, but the AI Labs no doubt are. They just can't release it right now because their bread and butter is still developers burning tokens, so there is still a business problem to solve.
Coding is probably solved. Honestly it’s probably going to be 5 years of runtime starting now depending on industry lag and uptake.
A team I was managing spent a whole year converting government forms into a web app. It took a while and there was a lot of discussion with the end users and a bunch of back and forth (keep in mind this is government software with a lot of onerous legal requirements). Then about 4 months ago, we just fed Codex the forms, along with the UI design and the legal requirements and it spat it out in 1 hour. Was it perfect? No. But our iterative loop became make the web app, show it to end users and then incorporate feedback and completely cut out the UX team, product etc. We were able to complete the conversion in 2 weeks with one guy (something like 150+ government forms/applications). For most simple to medium corporate development work it’s basically over.
We're in a similar boat. Entire eng team went full AI about a year ago, and honestly we're struggling to find things to do. Features that would take weeks, take minutes.
I just don't see how we keep going like this without layoffs. The company cannot responsibly continue paying for this many engineers, when half, or less, could do the same job.
Serious question from a consumer of software: if software can be improved so effortlessly now, why isn't it getting better?
Like, why is it all still so buggy? Why does no website seem to consider 96 PPI displays in its design? Why don't we have ostensibly useful features like undo trees everywhere? Why is everything still a RAM hog? Why does the proportion of websites that pay no heed to accessibility seem to be going up rather than down?
I'm told we're living in the future, but it sure doesn't feel like it.
For the same reason that increasing headcount never really solved the bug problem.
Flawless software isn't the goal, ever. Feature development is almost always the primary goal, and fixing bugs and performance issues is only done if the bug is bad enough.
I don't think you're wrong overall, and it jives with tripleee's reply, but I don't think wanting end-user software to be less buggy than it currently is is asking for "flawless" software.
Because the goal isn't to make the future better for humanity. It's to make the future more addictive for humanity. Big difference. More clicks, more ad views, more engagement, not necessarily more pleasant. I mean, that's obvious.
Lack of information. When you are on the building side you become blind to its faults. Consumers almost never tell you what isn't good about it.
> why is it all still so buggy?
Also lack of information. I haven't seen any bugs in years in the things I've worked on. LLMs have closed every last obvious testing gap, squashing any outstanding bugs there. That's not to say it is all bug-free now, but what bugs users may be experiencing aren't being reported.
AI is good at building software, but it is still not good at consuming software like actual consumers do.
If your company isn't throwing you into R&D right now everyone is looking at layoffs. You are no longer in an environment where doing "the same job" will keep you afloat.
I have been the lone developer on a project for quite some time and was already feeling that way. Yet, recently found out that a second is being added to the team and a third is to come. I guess I'll enjoy what is effectively a paid vacation before it all comes crumbling down.
It’s mainly mucus, from a piece of mouse colon that was cut open and gently flushed out, leaving the inner mucus layer attached. The actual mucus is black here by default but there is another channel you can turn on in the sidebar to see it.
On top of it we added bacteria (green) and bacteria sized beads (red). On the bottom are colon lining cells (blue).
The inner mucus layer of the colon was generally thought to be impenetrable to bacteria, serving as a way to protect you from the trillions of bacteria in the gut.
It turns out that actually, it’s really easy for bacteria to swim through mucus by using flagella. We found some bacterial mutants that swim through mucus (the green ones). The red beads serve as a control marking the top of the mucus since they can’t get through the mucus.
The real question is then - if they can get through the mucus at any time, why don’t they? And there are various theories about this - which I’m happy to elaborate on this you want. The basic idea is there is a risk whenever you cross the mucus (you get closer to the immune system) and mucus actually feeds and tames bacteria over time by providing them with various sugars and things like that. But if they get desperate/starve, then they will cross.
Mucus is still a mystery - there are substructures within mucus, take a look at this image for example:
I don't think this provides a lot of value for more complex applications. It just lowers the bar for getting started.
Once the app gets a bit more complex, it won't be that much different to any more traditionally vibe coded application (code, git/pr, test, deploy, ops). It only removes the barrier to get started, but doesn't solve any of the real problems. Problems like AI slop, code rot, bad architecture, security issues, data loss/breach because of bad code, and so on.
I mean, with a Cloudflare API key you can pretty much already do that. But keeping your data straight and putting in the right authentication guards, etc. are all still where the business problems meet technical and the AI is not enough unless the agent is both architect and customer.
That's what I do but normal people don't know what any of those terms mean so their daily AI app does that shopping for them and they don't need to get an "Appy Key".
The Saaspocalypse was a wet fart that never went anywhere. People aren't using these AI tools to kill or replace SaaS at their jobs as much as HN wants us to believe.
The argument there wasn't that everyone everywhere would use individually customized software* instead of packaged SaaS; it was that, because AI greatly reduces the cost of building an internal app or service, it improves enterprises' BATNA when dealing with SaaS vendors, which means surplus will shift from producers to consumers. It's not obvious to me that that's not happening, since it may have been masked by a general increase in automation since AI also synergizes significantly with SaaS. (The Stripe Economics post gets into this a little, though it's mostly less curious than I'd like about why the "SaaSpocalypse" didn't materialize. Of course, Stripe has an economic interest here.)
* Some people do think this (e.g., https://sockpuppet.org/blog/2026/09/25/what-even-is-an-os-no...), but I remain somewhat skeptical that this'll be true for the most economically significant applications. This is not because I'm an AI capabilities skeptic (I'm not, I'm more of a doomer), but because even in a world where AI is designing and implementing all the software, there are probably still significant benefits to standardization. I see it as sort of analogous to compilers; we might get a world where compilers are both used and maintained almost exclusively by LLMs, but we probably won't get one where compilers stop being used or maintained at all because LLMs write everything in assembly.
In stuff like wikis and ticket systems what I see is people now just making their own little tools around the ones we already use, this arguably just sements their position even deeper.
Also I dont get the inhouse argument, you dont really save anything. So instead of paying X to SaaS, you now propose to pay X to anthropic instead AND maintain said SaaS... what? how is that better, that just shifted the cost to another company and you got the burden of maintenance and responsibility.
People curate their brands on socials. Having a website is not really the badge it once was, and the ability to generate them like this further reduces the social proof they used to provide.
Look at squarespace + wix + lovable + framer + vercel + wordpress (and agencies), cloudflare, godaddy, etc. and think about how much more demand there is for a vertically integrated one stop shop without friction. This is a massive market even if it doesn’t become myspace.
This is the top of the funnel for most people building a tech product or marketing their business, so it’s very helpful if you intend to sell other services they might want to use down the line. (Cf Cloudflare except imagine if they also integrated ads into one of the most popular websites in the world)
You want to talk to a few guys not in the software industry to understand what the world is like first.
While there are many enthusiasts out there, lots and lots of people couldn't care less about ChatGPT allows them to "buuld their own website". It's simply irrelevant to their life.
The idea is great but the websites shown as examples are clearly made by AI because they make no sense.
Current AI is NOT going to replace web devs or coders because it cannot tabula rasa make human sense. I suspect the reverse with a Jevons thingy happening
They are desperate to find a moat cause models aren’t much more than a commodity.
What they might not realise is that stuff/services like these (ChatGPT Sites) are probably not much of a moat either in a world where majority of population has access to increasingly better models capable of building competitors in days if not hours. Even if they restrict usage of their models when it comes to creating competitor services there will always be other models that you can use to build competitor.
What is happening is that their products are putting people out of work but at the same time they are also enabling anyone to easily build competitors to their own products and are thus putting their own services “out of work”. They are essentially eating their own tail. Ouroboros.
Is noone seeing that?
The only truly lasting moat might be hardware/compute. At least for now.
I just had a tangential thought, but what if... AI completely recreated their own version of the internet from scratch? And that's what the internet eventually becomes.
Gonna wager good money they attempt to relaunch Apps or something of the sort, where you build, launch, host within ChatGpt, seamlessly making it easier really soon.
Idk about that, a company like Form3, Edenred etc are technically just software companies - but no one in their right mind is going to build a black box payment provider. Sure, if your business is wordpress websites, you're probably going to see a big downswing in new customers. But there is absolutely still going to be a place for highly specialised and security dependent software.
something that solves a real problem for businesses, that you are uniquely positioned to execute on well.
If you target the lowest common denominator, or consumers, or build something really quickly - good luck
My question ("Do you know what <details> and <summary> tags are?") doesn't appear to be here; but given the page functionality the answer seems to be no.
I get that this can be considered a "minor inconvenience", but I think it's pretty embarrassing on a page that's supposed to be marketing a website builder that's supposed to take care of knowing these things for you.
I took it for a spin. I used Claude to create a self contained HTML file with JSON data embedded to create a SVG bar chart with sort and filter buttons. I then gave the HTML file to ChatGPT Work and it created the site using the provided HTML code unadulterated.
I then asked the next question. How can I automatically update the embedded JSON data. It gave two answers which I thought was weird. First, it said I could using a plugin to an MCP server. Second, it said I could use an API and when I asked for which API... it said there was an API available in ChatGPT Sites but no details. So I assume it can be done.
What I am thinking of is to use ChatGPT sites as a replacement for SharePoint. I can create self contained HTML charts and let employees view it within our enterprise instance of ChatGPT. The piece I am working on is how to automatically update the embedded data from our database.
Something I've been thinking a lot about recently is how much we still need UIs vs if we'll just interact with agents.
We have a platform that enables building both and I still find UIs to be quite relevant. Especially if it's so easy to build like in this case because you can always throw together a visualization for something quickly and then never look at it again.
This looks really cool and because I'm in a tangential space I know how fun and useful it's been to build and host stuff on the web quickly.
We're geared towards fully-fledged (serverless) apps and agents for internal use only though, in case it's of interest to anyone: https://railcode.dev/
I’ve been using exe.dev to throw up micro sites. It has a built in web agent and first party mobile app agent. And any coding agent that can use “ssh” (all of them) can easily figure out how to drive it.
It also has a lot more developer and PaaS tools like GitHub integration.
RIP Heroku for imploding before the wave of vibe coding. We had the ability to “spin up a free app on the internet” before it was cool.
Depending on the idea this would be a great idea. Simple sites like portfolios for example are perfect for this job. Most people however are usually vague when describing what they want so I imagine you still need to be able to structure your toughts correctly. I'll definetly be trying this
I've been having such a blast making simple mapping apps that aggregate public data like GIS sources. I made a map for my city with every public parking space labeled, interactive with links to google maps and free hours, et cetera. I also made an educational map for myself to explore data on Jihadism.
I was imagining the other day what other "apps" that frontier labs can get into. If we had "Claude Mail" or "ChatGPT for Workspaces" it would be very interesting. Email service created by Claude, then docs (they have it already?) then all the business apps layer (heck integrate your vibe coded apps onto these! build your own slack, CRM)
If models are commoditized and apps is where the grab is - I actually wouldn't mind a new email provider - GMail has not practically changed in the last 10 years I feel.
For whatever reason most of the sites are Arabic-language prototypes for things like plumbing companies. I'm not sure why - if those are solving a real problem, or whether someone is running a scam, or what.
Cool project, but the hero art is a little strange (what is the dude on the left hammering? Aren’t the ones on the right both pulling the pulley down from opposite ends?)
Thanks! I think the fox is steadying the log as the raccoon hoists it. Agree the badger is sort of hammering the air, haha. The art was AI-generated but with some manual editing in Affinity, so hopefully it's not too sloppy.
Very polished. I suspect OpenAI will abandon this project at some point to focus on other core products. This kind of feature is great, but it spreads them thin, and other players like Cloudflare are rolling hard to provide this kind of convenience to agents with far greater depth. This is not a judgement against ChatGPT Sites - it's a well-executed product.
Sites has been one of the most valuable ships for us internally outside of general model releases imo. A few things that I think are incredibly underrated:
1. Plugin usage. I can build a site with some data vis and when someone else in the company logs in, it uses their access to evaluate those queries. The access model/BYO connection is really powerful.
2. Ephemeral sites. Setting up a site is so easy they can be ephemeral. I can set up a whole site for a presentation, or a complex form, or just a meme, and then never touch it again because it cost me basically none of my time.
3. Sites as plugins. Sites directly integrate with the plugin ecosystem in that a site itself can provide/be a plugin. Combine that with Sign in with Chat (byo subscription) and you can build some really cool experiences really easily. I've been working on a deck grader for Magic the Gathering the last few months for my own uses and now I could pretty trivially publish it as a site for others to use.
I don't see sites being abandoned any time soon :)
Sites is so good that it’s changed my workflow - I start in Opus 5.5, have it write a plan to Google Docs in markdown, then have Codex pick up the plan, shave it down, and build the prototype. A little annoying to switch tools, annoying enough to only do this on desktop, but the output is fantastic.
I think this is the next step in the evolution of AI based tools. We can develop the app or tool, but we need somewhere to easily publish it.
I actually recently shared my own version of this concept, but self hosted. I'm optimistic that there can't be much lock-in from the big providers, but I think there's a large market to capture
They'd better be proactive with checking for abusive users, because I can easily see people trying to spin up honeypots, crypto miners, scam sites, etc. on this. Especially given that the hosting appears to be free with undefined CPU/RAM limits.
We’ve been building something like this, but perhaps even more powerful, safe, and beginner friendly. If Sites looks like it’s the right speed for you, we’d love your feedback on https://vibes.diy/
I have been building https://github.com/openrundev/openrun. OpenRun allows you to build web apps, adding OIDC/SAML with RBAC to control who can access the apps. Supports MCP for app management. Also supports service bindings, which enables providing each app an isolated database account.
My guess is Google and Apple are going to eventually take all the market share for this "app builder" category. They own the device market-share and can offer way better integration
Does anyone know if Claude makes it easy to publish sites u vibe coded to a custom subdomain or new domain? That’s the biggest reason I use this feature in ChatGPT over Claude
You can ask Claude to deploy it with ShipStatic. It works pretty well for this use case, no account is required, and you can publish to a custom domain/subdomain without much setup.
Disclaimer: I’m one of the people behind ShipStatic.
I'm genuinely sad that what we learned with hard work and sweat, our craft, our identity, our soul; is now something you can buy. I've been in this profession for a long time and now I see people who have absolutely no fucking clue, but lots of money, and they buy all these agents and make things that work (way more broken, with absolutely shit quality), but they work and so they brag about their creations to me all day long, and man that makes me feel like a total loser.
I spent decades trying to master media theory, language design, computer graphics, compiler design, rewriting systems, mathematics, logic, build systems, graphs, data-flow systems, making GUI kits, layout systems, ... all to make a better system for thinking, and now the whole idea of thinking is outsourced (as altman says "sold as utility by the meter"). I need to become an addict or something and be wasted. This is not a life.
If it's any consolation, almost no craft will be spared. Software was simply first in line because of how much training data was available, but it won't be the last.
Are we looking at the same website? My eyes immediately go to the font and typeface styling and it registers immediately as AI generated. There’s a very distinct “AI” website aura that Claude and ChatGPT make that requires you to change a ton of the styling to make go away.
I think people are missing the point. To fix the sloppiness all you need is a non-technical person who wanted the site/app to tell it what to fix. Either way the technical middle-person is gone
You spent decades on this and you think that linked website is not slop? You say its "basic" but its really not, its overengineered nonsense if you look closely.
It's worth for the token budget. You can do so much with gpt-6.1-sol (high) on a small sub. If the money is not aan issue it's probably still worth it for Astra because it can occasionally be smarter than Fable 5.1 (or so I heard).
Could you just be an API to serve intelligence instead of moving into the app layer and putting your clients out of business. How much wealth and power do you need to accumulate seriously?
The "intelligence API" layer seems like a pretty rough business to be in. The smarter the models the more they're commodities (since they can use any harness or scaffolding) resulting in a race to the bottom on pricing. And then each new better model requires double the cost of the previous one.
I'm surprised there's not a bigger push into better and useful memory, since that would result in actual lock-in.
What is wrong with being a commodity, or a utility? Oil companies do just fine. So do electricity companies. Why does everyone have to seek world domination instead of building a good business. My point is by cannibalizing their clients they risk not being trusted by anyone. But Microsoft did the same thing and created a durable monopoly because Republicans abandoned anti trust enforcement and that seems likely here.
The question of them moving into the app layer is completely irrelevant. If you are threatened by this the reality you need to begin making peace with is that your business is not going to exist within 10 years. And that's being extremely generous.
This was always the play…
Wait until the models are trained to work on “capsules of your data” stored by OpenAI, and deploy your apps/services onto OpenAI cloud where you can _easily_ share the development with your team members’ AI agents. Ofcourse, they will all work only if you use and pay for their models via their subscriptions. A new global {K-V, Relational, Vector} database that your coding model is already trained to work with! will come very soon…
They know the API price wars are a race to the bottom with little/no profit margins.
I think this is going to back fire. I have talked to 5 tech leaders at companies and universities this week about setting up their own infrastructure for inference and training. They are worried about multiple rug pulls: increasing cost of tokens, stealing ideas, training on data etc. they are all willing to uses less intelligence models if they avoid these issues. They’ve seen me setting up propelcompute.com for my start up and propelcode.app neither of which are intended to be commercial products and want to do the same.
Don't worry, we had DreamWeaver, Wordpress and other tools for a long time now but it still did not replace web developers. Build quality tailor made websites that connect to humans and you will be fine.
DreamWeaver and WordPress both had limits on the complexity of site they could build. If you wanted more than what they can do, you look to web developers.
If you're wondering why OpenAI's models seem to be struggling to stay ahead of Anthropic and are falling further behind, I think it's because of side quests like this.
They have dots, they had Sora, so many other things that take away attention from just having the best model. There are already a bazillion ways to do this. Sure, the integration's nice, but it's just not needed.
Weird to describe something they could really easily sell to non-coders as a "side quest".
I think programmers make the mistake of thinking generative AI is for us, when it is much more ominously about us.
Newsflash: a company that plans to be the most valuable business on earth and has a desperate need for revenue is not going to stay out of anyone’s lane if there is money to be found in it. If you can make an AI wrapper, so can they.
This actually looks like it has meaningful potential, and given how easily they could train it to connect to a whole bunch of off-the-shelf APIs running in data centres they control, looks like a little glimpse of the future, and of the kind of power the industry has gleefully handed to the crypto-for-your-eyeball-scans guy.
Cloud AI has been the face-eating leopard of the tech industry for the last three years. This is them just starting to chew. How long before they tweak it to generate intranet apps, mini stores, specific AI wrappers? Why wouldn’t a firm that has so much influence over data centre rollout, RAM and processor prices simply directly start using their code generation tools and data centres to build and run the work product?
Maybe some of the big firms who have been helping train ChatGPT to eat their lunch will now wake up to the need for open weights.
I think at this point assessments like this are more qualitative than quantitative. I use Codex far more than I use Claude for coding and my results are generally higher quality. I would not, however, go to someone and tell them that Codex is better than Claude generally. In large part it depends on how you use the model and the harness.
And even if you do know how to use the model and the harness, Claude is just generally still a much a better experience overall.
Returned to Claude as my main driver after many months of Codex and my frustration index is in much, much better place. I can't recommend Codex/GPT to anyone, not just normies. It does make for a really nice reviewer though.
All of these "side quests" (and especially this one) are attempts to build a walled-garden.
Especially this. OpenAI will host it. OpenAI will throw "Sign in with ChatGPT" there. "Apps" will be built that depend on end users ChatGPT "connections". The app will fundamentally not work without an end-user having a ChatGPT account/subscription and "connected apps".
Having "the best model" will absolutely not support a $1T+ valuation. Owning the activity of 25-50% of internet users (assuming a split between OAI/Anthropic/Google/maybe-Meta), is.
I have been using Sites a few months now, really underrated feature IMO. Any time I get a quick app idea I just hand it off to Sites and usually get a working prototype within the hour.
For example, here's a game where you are a mouse hunting for cheese in higher dimensional mazes: https://manifold-six-axis-escape.nathan338932.chatgpt.site/
Idea came to me while at a concert, was ready to play that same night.
Polished and easy to use website builders have been around for a long time, from Wordpress all the way to Wix.
The fact is, most people aren't builders and never will be. Building software is about formalizing and automating processes, and 95% of folks aren't interested in putting even the most basic effort into improving their business by approching it at a strategic level. Too focused on keeping the lights on and/or more interested in field operations.
For another example look at the many businesses using @gmail addresses to run their business while Google Workplace has streamlined custom domains for a while.
So in practice, ChatGPT Sites will mainly enable software agencies to build websites and offload maintenance to the customer : just manage it from your ChatGPT account. The equivalent of Bring your own cloud for Mom and Pop businesses, and other small companies. Not even sure ChatGPT Sites will create meaningful downward pressure on agency rates.
Lesson learned the hard way.
> Polished and easy to use website builders have been around for a long time, from Wordpress all the way to Wix.
The issue with the Wordpress and Wix is that the site stay polished for approximately 1 second after it has been initially published. In 6mo~1yr it's unmaintainable pile of dogshit that is easier to redo from scratch.
Also: The Wordpress Block Editor is a mess and has a very steep learning curve.
could not agree more! most people really aren't builders. AI enables builders
Oh come on. Sure, it might not be a production-capable system. But comparing an Astra 6/Sol 6.1 powered sites to wordpress and wix is apples and oranges.
An Astra powered site is your representation of a polished and capable business website. For the average joe, a Wix website fits the bill, their representation is not the same as yours. They focus on items such as whether or not the theme is beautiful.
I would bet there’s a (very) nontrivial number of people who have attempted to build something in Wix etc and got bogged down and frustrated enough to quit.
Lots.
I’m a pretty sophisticated user (both a designer and software engineer) and I’ve ragequit several of these tools before.
I hate Wix. I tried it to see if it could streamline things for me about a year and a half ago, and I quickly realized I could set up an Eleventy site, customize it and host it way faster. Now that agents are better and Cloudflare's API is fully exposed, I can have the agent search for a domain, buy it, set it up, set up the site and get it working on pages and probably be done before Wix onboarding was even complete.
I recommend to all my customers - publish and use your own domain.
I mean Wordpress been around for decades, it’s solid, mature and everywhere. I’d even say way better than any vibe coded site
I tried to set up a Wordpress site for someone last month. I suggested it because of the reasons you said.
I’m convinced that’s the only reason Wordpress is still popular. The administrative experience was atrocious. Actually, that is an understatement.
*.chatgpt.site - I see many dead sites in very near future.
Is it possible to export site to own domain?
Yes, I've exported and self-hosted in the past. I'll sometimes use Sites to mockup something and then I'll download and pass it onto another agent for use in something bigger.
From what I can tell, "Sites" product supports custom domains so it's not an "export" per se but it is a way to map.
https://learn.chatgpt.com/docs/sites?surface=web#connect-a-c...
https://help.openai.com/en/articles/20001339-creating-and-us...
Really cool idea! Beat level 6! I'm still not sure how to explain how it is showing more than 3 dimensions on a 2D screen though! Would be nice to be able to choose which dimensions are represented at the various N-dimensions, but the UI might be tricky to get nice.
Yeah you have actually picked up on something that the game intentionally does to help the user: If the cheese can be reached by any of the dimensions you are currently on, then it will ensure that dimension is in the subset of dimensions you receive when you go down a dimension.
Otherwise I'd have to introduce some kind of dimension picker. Maybe a fun hardmode.
Oh I see! Thanks for explaining that - I did notice that sometimes a subset that I didn't expect was chosen.
Did the LLM explain its reasoning for how it is showing the higher dimensions? Or perhaps this is a standard way of attempting to display higher dimensional space?
I didn't ask about standard representations, but I think the graph representation works well when paired with the discrete moves this maze allows. If we wanted continuous spaces it would probably be harder to play and harder to represent, but maybe more interesting.
The dimension knob doesn't have user-select: none; and you probably have no idea what that means and why it's annoying.
As a developer I almost got defensive, but I think I feel what you're feeling. Dealing with change is hard. I am a developer who once felt value for knowing these things too. Today I wouldn't even bother opening the developer pane to test CSS changes. Tools that held so much potential to me are now headed for a museum, and it sucks.
user-select: none is considered harmful. Don't use it.
I made it through the 6th puzzle, it was a lot of fun. I basically just stayed in 3 dimensions until I had to move up and then went back to 3.
Yeah I think moving up and down dimensions is the best strategy. I find spamming movements in 6D works but only because the paths are dense enough.
Wow that game makes my head hurt. Cool idea!
Thanks! Yeah I find it helps to jump back down to lower (comprehensible) dimensions, move closer at that level, then jump back up.
Same. I went to a music festival and their website was hard to use, and the schedule was difficult to understand. And the area had bad connectivity. A few prompts and I was able to build a much better offline site.
Melted my brain. 10/10.
Super fun game!
pretty fun! beat level 6 after about 4 minutes... should add some more levels :)
Yeah if I were to make it harder I would introduce more sparse connections requiring the user to actually understand the layout, right now you can kind of "feel" your way towards the cheese. Most brutal maze would be one with only one possible path and lots of backtracking required.
Maybe I'll do that and post to HN later, didn't expect this much response!
Got so annoyed at this game. +/- Y should be up and down. Gets confusing with x being the same way.
Got curious and don't seem to be able to reproduce what you mean? In 2D Y+ corresponds to top of the screen and Y- to the bottom. X- is left and X+ is right. In 3D and above you can rotate the shape around so there is no real up or down? But maybe I'm missing some edge case where you go up into some high dimension position then back down?
People seem calm in here, but I really now think is replacing a lot of industries. Like Website Designer. I totally see people asking chatgpt to build them a website instead of paying them 2,000 USD.
Soon it will connect to Ads and it will runs for it as well.
Honestly since Opus 5.5, Opus is my interface for 95% off my work. I don’t open emails anymore i let it summarize draft and handle most of the work and I only talk to Claude and not read/write the emails myself.
I also barely go in analytics dashboards anymore, Most I do is creating artifacts to explain if text is not enough for me to understand.
I’m going to bed. I have analytics running for every cancellation on my software to see what goes on. I have a/b tests running that are getting interpreted automatically. My bugs are getting fixed automatically.
I can monitor my competitors and copy every feature a/b test resolution they’ve done instantly for (3$ of my budget)
I’m an early adaptor of the latest models. The labs will have better tools already, but it really feels like a breaking point. What’s normal for me today is normal for more tomorrow and the friction has never been lower...
Interesting..
Once everyone operates as you do, then websites will have little value. Just publish a .md file or API if you have interactivity.
I’m aware.
The killer app behind Azure and GCP in a lot of cases, the thing that really drove multi-billion dollar decisions made by the largest cloud customers in the world, was that they were not Amazon.
Why would Walmart want to let Amazon hold their business hostage, tell them all their secrets - literally pay to hand it over to them - and generate profits that would surely be re-invested in competing against them in retail? Even if AWS handled Walmart's account in 100% good faith and helped Walmart tremendously, they would still ultimately be paying to make themselves less competitive in a world where their single competitive edge (low-margin retail) would be eroded.
Would the consumer - or really anybody that likes having cheap, widely available products they don't personally make for themselves - ultimately benefit in a world where Walmart's only choice is AWS (startup-branding) and AWS(ea-branding) and AWS(IBM branding) and AWS(Microsoft-branding, if they were also evil chinese communists who maybe trained on the traces they paid for)?
better hope you're not walmart
It's a fascinating case of use our product to build websites & apps until we can get our product to entirely replace websites & apps.
AI is this strange double-threat for folks in software development: attacking both the demand and supply at the same time.
Me, I don't think they are super eager to be a hosting platform even though this is a very logical extension for them. I think they're hoping this is a catalyst to make login with OpenAI as ubiquitous as login with Google.
You don’t mind letting a company get 100% access to your personal data?
Unless you use a zero data retention inferencer?
Most people host their data on cloud anyway. Once you turn off training there is no difference to storing data with OpenAI/Anthropic vs Google or Meta.
Wall-E all the way...
I'll have my bots talk to your bots. Maybe our bots can do lunch.
Look at the "beneath the surface" example. Click on a creature close-up. Click the "rotate creature" button. The creature's rectangular jpeg, complete with black background, rotates on screen. No 3d or anything. This is their demo! It has that Potemkin village quality, much like AI apps, code, and answers.
They probably slopcoded the entire post including the examples, so nobody human even touched any of it or tested it before posting it.
They forgot to add "make the creatures rotate in 3D" to the prompt.
This seems to close a gap that both Claude and chatgpt had.
If you ask something like "please make a little website for me and my friends to plan our shopping lists on", I would expect the result to be:
"Here you go: https://shopping-lister.chatgpt.com/"
Unfortunately right now most AI models get into the weeds of asking the user to sign up to netlify or firebase or buying a domain etc - none of which most users are interested in doing.
I wonder if they'll combine this with the currently-in-closed-beta Sign In with ChatGPT to let people make sites that can make inference calls that get charged to the user's ChatGPT subscription rather than the site owner's API billing account, like Claude Artifacts already lets you do. That would synergize well with ChatGPT Sites being generally more robust and feature-complete than Claude Artifacts.
Good idea!
Disclaimer: I work on Sites
This is probably going to be hugely successful with the majority of the population.
It's a matter of weeks/months before they'll host a complete webapp, not just a site. The Saaspocalypse will be complete.
They’ve both been working on this for a while, I’m surprised they still haven’t launched.
Domains, auth, and databases are probably the stickiest software products of all time. I suspect it’s a combo of security concerns and the risk of eroding goodwill among developers (their most important customer base rn) if they move up the stack too quickly. Once you know that the goal is to become cloud vendors, train on/compete with their own customers, and own the entire software stack e2e it’s hard to really feel grateful that they’re dragging it out but planning on doing it anyway.
For those of us working in infra/SaaS outside these companies it’s pretty clear that the only viable path that doesn’t involve getting cannibalized is training your own agents/models. The new coding agent-infra-data business model is a path towards full commoditization and undifferentiated prompting in 3-4 vertically integrated walled gardens. If you ever start making real money on pure software infra they’ll just be able to eat you alive by implementing something similar, training on their own tools rather than yours, and integrating it into their stack (which all of your customers are on already).
Also, you should only ever use $20-200/mo subscriptions on work you want them to train against or you don’t mind automate yourself out of. Think a little bit about what you’re teaching them to do when you use their products, especially if they’re the primary interface you’re working in. People are going to start caring about this a lot when the AI companies feel safe enough to begin the “extinguish” phase of AI coding. Train your own models!
This is truly depressing and most likely very true.
I do not see enough people discussing this and trying to position their businesses in a way to avoid destruction at the hands of the labs.
> I do not see enough people discussing this and trying to position their businesses in a way to avoid destruction at the hands of the labs.
What can most even really do about it? Assuming they pull off what they're saying they will: it won't matter what you do, what vertical, what moat you think you have. It will be a concentration of capital and power we've never seen before, and frankly that's terrible for the world.
We need to build collaborative post-training pipelines to augment models with new custom capabilities without catastrophic forgetting/excessive loss of capability in other domains.
Currently what makes this too difficult for anybody but frontier labs is the lack of access to the full distribution of workloads/data they use to RL multiple separate envs/evals without regressing more than they advance in general capabilities. Because they are the primary buyers/builders of that stuff, they have no incentive or reason to allow anybody to replay or resample it but themselves. And it's all so very expensive to do in aggregate so there's not really demand for other product shapes except more openly available, collaborative/bundled RL that you could plug your specific workloads into (who the frontier labs would obviously not want to support with their business).
However, if someone were to build a collaborative rollout platform that you could use to train private workloads (ie create useful IP that doesn't just become profit for other labs/come from what they already can do), the upfront investment would be amortized over the very large number of potential buyers once it gets into the 5-8 figure range, who essentially have no other choice if they want to remain competitive in the technology industry.
Until openai/anthropic ipo the concentration of capital/spending and ndas/loss of employability is too concentrated for the best researchers to really do this without rocking the boat. And a lot of also-ran ai/saas have the same risk due to the lack of capitalized acquisition opportunities or AI vendors to partner with.
If the plan is ultimately to drive you out of business if you ever build anything profitable with their products, and monetize your knowledge without fairly rewarding or explaining their intent to do so, you might as well defect early and build what you inevitably would need anyway.
you could try to help build that!
It’s really only an existential problem for pure-software businesses/developers.
There’s a good chance that the next step after open weight models is something that makes it easier to post-train existing models on new workloads without running an entire pipeline just for yourself. Somebody has to actually build it though.
On one hand, I am incredibly tired of seeing the absolute train wreck of waste that is suave founders taking millions and blowing it on hundreds of personnel who do practically nothing— especially having seen what a single person can do with these tools.
On the other hand, death of a livelihood in a social structure that doesn’t care about you.
We’ve always had pump-and-dump trash. The terrible stuff loses and the good stuff wins. I don’t see why that would change.
The companies that ditch the management structures of yore and successfully build better software faster with the new tools will come out on top. I don’t think these companies will replace ALL saas subscriptions with in-house solutions, because one huge benefit of paying someone else to hold that bag is liability and regulatory overhead.
look at the accounting firms, you'll just have four or five companies that switch off between who holds the bag
I think this has always been the end result - when the user can enter a business problem and it spins up a site, and they don't have to think about the hosting, database, app, etc.
Vibe coding full circle
And it will be good enough for 90% of use cases.
> And it will be good enough for 90% of use cases.
I think it’ll just raise the automation bar that much higher. In other words, it will be good enough for 90% of yesterday’s use-cases. But tomorrow’s? I’m bearish.
I imagine the amount of onerous bureaucracy when interacting with any entity is going to explode if they expect you to do it in an automated fashion, for both private companies and the government. Especially as roles are cut because they can be “automated”. Except only 80% of the workload can be reliably automated, the other 20% requires a long tail of effort.
And whose neck can you choke when an error from the AI generated site forces you to refund customers or something worse?
Its not for those uses cases. though. THink about squarespace/wix hosting all the company landing pages, restaurant menus, daycares. This will replace it.
In my area many of those use cases are handled by Facebook pages already. For the general public this might still seem too technical.
given the percentage of business processes that still run on a manually updated spreadsheet, I am not sure that is actually worse.
It is likely better in some ways and worse in others...and few of hte people involved will have the understanding to tell the difference.
Honest question for devs out there... is anyone worried? I don't see enough frank conversation happening around this subject. Timelines. How long until salaries fall? Is it too early to stalk thinking about back up plans? Am I crazy?
> How long until salaries fall?
Hopefully you've been living below your means and saving.
No, you should be thinking about backup plans, because you should assume you’re not going to walk into another career (I am not suggesting that you are).
- People in other industries actually have specialist knowledge (this is something IT people forget)
- AI is coming for those jobs too
- and so are lots of other people, not just displaced techies but from any number of other industries.
Thinking about it in advance should include training yourself up for other sectors.
If AI is successful, it will be a very different world of jobhunting. If it’s unsuccessful there will still be job losses for a while because of overspending, and because of the wider economic issues.
You're not crazy. I think there's still room for the top 5% of devs working on genuinely hard technical problems.. but I don't see a bright future for the rest.
I think technically building this stuff is now easier for the masses. But knowing what to build, how to describe it, how to test it, figuring out how to understand the users -- that's the hard part, always was and I'll be employed doing that for a couple more years at least. But after that, I see this all going towards building for agents, building for users will be a thing of the past. Then I think its not as easy for people to relate to that, probably want a machine designing for other machines. UX and taste don't really matter much at that point.
There’s going to be a lot of chaos at some point but I think that “hybrid builders” (scientists, sales people, product developers, etc). true specialists, and new opportunities in hw/robotics/training/ will soak up a lot of talent, and be rewarding to work in. Feels very sadly k-shaped.
Software engineering on one end is going to be more like finance or medicine where the cost of poor performance is actually higher than the risk that they wont be worth what you pay them (ie there isn’t any amount of pay where it’s worth it to hire them). Then on the other it’ll be like knowing how to use PowerPoint. All the new tech categories have too many bottlenecks and challenges beyond just producing code to consolidate like that.
No. Saas hedge is to offer solid api + mcp for agents to use whilst still maintaining their own products alongside. That's it, that's the whole thing.
Are you going to trust a vendor who has been running for years with your business critical operations or brad_69's awesome saas skills and agents? Do you think these businesses are going to shift their whole product to be in Altman's walled garden when they don't need to?
Salaries won't fall, they'll go up. The number of roles is what'll fall. SWE is going to become a giant game of musical chairs, and people who can't find a seat will get pushed into human facing roles ("product" / "technical sales" / "customer support" ).
Naw, I'll just build a competitor to whoever sidelines me next :)
Nah. Website builders have been around for a long time in various forms. This isn't much different.
In a corporate environment the ability to code does not substitute a decent software engineer.
I've said it before but I'll repeat it here until someone changes my mind: jobs in IT are safe until using AI gets its equivalent of "no one ever got fired for buying IBM". Until there can be financial and legal responsibility for when things go bad, people are going to be needed. Once it's so good AI labs will be happy to accept liability, it's game over. It was never about code quality.
Slapping up a slop website is no threat to professional tech folk.
When they have full DevOps, with multiple environments for testing, compliance with security best practices, can do database updates and data migrations, move the app smoothly between different platforms, streamline costs, and scale to at least hundreds of thousands (if not millions) of users without crashing and burning... then go ahead and worry. We aren't anywhere near that.
On the other hand, if you are purely a coder, and cannot run a project that does all of the above... maybe start your worrying earlier.
> We aren't anywhere near that.
We aren't, but the AI Labs no doubt are. They just can't release it right now because their bread and butter is still developers burning tokens, so there is still a business problem to solve.
Coding is probably solved. Honestly it’s probably going to be 5 years of runtime starting now depending on industry lag and uptake.
A team I was managing spent a whole year converting government forms into a web app. It took a while and there was a lot of discussion with the end users and a bunch of back and forth (keep in mind this is government software with a lot of onerous legal requirements). Then about 4 months ago, we just fed Codex the forms, along with the UI design and the legal requirements and it spat it out in 1 hour. Was it perfect? No. But our iterative loop became make the web app, show it to end users and then incorporate feedback and completely cut out the UX team, product etc. We were able to complete the conversion in 2 weeks with one guy (something like 150+ government forms/applications). For most simple to medium corporate development work it’s basically over.
We're in a similar boat. Entire eng team went full AI about a year ago, and honestly we're struggling to find things to do. Features that would take weeks, take minutes.
I just don't see how we keep going like this without layoffs. The company cannot responsibly continue paying for this many engineers, when half, or less, could do the same job.
Serious question from a consumer of software: if software can be improved so effortlessly now, why isn't it getting better?
Like, why is it all still so buggy? Why does no website seem to consider 96 PPI displays in its design? Why don't we have ostensibly useful features like undo trees everywhere? Why is everything still a RAM hog? Why does the proportion of websites that pay no heed to accessibility seem to be going up rather than down?
I'm told we're living in the future, but it sure doesn't feel like it.
AI doesn't really make the process better or more reliable, it just speeds it up.
Like the old quip that a cup of coffee lets you do stupid things faster.
For the same reason that increasing headcount never really solved the bug problem.
Flawless software isn't the goal, ever. Feature development is almost always the primary goal, and fixing bugs and performance issues is only done if the bug is bad enough.
I don't think you're wrong overall, and it jives with tripleee's reply, but I don't think wanting end-user software to be less buggy than it currently is is asking for "flawless" software.
Because those don't increase business revenue. People aren't making their buying decisions off that
Because it hasn't been that long since the November singularity.
Because the goal isn't to make the future better for humanity. It's to make the future more addictive for humanity. Big difference. More clicks, more ad views, more engagement, not necessarily more pleasant. I mean, that's obvious.
> why isn't it getting better?
Lack of information. When you are on the building side you become blind to its faults. Consumers almost never tell you what isn't good about it.
> why is it all still so buggy?
Also lack of information. I haven't seen any bugs in years in the things I've worked on. LLMs have closed every last obvious testing gap, squashing any outstanding bugs there. That's not to say it is all bug-free now, but what bugs users may be experiencing aren't being reported.
AI is good at building software, but it is still not good at consuming software like actual consumers do.
If your company isn't throwing you into R&D right now everyone is looking at layoffs. You are no longer in an environment where doing "the same job" will keep you afloat.
> honestly we're struggling to find things to do.
I have been the lone developer on a project for quite some time and was already feeling that way. Yet, recently found out that a second is being added to the team and a third is to come. I guess I'll enjoy what is effectively a paid vacation before it all comes crumbling down.
You'll need less people but not no people.
Yeah, probably 10% of our current workforce is all that’s needed.
You could argue this was true prior to AI. A lot of the workforce is doing questionable value bureaucracy
What do you mean in a few months?
Yesterday my dot made a web app to view 3D microscopy files
https://mucus-atlas-viewer.pazimzadeh.chatgpt.site/?sample=k...
I didn't even know about the sites feature
That's so cool, what am I looking at?!
It’s mainly mucus, from a piece of mouse colon that was cut open and gently flushed out, leaving the inner mucus layer attached. The actual mucus is black here by default but there is another channel you can turn on in the sidebar to see it.
On top of it we added bacteria (green) and bacteria sized beads (red). On the bottom are colon lining cells (blue).
The inner mucus layer of the colon was generally thought to be impenetrable to bacteria, serving as a way to protect you from the trillions of bacteria in the gut.
It turns out that actually, it’s really easy for bacteria to swim through mucus by using flagella. We found some bacterial mutants that swim through mucus (the green ones). The red beads serve as a control marking the top of the mucus since they can’t get through the mucus.
Find this and other images in my paper: https://pmc.ncbi.nlm.nih.gov/articles/PMC11784811/
The real question is then - if they can get through the mucus at any time, why don’t they? And there are various theories about this - which I’m happy to elaborate on this you want. The basic idea is there is a risk whenever you cross the mucus (you get closer to the immune system) and mucus actually feeds and tames bacteria over time by providing them with various sugars and things like that. But if they get desperate/starve, then they will cross.
Mucus is still a mystery - there are substructures within mucus, take a look at this image for example:
https://mucus-atlas-viewer.pazimzadeh.chatgpt.site/?sample=f...
And here is a good paper too from my collaborator (who actually took these images) who discovered a sub-type of mucus and mucus-producing cells:
https://pubmed.ncbi.nlm.nih.gov/33859001/
I don't think this provides a lot of value for more complex applications. It just lowers the bar for getting started.
Once the app gets a bit more complex, it won't be that much different to any more traditionally vibe coded application (code, git/pr, test, deploy, ops). It only removes the barrier to get started, but doesn't solve any of the real problems. Problems like AI slop, code rot, bad architecture, security issues, data loss/breach because of bad code, and so on.
Don't underestimate it though. The easy on-ramp is why PHP was popular a decade ago despite being a crap language.
It is a complete web app, with a Cloudflare backend.
I mean, with a Cloudflare API key you can pretty much already do that. But keeping your data straight and putting in the right authentication guards, etc. are all still where the business problems meet technical and the AI is not enough unless the agent is both architect and customer.
That's what I do but normal people don't know what any of those terms mean so their daily AI app does that shopping for them and they don't need to get an "Appy Key".
The Saaspocalypse was a wet fart that never went anywhere. People aren't using these AI tools to kill or replace SaaS at their jobs as much as HN wants us to believe.
https://www.stripeeconomics.com/p/the-saaspocalypse-was-more...
The argument there wasn't that everyone everywhere would use individually customized software* instead of packaged SaaS; it was that, because AI greatly reduces the cost of building an internal app or service, it improves enterprises' BATNA when dealing with SaaS vendors, which means surplus will shift from producers to consumers. It's not obvious to me that that's not happening, since it may have been masked by a general increase in automation since AI also synergizes significantly with SaaS. (The Stripe Economics post gets into this a little, though it's mostly less curious than I'd like about why the "SaaSpocalypse" didn't materialize. Of course, Stripe has an economic interest here.)
* Some people do think this (e.g., https://sockpuppet.org/blog/2026/09/25/what-even-is-an-os-no...), but I remain somewhat skeptical that this'll be true for the most economically significant applications. This is not because I'm an AI capabilities skeptic (I'm not, I'm more of a doomer), but because even in a world where AI is designing and implementing all the software, there are probably still significant benefits to standardization. I see it as sort of analogous to compilers; we might get a world where compilers are both used and maintained almost exclusively by LLMs, but we probably won't get one where compilers stop being used or maintained at all because LLMs write everything in assembly.
In stuff like wikis and ticket systems what I see is people now just making their own little tools around the ones we already use, this arguably just sements their position even deeper.
Also I dont get the inhouse argument, you dont really save anything. So instead of paying X to SaaS, you now propose to pay X to anthropic instead AND maintain said SaaS... what? how is that better, that just shifted the cost to another company and you got the burden of maintenance and responsibility.
I don't understand which part of my comment, if any, you're disagreeing with.
Except for the companies that pay anti-competition money to OpenAI: cloning their applications will be restricted.
I doubt it, but I am open to being wrong.
People curate their brands on socials. Having a website is not really the badge it once was, and the ability to generate them like this further reduces the social proof they used to provide.
Look at squarespace + wix + lovable + framer + vercel + wordpress (and agencies), cloudflare, godaddy, etc. and think about how much more demand there is for a vertically integrated one stop shop without friction. This is a massive market even if it doesn’t become myspace.
This is the top of the funnel for most people building a tech product or marketing their business, so it’s very helpful if you intend to sell other services they might want to use down the line. (Cf Cloudflare except imagine if they also integrated ads into one of the most popular websites in the world)
You want to talk to a few guys not in the software industry to understand what the world is like first.
While there are many enthusiasts out there, lots and lots of people couldn't care less about ChatGPT allows them to "buuld their own website". It's simply irrelevant to their life.
Yeah, I'm seeing how I can move my dad's website to this so he can manage his site on his own.
Eternal AI September.
This is probably going to be as successful as Sora
The Sites feature seems a lot more monetizable than Sora ever could have been.
The idea is great but the websites shown as examples are clearly made by AI because they make no sense.
Current AI is NOT going to replace web devs or coders because it cannot tabula rasa make human sense. I suspect the reverse with a Jevons thingy happening
If you have a product manager directing the AI (instead of developers) to build what they want then it will make sense.
But then you quickly have tech debt build up because current LLMs cant plan over long scales
They don't do it by default all the time but if prompted to they're pretty good at designing and refactoring code
I'm increasingly finding my job is just making sure they don't go off the rails, and if they do it's a small effort of getting them back on track
I think you're underestimating your own role in making sure they are not going off the rails haha
yeah, I hope you're right. I find it hard to imagine how non or mildly-technical people would use it differently
Stuff like a wrapper over Plaid and some OCR is definitely dead for sure.
But currently they can't manage a vending machine as good as human, so I doubt they have our centuries long planning ability
Tech debt when doing a website? If it’s just a CMS, Claude and Codex are more than capable of building and maintaining it.
Also, planning over long scales is not something you should do if you do lean/agile. Premature optimization is a root of all evil ;)
I don't think that Claude or Codex could build and maintain a CMS as of yet.
Software is mostly in the last 5% of work
They are desperate to find a moat cause models aren’t much more than a commodity.
What they might not realise is that stuff/services like these (ChatGPT Sites) are probably not much of a moat either in a world where majority of population has access to increasingly better models capable of building competitors in days if not hours. Even if they restrict usage of their models when it comes to creating competitor services there will always be other models that you can use to build competitor.
What is happening is that their products are putting people out of work but at the same time they are also enabling anyone to easily build competitors to their own products and are thus putting their own services “out of work”. They are essentially eating their own tail. Ouroboros. Is noone seeing that?
The only truly lasting moat might be hardware/compute. At least for now.
I just had a tangential thought, but what if... AI completely recreated their own version of the internet from scratch? And that's what the internet eventually becomes.
I built a ChatGPT Site [0] for VT Code [1]: a browser-based WebMCP editor for exploring code, reviewing changes, and working with the agent.
[0] https://vtcode.vinhnx.chatgpt.site
[1] https://github.com/vinhnx/VTCode
Gonna wager good money they attempt to relaunch Apps or something of the sort, where you build, launch, host within ChatGpt, seamlessly making it easier really soon.
Then the app economy gets going within ChatGPT
already did, you can host plugins on ChatGPT Sites now, announced at devday
Looks like a jab at Lovable.
Wonder what domain they're going to pick next. I feel like many LLM-dependent businesses are going to get sherlocked by OpenAI and Anthropic.
honest question what is the sherlock proof business/app/saas we can do as independent builders
hard out here hard for startups too
Personal relationships, trust and a preexisting audience are your moat, from now until the end of time, in every field.
If your business is just software… it might be cooked.
Idk about that, a company like Form3, Edenred etc are technically just software companies - but no one in their right mind is going to build a black box payment provider. Sure, if your business is wordpress websites, you're probably going to see a big downswing in new customers. But there is absolutely still going to be a place for highly specialised and security dependent software.
something with a real life physical component
something that solves a real problem for businesses, that you are uniquely positioned to execute on well. If you target the lowest common denominator, or consumers, or build something really quickly - good luck
> Frequently asked questions
My question ("Do you know what <details> and <summary> tags are?") doesn't appear to be here; but given the page functionality the answer seems to be no.
I get that this can be considered a "minor inconvenience", but I think it's pretty embarrassing on a page that's supposed to be marketing a website builder that's supposed to take care of knowing these things for you.
I took it for a spin. I used Claude to create a self contained HTML file with JSON data embedded to create a SVG bar chart with sort and filter buttons. I then gave the HTML file to ChatGPT Work and it created the site using the provided HTML code unadulterated.
I then asked the next question. How can I automatically update the embedded JSON data. It gave two answers which I thought was weird. First, it said I could using a plugin to an MCP server. Second, it said I could use an API and when I asked for which API... it said there was an API available in ChatGPT Sites but no details. So I assume it can be done.
What I am thinking of is to use ChatGPT sites as a replacement for SharePoint. I can create self contained HTML charts and let employees view it within our enterprise instance of ChatGPT. The piece I am working on is how to automatically update the embedded data from our database.
Something I've been thinking a lot about recently is how much we still need UIs vs if we'll just interact with agents.
We have a platform that enables building both and I still find UIs to be quite relevant. Especially if it's so easy to build like in this case because you can always throw together a visualization for something quickly and then never look at it again.
This looks really cool and because I'm in a tangential space I know how fun and useful it's been to build and host stuff on the web quickly.
We're geared towards fully-fledged (serverless) apps and agents for internal use only though, in case it's of interest to anyone: https://railcode.dev/
Looks very useful.
I’ve been using exe.dev to throw up micro sites. It has a built in web agent and first party mobile app agent. And any coding agent that can use “ssh” (all of them) can easily figure out how to drive it.
It also has a lot more developer and PaaS tools like GitHub integration.
RIP Heroku for imploding before the wave of vibe coding. We had the ability to “spin up a free app on the internet” before it was cool.
Depending on the idea this would be a great idea. Simple sites like portfolios for example are perfect for this job. Most people however are usually vague when describing what they want so I imagine you still need to be able to structure your toughts correctly. I'll definetly be trying this
I've been having such a blast making simple mapping apps that aggregate public data like GIS sources. I made a map for my city with every public parking space labeled, interactive with links to google maps and free hours, et cetera. I also made an educational map for myself to explore data on Jihadism.
I was imagining the other day what other "apps" that frontier labs can get into. If we had "Claude Mail" or "ChatGPT for Workspaces" it would be very interesting. Email service created by Claude, then docs (they have it already?) then all the business apps layer (heck integrate your vibe coded apps onto these! build your own slack, CRM)
If models are commoditized and apps is where the grab is - I actually wouldn't mind a new email provider - GMail has not practically changed in the last 10 years I feel.
I’m getting dejavu. Is this the 2026 version of Yahoo GeoCities
I love the concept, but the `.chatgpt.site` suffix reflexively makes me think of those early-2000s era sites.
My fun little project has been Sherlocked! https://weejur.com
For whatever reason most of the sites are Arabic-language prototypes for things like plumbing companies. I'm not sure why - if those are solving a real problem, or whether someone is running a scam, or what.
Very nice. I really like the design.
I got tired of spinning up repos for basic pages, so made my own host last year - https://pageplane.app.
I think the SEO/AEO agencies are using them to seed content for clients.
Cool project, but the hero art is a little strange (what is the dude on the left hammering? Aren’t the ones on the right both pulling the pulley down from opposite ends?)
Very clean site design though.
Thanks! I think the fox is steadying the log as the raccoon hoists it. Agree the badger is sort of hammering the air, haha. The art was AI-generated but with some manual editing in Affinity, so hopefully it's not too sloppy.
Very polished. I suspect OpenAI will abandon this project at some point to focus on other core products. This kind of feature is great, but it spreads them thin, and other players like Cloudflare are rolling hard to provide this kind of convenience to agents with far greater depth. This is not a judgement against ChatGPT Sites - it's a well-executed product.
[Disclaimer, I work at OpenAI]
Sites has been one of the most valuable ships for us internally outside of general model releases imo. A few things that I think are incredibly underrated:
1. Plugin usage. I can build a site with some data vis and when someone else in the company logs in, it uses their access to evaluate those queries. The access model/BYO connection is really powerful.
2. Ephemeral sites. Setting up a site is so easy they can be ephemeral. I can set up a whole site for a presentation, or a complex form, or just a meme, and then never touch it again because it cost me basically none of my time.
3. Sites as plugins. Sites directly integrate with the plugin ecosystem in that a site itself can provide/be a plugin. Combine that with Sign in with Chat (byo subscription) and you can build some really cool experiences really easily. I've been working on a deck grader for Magic the Gathering the last few months for my own uses and now I could pretty trivially publish it as a site for others to use.
I don't see sites being abandoned any time soon :)
Funny how the showcased "Tidal house" has unreadable text on an image.
https://tidal-house-retreat.openai.chatgpt.site/
> Quiet cabins. A view of your own.
Also, there's a missing space in "Look insideeach cabin" and "Save astay plan".
luckily that's a prompt away from being fix^H^H^Himproved
Sites is so good that it’s changed my workflow - I start in Opus 5.5, have it write a plan to Google Docs in markdown, then have Codex pick up the plan, shave it down, and build the prototype. A little annoying to switch tools, annoying enough to only do this on desktop, but the output is fantastic.
I think this is the next step in the evolution of AI based tools. We can develop the app or tool, but we need somewhere to easily publish it.
I actually recently shared my own version of this concept, but self hosted. I'm optimistic that there can't be much lock-in from the big providers, but I think there's a large market to capture
They'd better be proactive with checking for abusive users, because I can easily see people trying to spin up honeypots, crypto miners, scam sites, etc. on this. Especially given that the hosting appears to be free with undefined CPU/RAM limits.
Probably the most underwhelming feature of Chatgpt. I dont know why its so slow and overall terrible as compared to Claude artifacts
Looks good. Like most ChatGPT things, though, it will probably take good prompting and iteration, to ensure that the sites have individuality.
Still, I suspect a lot of independent Web developers are sweating. It was already getting tough, but this is a coffin nail.
I wonder if they use version control underneath?
i love sites
i make it for anything, i made one as a food guide for the UW chow down https://uw-chow-down-2026.corruptbytes.chatgpt.site/
I've been using this for probably six weeks. It just popped up as an option one day in Codex.
hows your experience been?
I've found it to be very helpful. A fairly painless approach for hosting a quick prototype.
We’ve been building something like this, but perhaps even more powerful, safe, and beginner friendly. If Sites looks like it’s the right speed for you, we’d love your feedback on https://vibes.diy/
Clicked on it and wondered why it loaded slow.
Lo and behold the homepage bundle size is almost 10mb… Seems to mostly be caused by all of those images. I suggest you guys compress those ASAP.
Is there anything similar that's open source and self hosted and can be hooked into any agent via MCP?
I know that Vercel or something similar is quite close to that, but it's not that well integrated into MCP/Skills.
I have been building https://github.com/openrundev/openrun. OpenRun allows you to build web apps, adding OIDC/SAML with RBAC to control who can access the apps. Supports MCP for app management. Also supports service bindings, which enables providing each app an isolated database account.
My guess is Google and Apple are going to eventually take all the market share for this "app builder" category. They own the device market-share and can offer way better integration
If Apple doesn't up its AI game soon, their entire ecosystem will be rebuilt by AI.
I'd like to have one of this but where the content is dynamically updated by a 'cloud' agent on schedule, is this feasible today?
Yep. https://help.openai.com/en/articles/20001339-creating-and-us...
Does anyone know if Claude makes it easy to publish sites u vibe coded to a custom subdomain or new domain? That’s the biggest reason I use this feature in ChatGPT over Claude
You can ask Claude to deploy it with ShipStatic. It works pretty well for this use case, no account is required, and you can publish to a custom domain/subdomain without much setup.
Disclaimer: I’m one of the people behind ShipStatic.
Pretty sure if you give Claude your AWS keys, it will easily deploy the stock CloudFront setup in minutes. But that's not as seamless.
Not sure I wanna give them my AWS keys...
Put them in a .env file and tell it to never look there but what the env vars are named. Problem solved.
Are there any significant differences between Claude Artifacts and this?
This is full stack apps and deployments for you. BE, database, blob storage functionalities too instead of only static html
Sadly while the example: https://below-the-surface-ocean.openai.chatgpt.site can be detected as SLOP on the first contact of photon with the retina, I honestly cannot detect https://frame-studio-motion.openai.chatgpt.site and it is giving me depression. The website is basic, but it doesn't feel slop. And I guess that is something.
I'm genuinely sad that what we learned with hard work and sweat, our craft, our identity, our soul; is now something you can buy. I've been in this profession for a long time and now I see people who have absolutely no fucking clue, but lots of money, and they buy all these agents and make things that work (way more broken, with absolutely shit quality), but they work and so they brag about their creations to me all day long, and man that makes me feel like a total loser.
I spent decades trying to master media theory, language design, computer graphics, compiler design, rewriting systems, mathematics, logic, build systems, graphs, data-flow systems, making GUI kits, layout systems, ... all to make a better system for thinking, and now the whole idea of thinking is outsourced (as altman says "sold as utility by the meter"). I need to become an addict or something and be wasted. This is not a life.
If it's any consolation, almost no craft will be spared. Software was simply first in line because of how much training data was available, but it won't be the last.
> I honestly cannot detect https://frame-studio-motion.openai.chatgpt.site and it is giving me depression. The website is basic, but it doesn't feel slop. And I guess that is something.
Are we looking at the same website? My eyes immediately go to the font and typeface styling and it registers immediately as AI generated. There’s a very distinct “AI” website aura that Claude and ChatGPT make that requires you to change a ton of the styling to make go away.
I think people are missing the point. To fix the sloppiness all you need is a non-technical person who wanted the site/app to tell it what to fix. Either way the technical middle-person is gone
The copy for that one is very slop, too.
Yeah OP comment is really, really weird. It sounds as slop as the links shared.
Forget the design, you can always tell these are slop by just reading the copy.
I agree the second site is nice looking though.
You spent decades on this and you think that linked website is not slop? You say its "basic" but its really not, its overengineered nonsense if you look closely.
uBlock needs a slop filter post haste.
And I wish there was a SponsorBlock type, crowdsourced slop filter for Youtube.
Been using it for a while. Very good experience so far. Maybe even better than claude or replit
Add WYSIWYG site builders to the list of businesses on the chopping block.
At least they lived a good life as free-range businesses.
Is this better than Claude? Probably not, right?
I haven't used ChatGPT in over a month. I bounce between Grok, Gemini & Claude. Is ChatGPT even worth it.
Grok can do things the other two can't, like read and fetch websites sometimes where Claude and Gemini cannot.
Gemini is just someone to feed javascript to when you want to save on Claude tokens.
It's worth for the token budget. You can do so much with gpt-6.1-sol (high) on a small sub. If the money is not aan issue it's probably still worth it for Astra because it can occasionally be smarter than Fable 5.1 (or so I heard).
This is my impression as well; ChatGPT/Codex is very affordable for what you get.
I’d be firing my entire marketing team if this is all they came up with.
If you want a solution that works with any model/harness, try ht-ml.app!
impressive, but brand websites mostly "eaten" by curated accounts in social media
Seems like Ploy, I think.
rip base 44 /wix
geocities vibes?
wow there goes lovable's moat
Could you just be an API to serve intelligence instead of moving into the app layer and putting your clients out of business. How much wealth and power do you need to accumulate seriously?
The "intelligence API" layer seems like a pretty rough business to be in. The smarter the models the more they're commodities (since they can use any harness or scaffolding) resulting in a race to the bottom on pricing. And then each new better model requires double the cost of the previous one.
I'm surprised there's not a bigger push into better and useful memory, since that would result in actual lock-in.
What is wrong with being a commodity, or a utility? Oil companies do just fine. So do electricity companies. Why does everyone have to seek world domination instead of building a good business. My point is by cannibalizing their clients they risk not being trusted by anyone. But Microsoft did the same thing and created a durable monopoly because Republicans abandoned anti trust enforcement and that seems likely here.
The question of them moving into the app layer is completely irrelevant. If you are threatened by this the reality you need to begin making peace with is that your business is not going to exist within 10 years. And that's being extremely generous.
I hope you support UBI.
This was always the play… Wait until the models are trained to work on “capsules of your data” stored by OpenAI, and deploy your apps/services onto OpenAI cloud where you can _easily_ share the development with your team members’ AI agents. Ofcourse, they will all work only if you use and pay for their models via their subscriptions. A new global {K-V, Relational, Vector} database that your coding model is already trained to work with! will come very soon…
They know the API price wars are a race to the bottom with little/no profit margins.
I think this is going to back fire. I have talked to 5 tech leaders at companies and universities this week about setting up their own infrastructure for inference and training. They are worried about multiple rug pulls: increasing cost of tokens, stealing ideas, training on data etc. they are all willing to uses less intelligence models if they avoid these issues. They’ve seen me setting up propelcompute.com for my start up and propelcode.app neither of which are intended to be commercial products and want to do the same.
Don't worry, we had DreamWeaver, Wordpress and other tools for a long time now but it still did not replace web developers. Build quality tailor made websites that connect to humans and you will be fine.
DreamWeaver and WordPress both had limits on the complexity of site they could build. If you wanted more than what they can do, you look to web developers.
That limitation doesn't exist with ChatGPT
The limitation actually is: your fantasy and knowledge how the thing you want actually should work.
I'll make a note to come back to this comment in a year.
They owe investors something like $200 billion dollars, so no.
Would be great if all AI slop sites lived under a common namespace...
another easily blockable domain at least
If you're wondering why OpenAI's models seem to be struggling to stay ahead of Anthropic and are falling further behind, I think it's because of side quests like this.
They have dots, they had Sora, so many other things that take away attention from just having the best model. There are already a bazillion ways to do this. Sure, the integration's nice, but it's just not needed.
Weird to describe something they could really easily sell to non-coders as a "side quest".
I think programmers make the mistake of thinking generative AI is for us, when it is much more ominously about us.
Newsflash: a company that plans to be the most valuable business on earth and has a desperate need for revenue is not going to stay out of anyone’s lane if there is money to be found in it. If you can make an AI wrapper, so can they.
This actually looks like it has meaningful potential, and given how easily they could train it to connect to a whole bunch of off-the-shelf APIs running in data centres they control, looks like a little glimpse of the future, and of the kind of power the industry has gleefully handed to the crypto-for-your-eyeball-scans guy.
Cloud AI has been the face-eating leopard of the tech industry for the last three years. This is them just starting to chew. How long before they tweak it to generate intranet apps, mini stores, specific AI wrappers? Why wouldn’t a firm that has so much influence over data centre rollout, RAM and processor prices simply directly start using their code generation tools and data centres to build and run the work product?
Maybe some of the big firms who have been helping train ChatGPT to eat their lunch will now wake up to the need for open weights.
> There are already a bazillion ways to do this.
Which means non-technical ChatGPT users are a bit bewildered as to how to pick.
This makes that easy; I suspect normies will fucking love it. My boss will.
I think at this point assessments like this are more qualitative than quantitative. I use Codex far more than I use Claude for coding and my results are generally higher quality. I would not, however, go to someone and tell them that Codex is better than Claude generally. In large part it depends on how you use the model and the harness.
And even if you do know how to use the model and the harness, Claude is just generally still a much a better experience overall.
Returned to Claude as my main driver after many months of Codex and my frustration index is in much, much better place. I can't recommend Codex/GPT to anyone, not just normies. It does make for a really nice reviewer though.
All of these "side quests" (and especially this one) are attempts to build a walled-garden.
Especially this. OpenAI will host it. OpenAI will throw "Sign in with ChatGPT" there. "Apps" will be built that depend on end users ChatGPT "connections". The app will fundamentally not work without an end-user having a ChatGPT account/subscription and "connected apps".
Having "the best model" will absolutely not support a $1T+ valuation. Owning the activity of 25-50% of internet users (assuming a split between OAI/Anthropic/Google/maybe-Meta), is.
OpenAI is trying to build a business to justify their valuation, while they pursue ASI at the same time.
Anthropic is just trying to get to ASI and relying on the blind faith of investors that there will be a return at the end.
What matters is which strategy is more effective for attracting investor capital, because neither can survive without it.
Is it surprising? Most devs at my company (2k+ employees) are using Oppus and casually burning millions token a day.
This take sounds like engineering not "getting" what product even does. OpenAI has a company to run and a business to build.
The frontier companies are scrutinized heavily for their trillion dollar valuations all without a clear mass-consumer use-case. They're trying stuff.
It’s pretty obvious that they are trying to lock in consumers.
They are trying to build an ecosystem that can capture the value of AI, whereas LLMs are quickly becoming a commodity.
Where is the moat? How can they convert users?
If you were sitting on a rocket burning cash to not crash you would do that too.
OpenAI has an enormous team working on Codex. The folks who train models are different people.
OpenAI models are not falling behind Anthropic. They just have different strengths.
in terms of building sites, gpt is ahead e.g. computer vision etc.