Au contraire, based on their recent hires, they're just beginning their Facebook phase. Expect to see a lot of ads 2 years from now, and expect to see a LOT of money being made by the company a year after that.
I predict this won’t be the case. The switching cost to a non Ad LLM is too low for me to be bothered with an ad based LLM.
OpenAI must justify ads when its competitors are not sending ads. Why would they? OpenAI models must be so good that it should be worth dealing with ads compared to say Claude or DeepSeek.
I predict it will be sneaky enough and the ChatGPT brand is strong enough to take it. Your switching cost is low - but you’re likely a tech enthusiast who likes researching and trying alternatives. Most people aren’t like that. The one competitor with anywhere close consumer recognition is Gemini and I’m sure Google would love to push ads too. Maybe not yet as they’re in the reputation building phase, but if it makes money I’m fairly sure everyone will end up doing it.
I suspect that ad powered LLMs will also dramatically sacrifice quality- they are a cost center suddenly, no reason to run a model with 10s or 100s of billions of parameters when some 500m thing will provide a minimally plausible scaffold for ad delivery.
Like everything else, most people won't notice a difference and prioritize free over paid.
The thing is, ads don’t pay that much. Facebook’s ad revenue works out to about $50/user per year . I probably don’t like FB enough to pay $5/mo for it (and even if I did, all the people I interact with that make it useful wouldn’t) but something that’s a personal assistant? Easy.
I do wonder whether we will switch back to ads or there is/will be a big switch to the subscription type model.
Would you trust output from OpenAI which is sponsered? I mean it's bad enough now in the ad space where they are increasingly trying to make the ads look more like content - imagine that in woven into your ChatGPT output?
The younger generation are quite used to subscription models - netflix, spotify, various gaming platforms etc. Perhaps access just becomes part of your internet access bundle.
The problem with charging more for subscription is they have like 5 competitors who can give you the same thing, and charging more is giving them an open opportunity to trade revenue for market share if you undercut them.
I think they're realizing what most capitalists have realized - if you don't own the whole value chain, you don't own the value.
I bet Alex won't have a dropdown to switch to Claude Sonnet for very long.
> I bet Alex won't have a dropdown to switch to Claude Sonnet for very long.
> We plan to continue service for existing users, but will stop new downloads of the app on October 1st. As long as you have the app installed, our plan is to continue serving you. But there won’t be any new features released.
Sounds like there won't be an Alex at all after very long.
From my perspective, I'm so tired of ads being the main source of revenue on the internet. If they're added to chatgpt for example I will most definitely be moving to another that has no ads. Probably I don't represent the masses but I hate this rush to the bottom with adverts.
Ads are a way to get money from users who don’t want to pay with their money. You can’t beat free on price and crucially you can’t beat free on convenience - paying even 1 cent above free introduces complexity and friction.
Solve this and you’ll solve the ad problem, but I’m afraid it isn’t possible because money involves controls and regulations which you can’t weasel out of and not end up in jail.
I was watching a TV show on Prime for the past months. They introduced ads last month and I instantly returned to piracy for this specific show. I hope my main steaming service (for local content) doesn't get the same idea. I'd happily pay extra if they add 4k, but I don't like being extorted into paying (over 50%) more for the same service.
Yeah it could be insanely valuable and it'll give them an advantage by letting them add more features to the free ad supported product to expand their reach and cement their position.
Quick summary, I believe consumer AI experiences will feature ads because the profit opportunity is too large and company valuations depend on it. The hiring of Fidji Simo (ads at Facebook) at OpenAI + and just this week, Vijaye Raji/Statsig also point that way.
We’re in a golden period where AI results are ad-free.
One thing I’ve been doing is querying and storing results. For example, “what are the best books on X topic” for every topic I can possibly think that I may want to read about in the future.
I’ve found the results to be amazing if you give a sufficiently detailed prompt. I have enough reading to see me through to exit.
I've already started getting results infiltrated by SEO but for AI. Deep research did return seemingly a top book when I was looking into Ansible. I independently verified it with my own searching in a few different places.
But the other recommendations seemed like crap and when I followed the sources they seemed like AI generated garbage for AI that I couldn't find doing my normal searching.
One possibility is that if they can manage to lower interest rates back to zero ish, we might see a hiring frenzy in machine learning/llm/genai, starving upstarts and free software of talent, slowing progress in custom local models? Or is this too pessimistic/ "out there" of a take that requires the stars to align just right?
Some models will still trickle down; hell, better/cheaper hardware should enable to run hefty models available today, and they seem to be already okay-ish with such queries.
The training data is full ads. For books, you have publisher-influenced rankings, SEO slop and promotional social media posts. It's GIGO, and has been that way from the start.
I like the thesis, (and probably not a fully informed opinion here), it's easier for openai like router to go the affiliate model than ads model. Router can determine how much a query is worth (eg: help plan a vacation is worth $50 vs say what is the capital of Australia as $0). This further informs how much compute to use, and this can lead to how much they get on affiliate fees if a link is clicked. Ads here are counter to the proposition in the sense that results are to be trusted. Then, despite everything, while contextual ads are good, personalized ads are still going to be more effective (gut feel). To get there, openai needs the kind of infra Facebook and Google has. They can get the same value (perhaps more) from affiliate links - especially in a world where they are kind of gateway to the discovery - and don't have to do as much work on the infra side. This also aligns incentives for all three - companies, consumers, and middlemen, in a way it only happened with Google before this.
ChatGPT is already returning lists of products (with photos and rating) saying they are impartial, I guess to collect affiliate fees. It's not a big jump to have sponsored products showing up first.
If folks corrupt the integrity of LLM responses, as it were, they’ll destroy the value proposition.
More likely the model will be payment for low friction enablement of transactions rather than overt steering. Pick Door #1: the LLM states the product is fit for purpose. Pick Door #2: the LLM will directly complete the transaction or close to it.
> If folks corrupt the integrity of LLM responses, as it were, they’ll destroy the value proposition.
Highly unlikely. Google's SERP is an ad-infested abomination that sometimes shows useful results, and yet people still use Google Search.
The same will happen with LLMs, except in far more subtle and insidious ways. Instead of showing you ads directly, they will be naturally interwoven in conversations, suggestions, and generated content. You won't be able to tell whether the content is genuine or promoted, as is common on the web today.
The ads will target you more accurately than ever before based on not just the data you've given them, but on the context of the conversation, your surroundings, and any other piece of real-time information they can use to secure a conversion, or to influence your thoughts on a particular matter. You will trust it more than any current ad channel since the AI will be personal, and the tone will be friendly.
As with the web, ad-free services will exist, but the only way to escape this entirely will be to use local and self-hosted models.
> Google's SERP is an ad-infested abomination that sometimes shows useful results, and yet people still use Google Search.
I've personally stopped doing that. I understand it's not a mainstream decision since most people don't even know what alternatives are there, but doesn't mean that in a couple of years google won't start to feel it.
OTOH with the help of local models you should be able to post process any and all content on device to at least highlight, desensationalize if not outright remove quite a lot of those ads, at least until DRM folks lay their hands on it.
Definitely OpenAI. They’re not profitable and their core product is expensive to serve. They also have the disadvantage that their better models and features (eg agents, research, CoT, etc) are more expensive, but are better hooks to prove utility to new users - fundamentally they need a method to cheapen the cost of serving free users better features. Google can afford to see how it shakes out and the impact to OpenAI.
OpenAI already exploring and experimenting with different ad modalities privately. They’re also have a much better brand, so they might be able to avoid too much churn of customers.
ooo charge advertisers for % of time that their mcp context is influencing the response that gets generated. insiiidious. you wouldnt even know it was an ad.
Expensive part is providing the services at loss to get the biggest part of the market, which is the current state of affairs. At some point the market leader will emerge and it will price things according to the actual cost of running this thing + markup, and ALL efforts would be going into squeezing as much profit as possible from the leading position. At the same time usability and quality would not be priority anymore. We are enjoying the late-capitalism at it's peak here.
I'm just saying what Anthropic is saying: individual models are printing money. The problem is all the capital needed to get the next model before the previous ones are obsolete; investors capitalize the business because the future models are expected to print even more money, at least a few months ago. It's increasingly obvious that the new models won't print as much money as expected to justify the expenses, but if they all stopped training today they'd be swimming in cash in a couple years.
>if they all stopped training today they'd be swimming in cash in a couple years.
If they stopped training, then the data cutoff date is one year farther each year? How do you make money from model which is stale on data and doesn't include any recent stuff?
So far it's about as subtle as a slap to the face. If you set ChatGPT's personality to 'straight-shooting' it starts every answer with "blunt", "tell it like it is" or "unvarnished" and permeates that across the whole reply; by its very nature, it is unable to be subtle about anything.
I have Default personality with the traits field set to straight-shooter ("Tell it like it is; don't sugar-coat responses.") I'll try the robot one, thanks.
Yeah, that's a recent addition, as far as I can tell. It became available in my account a few weeks ago. I currently use 'Robot' mode, call myself a 'Researcher', and use these custom instructions:
Answer concisely when appropriate, more extensively
when necessary. Avoid rhetorical flourishes, bonhomie,
and (above all) cliches. Take a forward-thinking view.
OK to be mildly positive and encouraging but NEVER
sycophantic or cloying. Above all, NEVER use the
phrase "You're absolutely right." Rather than "Let me
know if..." style continuations, you may list a set
of prompts to explore further topics, but only when
clearly appropriate.
Pretty happy with the results so far. Very low BS factor (although it does ignore the last part sometimes.)
This is true, but it doesn't go far enough. Google could serve you ads, like LLMs will.
But LLMs can be used to astroturf internet spaces. Which means they allow everyone the ability to serve ads and manipulate. It's no longer just limited to the company providing the original service.
Check out https://deep.ai lots of people who don't have $20/month to spend on a chatbot are happy to have ad impressions subsidize what model they get to use
Given employees (especially at higher levels) of companies wouldnt pass their own "interview" process its a legitimate tactic to getting into the buzier companies these days
The details are scant, but if I had to guess Alex probably ran out of money and after Apple showed off their Xcode AI integrations they couldn't get more funding and shopped around for an acquisition.
Daniel and his team did a great job at doing what Apple wasn't but I can't help but feel this was an inevitable outcome.
Since time immemorial acquired companies have claimed this sort of thing and then the acquiring companies have shut it down quick. Practically a meme at this point.
I still can't forgive Apple to buy darksky and then killing it. It was so good and Apple Weather is still not even close. Why are these big companies buying small companies that have a great product and just let it die without using their tech?
A coding agent that receives no updates will be useless very soon. In a few months they’ll tell you to switch to Codex CLI, and that will in fact be a good recommendation.
The failed Windsurf bid and this makes me think OpenAI feels they need to focus more on the coding agent use case.
Still thinking about the endgame. Its not obvious to me if OpenAI/Anthropic will become competitors to coding startups like Cursor or continue to be model providers.
I think the endgame is a shift toward a platform of services that tightly bind users to a single LLM provider.
Right now, many small startups are essentially just thin wrappers around ChatGPT. Once it becomes clear which ideas and solutions gain real traction, providers like OpenAI/Anthropic can simply roll out those features natively removing any need for a third party.
In a sense, a lot of what happened with the mobile market. For example, there's no need for a QR scanner or document scanner app anymore, if your phone starts to offer it natively.
Does anyone else remember how we used to have flash light apps all over the playstore and how they quickly varnished once the feature was implement natively?
Almost everything in ios since ios1 has been a sort of in house copy of a jailbroken tweak. I feel like that is just how these huge companies actually innovate featuresets now. No one has agency to drive change from within, so they see what the community has built and is using and shamelessly copy it as it. Easier to sell to management that you should steal an already bright idea than to try and prove your own novel idea is worth pursuing.
This seems correct to me. I'm unusually appreciative of vertical integration (life-long Apple user, etc.), and I can already feel the vendor lock-in tightening. I have no need for anything other than my ChatGPT subscription, and adding other tools appears to offer marginal gain at double the cost.
This type of bundling appears to be one of the strongest forces in the economy today, and I think comes about consistently due to a confluence of efficiencies of scale, coordination, and second-order effects of prestige (being able to hire and pay large numbers of outlier high performing employees, etc.)
I've learned not to bet against it, except in niche areas.
Neither type has agents in their DNA. The IDE agent companies are dragging an enormous vestigial appendage, and the the scientists at frontier labs are stuck in the next token prediction mindset. This needs a systems engineering approach with online learning from feedback and high throughput optimization experiments run end to end.
> Its not obvious to me if OpenAI/Anthropic will become competitors to coding startups like Cursor or continue to be model providers.
They’ll do both: continue to be model providers while also leveraging their position as model providers to own as many of the valauble markets in which models are used as possible. Kind of like Amazon and its role as both infrastructure provider and direct competitors to other sellers (on the shopping/logistics side) and SaaS vendors (on the AWS side).
Could be a nice move for them and it sounds like they have some valuable skills they could add to the Codex team.
I'm pretty impressed with the web ui of Codex lately. The cli ui didn't really impress me when it came out (a bit flaky and buggy and tedious to use). But the web UI is nice. I've created a few alright PRs with it. I think I have about 60-70% merge rate, some with manual changes made by me on the same branch. I like the mode of just keeping that stuff on a branch and interacting via Git.
It has its limitations but I do like this UX and DX. And I've so far not experienced any rate limiting on the plus plan.
I'm curious to learn how others are feeling about this. I know Claude Code and other solutions are popular. But how do they stack up in terms of usability and utility compared to codex?
Hrm, this makes sense their new model Alex was okay but due to XCode having Chatgpt and Claude access if Chatgpt wants people that have more experience with Swift / SwiftUI and all of the platforms. I have used Alex codes for a long time and was already planning to move to XCode's integration with Claude since I find Chatgpt 5 to be lacking. I also wouldn't see how viable the app is long term expecially with the power users who pay $200 for it not breaking even since they are still proxying another service (but they did create their own LLM which IMHO was worse than claude).
So companies like this one are being acquired because they just have really good/niche system prompts? I mean, they all use the same models as everyone else, right?
I'm speculating here, but my guess is that the urgency for OpenAI isn't that the Alex team has a "secret prompt." It's that they've already done the heavy lifting to make LLM coding assistants actually usable in a domain (Apple platforms) where OAI doesn't have integrations yet.
Acquiring them gives OAI:
- ready-made team that understands the IDE plumbing and the developer UX at a deep level
- head start in a platform ecosystem that's hard to crack
3. team that knows how to push the models into productized, developer-ready experiences
So it's not the prompts. It's engineering the scaffolding and UX around the model so it feels like magic to the user. That's what OpenAI is buying.
Also possible that Alex solved a lot of UX and integration challenges that could translate to other Apple contexts (productivity apps, design tools, even consumer-facing AI on macOS/iOS).
Without having used Alex myself (I don't do iOS or macOS development), I would guess that all the retrieval, context slicing, editor integration yada yada that they've built aren't necessarily unique to coding. Same scaffolding could support things like AI-driven writing, design, or general productivity in an Apple-native way.
I personally enjoy using Alex code even though I also use claude code and cursor. But saw they will join openAi today, and I am wondering whether this is more like joining openAi without an interview or openAi really believes Alex code is an excellent product and would like to integrate the features into their codex?
The word joining always annoys me. I doubt very much that the Alex team is joining OpenAI, they where bought out. Joining implies some sort of semi-equal partnership, which I very much doubt this is.
Something mentioned in WWDC stuck out to me (think it was Craig) about current state of LLMs not living up to their high standard of quality. Because they absolutely could (and kind of are, with the foundation models stuff) build their own. Letting Claude or OpenAI handle the Xcode changes via their integration in the next Xcode makes more sense in that context.
Are they basically buying users here? I agree with the other comments about Claude being better, and now that the new Xcode will ship with a Claude integration I wonder about this.
absolutely no. Alex is heavily optimized for xcode. If you work on an ios project that has more than 700 files, you'll understand how accurate it captures the context.
It's not that Claude can't but iOS / Apple development is such a bear and beast. You could really build a whole model just to solve for that ecosystem. There are huge issues in the Apple ecosystem with documentation and so much tribal knowledge. In many cases it's hard to know what is the right thing to do, and there is keeping up with all of Apple's required changes.
So Claude could do it, it just seems like they're focusing on a different set of developers for the moment.
I don't think that makes it as special as you think it does.
> There are huge issues in the Apple ecosystem with documentation and so much tribal knowledge.
I struggle to come up with an ecosystem where that doesn't apply. React, Angular, .NET, .... Though some of them probably even suffer from overdocumentation, e.g. React with the same beginner level tutorials / open source code regurgitating bad patterns, and you then have the challenge of separating the wheat from the chaff.
The question is really whether maintaining an ecosystem-specific model would be able outperform a better generalized coding model, and even further whether the marginal improvements would justify the additional maintenance process/cost.
In your mind is 700 files a lot or a little? It feels very small to me, and Xcode really ought to be able to handle that tiny scale on modern machines with ease.
I struggle to imagine a team of more than 10 people writing an iOS app with less than 700 files.
In fairness, as a mere generator of eyeball time that gets mis-sold* to advertisers, I'd say the FB user experience is very much "doing something wrong".
* dick pills and boob surgery, also government announcements for a country I don't live in, also offers to help renounce a citizenship I never had in the first place
I mean.. say its an enterprise mobile app. Maybe there are 2 shells, each shell has 5 tabs. Each tab might have 5 screens on it.. that's 50 files already just for the screens. Each screen might have various UI components or steppers, etc.
Most noobs, such as those who think 700 files is too many because they've only worked on apps they never published, might just cram everything into that one file.
However, there would be various files for components, functions, etc. Code that's single responsibility and easy to test might mean there are lots of files. There might be upload queues, offline functionality, custom code to go beyond what the ios/android SDKs offer, and so on. DTOs, DAOs, etc. various services..
This was such a fun read. Eels sound like they’re living in a sci-fi cycle: mud, rivers, Freud’s lab, then the Bermuda Triangle. Nature really doesn’t care about fitting into neat categories.
At this rate it’s better to start a company and get aquihired vs applying and getting hired.
Seems like OpenAI speed ran through the Facebook phase and are out of ideas
Au contraire, based on their recent hires, they're just beginning their Facebook phase. Expect to see a lot of ads 2 years from now, and expect to see a LOT of money being made by the company a year after that.
If they can maintain runway until then.
I predict this won’t be the case. The switching cost to a non Ad LLM is too low for me to be bothered with an ad based LLM.
OpenAI must justify ads when its competitors are not sending ads. Why would they? OpenAI models must be so good that it should be worth dealing with ads compared to say Claude or DeepSeek.
I predict it will be sneaky enough and the ChatGPT brand is strong enough to take it. Your switching cost is low - but you’re likely a tech enthusiast who likes researching and trying alternatives. Most people aren’t like that. The one competitor with anywhere close consumer recognition is Gemini and I’m sure Google would love to push ads too. Maybe not yet as they’re in the reputation building phase, but if it makes money I’m fairly sure everyone will end up doing it.
> The switching cost to a non Ad LLM is too low for me to be bothered with an ad based LLM.
Provided one with sufficient quality exists.
Yes so what amount of quality are you willing to give up so that you won't be bombarded with ads?
Or, what extra money would you pay OpenAI to get a non ad powered LLM model?
I suspect that ad powered LLMs will also dramatically sacrifice quality- they are a cost center suddenly, no reason to run a model with 10s or 100s of billions of parameters when some 500m thing will provide a minimally plausible scaffold for ad delivery.
Like everything else, most people won't notice a difference and prioritize free over paid.
The thing is, ads don’t pay that much. Facebook’s ad revenue works out to about $50/user per year . I probably don’t like FB enough to pay $5/mo for it (and even if I did, all the people I interact with that make it useful wouldn’t) but something that’s a personal assistant? Easy.
It’s not hard to imagine ChatGPT just charging.
Unless they all have ads, which is a good way to justify spending all that capex on training.
That requires you to be able to notice the ads.
I do wonder whether we will switch back to ads or there is/will be a big switch to the subscription type model.
Would you trust output from OpenAI which is sponsered? I mean it's bad enough now in the ad space where they are increasingly trying to make the ads look more like content - imagine that in woven into your ChatGPT output?
The younger generation are quite used to subscription models - netflix, spotify, various gaming platforms etc. Perhaps access just becomes part of your internet access bundle.
The problem with charging more for subscription is they have like 5 competitors who can give you the same thing, and charging more is giving them an open opportunity to trade revenue for market share if you undercut them.
I think they're realizing what most capitalists have realized - if you don't own the whole value chain, you don't own the value.
I bet Alex won't have a dropdown to switch to Claude Sonnet for very long.
> I bet Alex won't have a dropdown to switch to Claude Sonnet for very long.
> We plan to continue service for existing users, but will stop new downloads of the app on October 1st. As long as you have the app installed, our plan is to continue serving you. But there won’t be any new features released.
Sounds like there won't be an Alex at all after very long.
> if you don't own the whole value chain, you don't own the value.
Good rule. Just like using a service based AI instead of a self-hosted one if you are a developer or artist.
From my perspective, I'm so tired of ads being the main source of revenue on the internet. If they're added to chatgpt for example I will most definitely be moving to another that has no ads. Probably I don't represent the masses but I hate this rush to the bottom with adverts.
Ads are a way to get money from users who don’t want to pay with their money. You can’t beat free on price and crucially you can’t beat free on convenience - paying even 1 cent above free introduces complexity and friction.
Solve this and you’ll solve the ad problem, but I’m afraid it isn’t possible because money involves controls and regulations which you can’t weasel out of and not end up in jail.
What about tiers like YouTube lite where you pay and you still see advertisement ?
How does that fit in your price model because to me it makes no sense at all yet people buy that. Same for Netflix tier with ads.
There's money to take, maybe, so why not take it.
I was watching a TV show on Prime for the past months. They introduced ads last month and I instantly returned to piracy for this specific show. I hope my main steaming service (for local content) doesn't get the same idea. I'd happily pay extra if they add 4k, but I don't like being extorted into paying (over 50%) more for the same service.
Yeah it could be insanely valuable and it'll give them an advantage by letting them add more features to the free ad supported product to expand their reach and cement their position.
> Expect to see a lot of ads 2 years from now
I think I am just slow today - but could you please elaborate?
I wrote about this idea here: https://digitalseams.com/blog/the-ai-lifestyle-subsidy-is-go...
Quick summary, I believe consumer AI experiences will feature ads because the profit opportunity is too large and company valuations depend on it. The hiring of Fidji Simo (ads at Facebook) at OpenAI + and just this week, Vijaye Raji/Statsig also point that way.
We’re in a golden period where AI results are ad-free.
One thing I’ve been doing is querying and storing results. For example, “what are the best books on X topic” for every topic I can possibly think that I may want to read about in the future.
I’ve found the results to be amazing if you give a sufficiently detailed prompt. I have enough reading to see me through to exit.
I've already started getting results infiltrated by SEO but for AI. Deep research did return seemingly a top book when I was looking into Ansible. I independently verified it with my own searching in a few different places.
But the other recommendations seemed like crap and when I followed the sources they seemed like AI generated garbage for AI that I couldn't find doing my normal searching.
By the time this „golden“ ad free time is over in 1-2 years you should be able to run your own custom model locally
I hope so but we could also be staying in an eternal state of FOMO as the proprietary models keep getting marginally (or a lot) better.
How are they better if they spam the user with ads?
One possibility is that if they can manage to lower interest rates back to zero ish, we might see a hiring frenzy in machine learning/llm/genai, starving upstarts and free software of talent, slowing progress in custom local models? Or is this too pessimistic/ "out there" of a take that requires the stars to align just right?
Interest rates at zero and there won't be any clients for these upstarts as people will be in bread lines from their worthless money.
Some models will still trickle down; hell, better/cheaper hardware should enable to run hefty models available today, and they seem to be already okay-ish with such queries.
Would you consider sharing these somewhere?
The training data is full ads. For books, you have publisher-influenced rankings, SEO slop and promotional social media posts. It's GIGO, and has been that way from the start.
(me again) and Drew Breunig just posted about the tensions of actually getting those ads in: https://www.dbreunig.com/2025/09/02/considering-ad-models-fo...
I like the thesis, (and probably not a fully informed opinion here), it's easier for openai like router to go the affiliate model than ads model. Router can determine how much a query is worth (eg: help plan a vacation is worth $50 vs say what is the capital of Australia as $0). This further informs how much compute to use, and this can lead to how much they get on affiliate fees if a link is clicked. Ads here are counter to the proposition in the sense that results are to be trusted. Then, despite everything, while contextual ads are good, personalized ads are still going to be more effective (gut feel). To get there, openai needs the kind of infra Facebook and Google has. They can get the same value (perhaps more) from affiliate links - especially in a world where they are kind of gateway to the discovery - and don't have to do as much work on the infra side. This also aligns incentives for all three - companies, consumers, and middlemen, in a way it only happened with Google before this.
Yeah, I agree. Almost mentioned in the post how I imagine an ad PM at OpenAI is jealous of an ad PM at Perplexity.
Affiliate links are literally ads
ChatGPT is already returning lists of products (with photos and rating) saying they are impartial, I guess to collect affiliate fees. It's not a big jump to have sponsored products showing up first.
If folks corrupt the integrity of LLM responses, as it were, they’ll destroy the value proposition.
More likely the model will be payment for low friction enablement of transactions rather than overt steering. Pick Door #1: the LLM states the product is fit for purpose. Pick Door #2: the LLM will directly complete the transaction or close to it.
> If folks corrupt the integrity of LLM responses, as it were, they’ll destroy the value proposition.
Highly unlikely. Google's SERP is an ad-infested abomination that sometimes shows useful results, and yet people still use Google Search.
The same will happen with LLMs, except in far more subtle and insidious ways. Instead of showing you ads directly, they will be naturally interwoven in conversations, suggestions, and generated content. You won't be able to tell whether the content is genuine or promoted, as is common on the web today.
The ads will target you more accurately than ever before based on not just the data you've given them, but on the context of the conversation, your surroundings, and any other piece of real-time information they can use to secure a conversion, or to influence your thoughts on a particular matter. You will trust it more than any current ad channel since the AI will be personal, and the tone will be friendly.
As with the web, ad-free services will exist, but the only way to escape this entirely will be to use local and self-hosted models.
> Google's SERP is an ad-infested abomination that sometimes shows useful results, and yet people still use Google Search.
I've personally stopped doing that. I understand it's not a mainstream decision since most people don't even know what alternatives are there, but doesn't mean that in a couple of years google won't start to feel it.
OTOH with the help of local models you should be able to post process any and all content on device to at least highlight, desensationalize if not outright remove quite a lot of those ads, at least until DRM folks lay their hands on it.
>yet people still use Google Search
Google search is widely acknowledged as drastically drying up in the last year or so, that's despite more worldwide internet usage.
IMHO a huge part of people choosing ChatGPT over Google for search is due to the absence of ads and other distractions.
Doing search with ChatGPT feels like doing research in a library setting whereas Google search feels like doing research in Times Square.
Many people still use google search probably because the LLM response is at the top.
Insightful article, thanks for writing/sharing!
Very good catch. For all my interest this space, it's embarrassing that I did not remotely see this coming...
Ads on ChatGPT as a way to extract more money from users
And I'm betting they won't be shown as a clearly marked box that says "Ad". It'll be woven directly in the response like normal content.
They are next. Perplexity's Comet browser tracks the ** out of you already because ads.
OpenAI already runs ads on Instagram for its ChatGPT app, advertising what it can do for you and your day to day things.
Who will pull the ads trigger first, OpenAI or Google I wonder. Whoever leads could lose a lot of users to the other, but they will soon follow.
Definitely OpenAI. They’re not profitable and their core product is expensive to serve. They also have the disadvantage that their better models and features (eg agents, research, CoT, etc) are more expensive, but are better hooks to prove utility to new users - fundamentally they need a method to cheapen the cost of serving free users better features. Google can afford to see how it shakes out and the impact to OpenAI.
OpenAI already exploring and experimenting with different ad modalities privately. They’re also have a much better brand, so they might be able to avoid too much churn of customers.
Gemini can already sponsor other Google products like Google docs sheets and youtube
Their non-profit status may be an impediment to maximising profits.
I see mcps from ebay alike coming to openai.
ooo charge advertisers for % of time that their mcp context is influencing the response that gets generated. insiiidious. you wouldnt even know it was an ad.
Exactly. Definitely coming.
oAI rev exceeds $1b per month right now.
still not profitable? All this revenue is burned on compute
Individual models are probably profitable because why not.
Not staying behind is the expensive part.
Expensive part is providing the services at loss to get the biggest part of the market, which is the current state of affairs. At some point the market leader will emerge and it will price things according to the actual cost of running this thing + markup, and ALL efforts would be going into squeezing as much profit as possible from the leading position. At the same time usability and quality would not be priority anymore. We are enjoying the late-capitalism at it's peak here.
> providing the services at loss to get the biggest part of the market
Would note that market share confers revenue, users and data. The last is uniquley valuable for LLM builders.
I'm just saying what Anthropic is saying: individual models are printing money. The problem is all the capital needed to get the next model before the previous ones are obsolete; investors capitalize the business because the future models are expected to print even more money, at least a few months ago. It's increasingly obvious that the new models won't print as much money as expected to justify the expenses, but if they all stopped training today they'd be swimming in cash in a couple years.
>if they all stopped training today they'd be swimming in cash in a couple years.
If they stopped training, then the data cutoff date is one year farther each year? How do you make money from model which is stale on data and doesn't include any recent stuff?
1. They can't stop training
2. I don't trust them anyway that without training they are making money.
Doesn’t have to be profitable atm
For sure. Needs to drain as much money as possible from Microsoft first. I approve!
Be careful what you wish for if you use anything from MS. Further enshittification will be mercyless.
[dead]
I do what I can to waste OpenAI cash by getting the free version of chatgpt to write dirty limericks about Sam Altman.
Are people really going to keep using these AI tools if they start shoving ads down our throats?
Just by the nature of the product, the manipulation will be a lot more subtle, and likely more successful than traditional ads..
So far it's about as subtle as a slap to the face. If you set ChatGPT's personality to 'straight-shooting' it starts every answer with "blunt", "tell it like it is" or "unvarnished" and permeates that across the whole reply; by its very nature, it is unable to be subtle about anything.
That hasn't been true for a while IME. Did you select 'Robot' as the personality?
I have Default personality with the traits field set to straight-shooter ("Tell it like it is; don't sugar-coat responses.") I'll try the robot one, thanks.
Yeah, that's a recent addition, as far as I can tell. It became available in my account a few weeks ago. I currently use 'Robot' mode, call myself a 'Researcher', and use these custom instructions:
Pretty happy with the results so far. Very low BS factor (although it does ignore the last part sometimes.)"I feel like something is missing in my life"
> Thinking...
> Have you ever had that energetic and refreshing feeling after a Baja Blast(TM)?
This is true, but it doesn't go far enough. Google could serve you ads, like LLMs will.
But LLMs can be used to astroturf internet spaces. Which means they allow everyone the ability to serve ads and manipulate. It's no longer just limited to the company providing the original service.
I think they'd inject ads during training (too). You and everyone cannot do that; Google can.
why would they inject ads during training?
It's terrifying.
And yet somehow it's even worse.
Maybe something good comes out of the work to only push products that actually exist. /s
Worked for every tech company so far.
Check out https://deep.ai lots of people who don't have $20/month to spend on a chatbot are happy to have ad impressions subsidize what model they get to use
Do you use any free web search tool?
I actually switched from Google to Kagi a few years ago. Whenever possible I avoid products with advertisements.
From what I've seen you still need to be well connected.
I have at least one more idea in me and invest my own money to make it happen, but I don't think I'll have a 2 billion dollar exit.
Given employees (especially at higher levels) of companies wouldnt pass their own "interview" process its a legitimate tactic to getting into the buzier companies these days
I love your username. So many good memories from that show.
> At this rate it’s better to start a company and get aquihired vs applying and getting hired.
Then do it.
Since when did HN become like Reddit? Always negative about everything?
Feels like all the losers from Reddit have somehow migrated here.
The details are scant, but if I had to guess Alex probably ran out of money and after Apple showed off their Xcode AI integrations they couldn't get more funding and shopped around for an acquisition.
Daniel and his team did a great job at doing what Apple wasn't but I can't help but feel this was an inevitable outcome.
I don't use Alex so this doesn't affect me, but:
> our plan is to continue serving you.
How many months until OpenAI no longer feels it?
Since time immemorial acquired companies have claimed this sort of thing and then the acquiring companies have shut it down quick. Practically a meme at this point.
I still can't forgive Apple to buy darksky and then killing it. It was so good and Apple Weather is still not even close. Why are these big companies buying small companies that have a great product and just let it die without using their tech?
It's not only a meme, it's a Tumblog too, and I first saw it many years ago.
Google "our incredible journey".
https://www.google.com/search?q=our+incredible+journey
https://www.gyford.com/phil/writing/2013/02/27/our-incredibl...
https://www.tumblr.com/ourincrediblejourney
" but will stop new downloads of the app on October 1st."
This tells me that they wont be supporting it forever and at some point will sunset it. I could be totally wrong of course.
12-60, but most of the acquisitions I seen executed: the teams are subsumed and the product put on life-support around 18 months in.
A coding agent that receives no updates will be useless very soon. In a few months they’ll tell you to switch to Codex CLI, and that will in fact be a good recommendation.
Highly likely the next major Xcode update will break it.
December
The failed Windsurf bid and this makes me think OpenAI feels they need to focus more on the coding agent use case.
Still thinking about the endgame. Its not obvious to me if OpenAI/Anthropic will become competitors to coding startups like Cursor or continue to be model providers.
I think the endgame is a shift toward a platform of services that tightly bind users to a single LLM provider.
Right now, many small startups are essentially just thin wrappers around ChatGPT. Once it becomes clear which ideas and solutions gain real traction, providers like OpenAI/Anthropic can simply roll out those features natively removing any need for a third party.
In a sense, a lot of what happened with the mobile market. For example, there's no need for a QR scanner or document scanner app anymore, if your phone starts to offer it natively.
Does anyone else remember how we used to have flash light apps all over the playstore and how they quickly varnished once the feature was implement natively?
Almost everything in ios since ios1 has been a sort of in house copy of a jailbroken tweak. I feel like that is just how these huge companies actually innovate featuresets now. No one has agency to drive change from within, so they see what the community has built and is using and shamelessly copy it as it. Easier to sell to management that you should steal an already bright idea than to try and prove your own novel idea is worth pursuing.
This seems correct to me. I'm unusually appreciative of vertical integration (life-long Apple user, etc.), and I can already feel the vendor lock-in tightening. I have no need for anything other than my ChatGPT subscription, and adding other tools appears to offer marginal gain at double the cost.
This type of bundling appears to be one of the strongest forces in the economy today, and I think comes about consistently due to a confluence of efficiencies of scale, coordination, and second-order effects of prestige (being able to hire and pay large numbers of outlier high performing employees, etc.)
I've learned not to bet against it, except in niche areas.
Neither type has agents in their DNA. The IDE agent companies are dragging an enormous vestigial appendage, and the the scientists at frontier labs are stuck in the next token prediction mindset. This needs a systems engineering approach with online learning from feedback and high throughput optimization experiments run end to end.
In my view Anthropic is already a competitor to Cursor, while also being a model provider to them.
OpenAI has been trying to get into the space with their multiple product offerings all called “codex” but execution has been lacking.
So this is very much a play at becoming more competitive in the space.
> Its not obvious to me if OpenAI/Anthropic will become competitors to coding startups like Cursor or continue to be model providers.
They’ll do both: continue to be model providers while also leveraging their position as model providers to own as many of the valauble markets in which models are used as possible. Kind of like Amazon and its role as both infrastructure provider and direct competitors to other sellers (on the shopping/logistics side) and SaaS vendors (on the AWS side).
> Still thinking about the endgame.
It is said by some that the endgame is agents, like Devin. The IDEs are just the start of AI coding.
I know that we like unmodified titles here, but it would have been helpful to spell out who "we" refers to in this case.
I'm learning live what I had only previously vaguely understood as "exit strategy"
I'd never heard of Alex before today
Could be a nice move for them and it sounds like they have some valuable skills they could add to the Codex team.
I'm pretty impressed with the web ui of Codex lately. The cli ui didn't really impress me when it came out (a bit flaky and buggy and tedious to use). But the web UI is nice. I've created a few alright PRs with it. I think I have about 60-70% merge rate, some with manual changes made by me on the same branch. I like the mode of just keeping that stuff on a branch and interacting via Git.
It has its limitations but I do like this UX and DX. And I've so far not experienced any rate limiting on the plus plan.
I'm curious to learn how others are feeling about this. I know Claude Code and other solutions are popular. But how do they stack up in terms of usability and utility compared to codex?
Hrm, this makes sense their new model Alex was okay but due to XCode having Chatgpt and Claude access if Chatgpt wants people that have more experience with Swift / SwiftUI and all of the platforms. I have used Alex codes for a long time and was already planning to move to XCode's integration with Claude since I find Chatgpt 5 to be lacking. I also wouldn't see how viable the app is long term expecially with the power users who pay $200 for it not breaking even since they are still proxying another service (but they did create their own LLM which IMHO was worse than claude).
So companies like this one are being acquired because they just have really good/niche system prompts? I mean, they all use the same models as everyone else, right?
I'm speculating here, but my guess is that the urgency for OpenAI isn't that the Alex team has a "secret prompt." It's that they've already done the heavy lifting to make LLM coding assistants actually usable in a domain (Apple platforms) where OAI doesn't have integrations yet.
Acquiring them gives OAI:
- ready-made team that understands the IDE plumbing and the developer UX at a deep level
- head start in a platform ecosystem that's hard to crack
3. team that knows how to push the models into productized, developer-ready experiences
So it's not the prompts. It's engineering the scaffolding and UX around the model so it feels like magic to the user. That's what OpenAI is buying.
Also possible that Alex solved a lot of UX and integration challenges that could translate to other Apple contexts (productivity apps, design tools, even consumer-facing AI on macOS/iOS).
Without having used Alex myself (I don't do iOS or macOS development), I would guess that all the retrieval, context slicing, editor integration yada yada that they've built aren't necessarily unique to coding. Same scaffolding could support things like AI-driven writing, design, or general productivity in an Apple-native way.
could just be the git/apply changes functionality they’re interested in. obviously speculation
I personally enjoy using Alex code even though I also use claude code and cursor. But saw they will join openAi today, and I am wondering whether this is more like joining openAi without an interview or openAi really believes Alex code is an excellent product and would like to integrate the features into their codex?
The word joining always annoys me. I doubt very much that the Alex team is joining OpenAI, they where bought out. Joining implies some sort of semi-equal partnership, which I very much doubt this is.
I read the GP comment as referring to an acquihire situation, not joining as any level of partnership role in the OpenAI business.
I totally do not understand why Apple did not buy them. It’s not that they have anything like that already.
Something mentioned in WWDC stuck out to me (think it was Craig) about current state of LLMs not living up to their high standard of quality. Because they absolutely could (and kind of are, with the foundation models stuff) build their own. Letting Claude or OpenAI handle the Xcode changes via their integration in the next Xcode makes more sense in that context.
The high standard of quality that is Xcode?
lol
As other comments have mentioned, Xcode 26 has sherlocked a lot of the functionality of Alex.
Not everything, but the majority of the features are built in.
The product is likely just not very viable after that
Are they basically buying users here? I agree with the other comments about Claude being better, and now that the new Xcode will ship with a Claude integration I wonder about this.
No they're buying engineers.
Congrats to you and the whole Alex team Huge milestone, and well deserved. Excited to see what you’ll build with the Codex team at OpenAI!
Tech is full of bandwagons and its certainly one way to get paid well.
Hopefully they make a chunk.
let's build AI together they said, it'll be fun they said :)
I do not know it's proper to say, but congrats!!
makes sense, it kinda not super useful once xcode added native AI coding features
I misread the source as "alexjones dot app" first and I was impressed.
stoked for this, think it will be great for Codex!
I actually feel a little bit disappointed as they no longer rolls out new features after oct 1st :(
isn't this useless with Claude Code now?
absolutely no. Alex is heavily optimized for xcode. If you work on an ios project that has more than 700 files, you'll understand how accurate it captures the context.
> Alex is heavily optimized for xcode.
Is there a reason why the Claude cannot do the same thing?
It's not that Claude can't but iOS / Apple development is such a bear and beast. You could really build a whole model just to solve for that ecosystem. There are huge issues in the Apple ecosystem with documentation and so much tribal knowledge. In many cases it's hard to know what is the right thing to do, and there is keeping up with all of Apple's required changes.
So Claude could do it, it just seems like they're focusing on a different set of developers for the moment.
I don't think that makes it as special as you think it does.
> There are huge issues in the Apple ecosystem with documentation and so much tribal knowledge.
I struggle to come up with an ecosystem where that doesn't apply. React, Angular, .NET, .... Though some of them probably even suffer from overdocumentation, e.g. React with the same beginner level tutorials / open source code regurgitating bad patterns, and you then have the challenge of separating the wheat from the chaff.
The question is really whether maintaining an ecosystem-specific model would be able outperform a better generalized coding model, and even further whether the marginal improvements would justify the additional maintenance process/cost.
If your iOS project has more than 700 files... you might be doing it wrong?
In your mind is 700 files a lot or a little? It feels very small to me, and Xcode really ought to be able to handle that tiny scale on modern machines with ease.
I struggle to imagine a team of more than 10 people writing an iOS app with less than 700 files.
Bear in mind 600 of those files are icon and screenshot variants for various screen dpis and spec ratios..
Translation files, themes, drawables .etc...the list is endless. Even a simple app will easily have a thousand files.
Instagram.app likely has 30,000 files for iOS. And it produces 10-figures of revenue. So how is that wrong?
https://www.reddit.com/r/programming/comments/3h52yk/someone...
In fairness, as a mere generator of eyeball time that gets mis-sold* to advertisers, I'd say the FB user experience is very much "doing something wrong".
* dick pills and boob surgery, also government announcements for a country I don't live in, also offers to help renounce a citizenship I never had in the first place
I mean.. say its an enterprise mobile app. Maybe there are 2 shells, each shell has 5 tabs. Each tab might have 5 screens on it.. that's 50 files already just for the screens. Each screen might have various UI components or steppers, etc.
Most noobs, such as those who think 700 files is too many because they've only worked on apps they never published, might just cram everything into that one file.
However, there would be various files for components, functions, etc. Code that's single responsibility and easy to test might mean there are lots of files. There might be upload queues, offline functionality, custom code to go beyond what the ios/android SDKs offer, and so on. DTOs, DAOs, etc. various services..
You probably (won't) get the gist but yeah.
and a thousand lines of code? oh my? whatever shall we do?
[dead]
Is this related to that awful "ai interviewer" also named alex?
God, I sure hope not. That guy sucked!
This was such a fun read. Eels sound like they’re living in a sci-fi cycle: mud, rivers, Freud’s lab, then the Bermuda Triangle. Nature really doesn’t care about fitting into neat categories.
https://news.ycombinator.com/item?id=45115941