AI is a technology not a product
daringfireball.net
daringfireball.net
The ideal implementation of AI for Apple is probably to finally make Siri work. This isn’t necessary fancy, just let me set some calendar events without knowing the magic words or tell it to open Overcast and play the new Gastropod episode. Better yet, for power users, let me set up reusable shortcuts using natural language.
The most important part of this is it doesn’t necessarily feel like AI. The user does not like AI for its own sake or the weirdos who ramble about putting them into a permanent underclass. The user likes messaging their friends and playing music.
To much of this hype cycle has no user in mind.
This isnt unprecedented, its what happened in the dotcom bubble as well. But then that tech started getting used properly as well. So i think its a matter of time before claude code levels of value is avialable to normal users
Please elaborate
I want a "how do I?" function alongside search that will explain their product to me. Especially since so many SaaS products have absolutely terrible UX - it looks lovely, but you cannot discover anything, and you cannot intuit how to do something. Menus auto-hide, scroll bars don't work so you don't realise there's another half of the page you're looking at, buttons don't have tooltips or any explanation of what they do, icons are lovely but don't actually describe the thing they do, colours are lovely but I'm colourblind so aren't helpful, there's no useful help page for "this is how to do the really obvious thing you're trying to do...", or at least not one that I can find using the search terms that make sense to me.
I think I'm at 1 success out of 15 attempts for Gemini to explain how to do something in Google Sheets/Docs, though, so I'm not hopeful that anyone can actually implement this.
People love to talk about this as one of the helpful features of AI (knowledge extraction from documents/summarizing), but I'm really not convinced. The last generation of models seem to have 70-90% accuracy on tasks like this, which is way below what i'd consider a reliable tool
e.g. https://www.frontiersin.org/journals/digital-health/articles...
https://pmc.ncbi.nlm.nih.gov/articles/PMC11197181/
I don't know if there are any benchmarks for this sort of task, maybe the new ones are improved but I also doubt that people are using GPT5.5 pro ultrathink for these tasks anyways
Reverse dictionary
Stack Overflow clone, except you're guaranteed to get an unreliable answer promptly instead of waiting for a human to give it
OCR, with new and exciting failure modes
Machine translation, with new and exciting failure modes
Endless possibilities for exploiting the stupid and ignorant while destroying the web in the process
Note that only the first two are unalloyed good, and they can be done with embeddings without generative AI.
They lost the plot long ago. They're firmly in extraction mode now: how much value can they get from end-users?
Beyond that though is the dream that highly persuasive efforts will be effective at overcoming hesitation and converting it into new desires and preferences. Like the way it has worked under so many situations. But with survivor bias firmly in mind, those are the orgs where no miracle was actually required before it could lead to a windfall.
Wouldn't the simplest solution be to auction off Siri's back end the way Apple does Safari's search bar in iOS?
But this is contingent on the same services not being able to replace Siri and being able to reserve its APIs for Apple's exclusive use, and they have a pretty tenuous grasp on that these days.
https://9to5mac.com/2026/05/05/ios-27-will-let-you-choose-be...
https://blog.google/company-news/inside-google/company-annou...
I’m picturing a combination of on-board facilities and online services from the Apple cloud that Apple product holders could use to flag and filter LLM slop. As a value added prospect, iPhone users who read HN or used TikTok would be seeing clear UI-level indications of when they’re interacting with slop with options to kill it.
In my estimation it would provide platform benefits without losing capabilities, leverage Apples hardware and not advertising positioning, fix critical issues of spam and scams, and let them market a higher calibre of online experience. Also, they could un-eff Siri - “play album X starting at track Y”, come on, it’s 2026.
Isn’t this the proverbial ”faster horse”? Ie let me do exactly what I can do now, in a very slightly different, possibly very slightly more convenient way?
As UX / UI professional of 17 years I think design is a dying field the above would kill digital UI design quicker. Yet the UX would be less steps / friction to complete tasks which is the harbinger of UX design…less is more.
On a side note I’m just in medical school studying a mid level Concentration. I don’t foresee a LONG term future in digital design and development much anymore.
Less steps isn’t always better. Friction has its place.
A basic example is an “Are you sure?” confirmation before a destructive action.
I wish there wasn’t so much focus on “less clicks.” It’s often to the company’s benefit at the detriment of the user
If the user asks for a faster horse and you sell them a trebuchet, you lose, no matter how fast the trebuchet would technically get them to their destination.
And you may be able to sell them what they are already asking for a lot faster than what they are not.
Now if you are trying to sell them something that they would rather not even have at all, that's another story too.
(Arguably the car affords you better control than an unruly horse. Self-driving cars are moving us closer to the horse again. ;))
Maybe there is some parallel to the way that AI is moving "cutting edge" programming closer to the mainframe/dumb-terminal paradigm.
I've seen a horse get a black out drunk rider "safely" to a hammock and then continue to it's stable to rest
Agreed. But it is a good UI for some things, and which things is probably situation and user specific. Many people’s frustration with Siri is that many of the things it should be good at based on their decision to try, Siri cannot do.
Otherwise humans hate that interface.
It's crazy to me that even with a strong accent ChatGPT can nail my voice messages. If Siri can suddenly do that (and there's no reason it can't anymore) the device becomes much more useful to everyone that doesn't speak English and doesn't have an American accent.
Upon learning about LLM's however many years ago (3? 4?), literally my first thought was:
"Oh, how Siri is supposed to work."
It's the single most obvious application.
"Hey Siri turn on the livingroom lights" "Hey Siri set the thermostat to 19'
Being able to go "Hey Siri turn on the livingroom lights and set the temperature to 19" would be so much easier.
For a real AI this would be no issue. But Siri is completely hand scripted.
They literally have forgotten as they’re just following orders from investors (public or private).
Before those orders were “get as many users as possible” as valuation is based on credibility of reaching TAM.
Now the orders are “just use AI as this will be necessary to stay relevant and avoid being displaced” in addition to of course “profitability now”
"You have to work backwards from the customer experience."
AI was never going to be on Apple's roadmap in a significant way because it's in their DNA to differentiate technology from products.
I remember my first meeting I went to at another company that was just a guy talking with a PowerPoint. I couldn’t believe we didn’t have the data or time to ask probing questions. We’re just supposed to take this guy for his word? Crazy
- no PowerPoint
- 1-6 page write up of the problem, proposed solution and timeline, and alternate methods that were not chosen
- meeting participants ideally have already read the paper, but given 10-20 minutes in silence to read and mark up their thoughts.
- presenter says their piece, mostly just summarizing the paper and clarifying tricky sections
- intellectual bloodbath as all participants try to poke holes and see around corners not foreseen by the presenter
- follow up next week, until the group/manager is satisfied about the direction of the project
The what is the idea behind the "ideally have already read the paper, but given 10-20 minutes in silence" part?
The fact that people that have already read it have nothing to do and waste time sitting around bored sounds like an obvious flaw, are we missing something?
Also, 20 minutes of respite isn’t necessarily “waste.” Having 20 minutes of time to think deeply on something is often a gift!
I think this gets to the heart of my complaint. I thought this would increase the water level on our technical discussions. It hasn't. Like I say, could be the documents or people are poor. In any event, I'm one person that believed in this right up until my org implemented it.
I have no idea how your org implemented the table read, so I can’t comment on why it may not have been effective there.
It's the difference between peer review by leaders in the field trying to make your paper better, and juniors wanting to be heard or insecure academics trying to get an ego boost by nit picking and wasting time.
Intellectual bloodbath sounds like so much of the latter with point scoring being the goal.
Intellectual honesty, saying "I don’t know", for example, is only possible in low-politics environments. Otherwise, you make yourself vulnerable to the wolves.
Summarizing with AI isn’t usually a problem, but the objective of the narrative is to gain a deep and detailed understanding of the proposal or problem described within it. The reader or decision maker often can’t do their job well unless they read the whole thing. These narratives are often thoroughly marked up with commentary during the review, sometimes every paragraph.
SRI -> SiRI Inc.
[1] https://www.sri.com/75-years-of-innovation/75-years-of-innov...
answer... ditch phone/screen, just have an earpod you talk to.
(I don't know if they exist)
Sounds heinous, please never design the UX for a product I’ve got to use.
While I agree with the thesis, the response is total reality distortion field.
He says "you have to start with the customer experience" rather than the technology.
Then he name drops 4-5 technologies that were speculative endeavors and says when Apple put them all together to make the laser printer: "we can sell this".
Ok. Let's try that with some basic needs. And I'm totally serious. Let's go. I am abroad, walking in a city. I look for a book store. I get my Apple phone, open maps, OK, that works.
Now I have to go to the bathroom. Hmm, is there an app for that?
How do I convert this phone into a nice and clean toilet? Stupid question you say? I'm the user, remember, and I have __one__ need right now.
Wait, I'm supposed to use maps again to find a public toilet? Chances are it's going to be smelly and dirty. Not the great UX I am looking for, Apple.
Seriously, Apple has been addressing the wrong problems for far too long now. They are not looking from the user's perspective, but rather from the viewpoint of: we have a CPU and a touchscreen and a camera, what can we make with that so that more people will buy it? And how can we sell people even more stuff __through__ it?
But of what use is a better camera if the device can't even solve basic needs?
If you want to call yourself a revolutionary company, you gotta step back and think different.
"You have to work backwards from the customer experience."
To do this right, you probably need to learn from the many attempts others made before. I bet nobody knows yet what a good customer experience for AI will be. They are all still experimenting until somebody puts together all the parts in a successful package.
All the major AI companies are trying to manufacture their own ecosystems to become less disposable. They'll get away with it for a while, but only insofar as hardware prevents advanced use. Once we get that hardware[1] there will only be two types of AI companies: hardware manufacturers, and labs. Just like sync became trivial and ancillary, so will AI inference.
It may well be that the user interface of your "phone", and how you use it, changes over time as we progress toward AGI, but as long as Apple keep to the Job's aesthetic of making well designed products that get out of the way and just "do the thing", they should be fine. Of course Apple will eventually fall, as all companies do, but I don't think the reason for it will be that the "phone" market was rendered obsolete by AI.
Perhaps if phones becomes more of a "pocket assistant" than a device to run discrete apps, then they will becomes harder to differentiate based on software, and more of a generic item rather than a status/luxury one ... who knows? Anyone else have any theories of how Apple may eventually fall?
There is one potential AI risk to Apple, that they are at a disadvantage due to not having their own frontier models and datacenters to run them on, but I think there will always be someone willing to sell them API access, and they will adapt as needed. Good enough AI is only going to get cheaper to train and serve, and Apple not trying to compete in this area may well turn out to have been a great decision, just as Microsoft seem to be doing fine letting OpenAI take all the risk.
It's not going away in the next few years. Which means Apple doesn't have to rush to release an AI product for the sake of it à la Giannandrea.
Well before the iPhone flew off the shelf, using the the previously established smartphones I never had to settle for less than a week of battery life.
Plus anybody could just slap in another spare battery whenever they wanted to, whether they were off the grid for an extended period or not.
Never thought it was going to end, only get better not worse.
Given that humans sleep at night, recharging the phone at night is a reasonable price to pay for the benefit of a smartphone vs flip phone, but a device that needed charging during the day as well (e.g. due to a form factor with a tiny battery) would almost certainly be a product killer.
If the signal is clear – if you have observed the same person facing the same problem in the same workflow – then the AI feature deserves its place in the product by automating one step that they hate. The outcome does not necessarily need to be AI-powered. The user simply stops facing that problem anymore.
The Gruber's logic works on the level of the whole product. But there is also a diagnostic implication here – the louder the product sells its AI capabilities, the less the team understands what exactly the product does.
We are in the midst of a paradigm shift, and the perspective in the daring fireball post aligns exactly with this author’s perspective:
https://rebecca-powell.com/posts/return-on-intelligence-01-e...
They don't have a social network business because they tried that and failed. [1]
> By the end of this decade, it’s unlikely that people will swipe on their phones to tap on Uber or Lyft. They will just tell their always-on AI agent to get them home. Or that agent will have already figured out where they need to go, and the car will be waiting without the friction of a request. “There’s an app for that,” may be replaced by “Let the agent do that.”
Who TF are these people who think this kind of future is desirable? I basically think it's just people that want to broadcast that they're so important and busy that they can't take the 5 seconds it takes to hail an Uber. Its like all that "productivity optimization" porn that people spew online to show how focused they are.
I was reading article recently that said that a majority of people interviewed did not want to use AI agents simply because they didn't have much stuff in their life worth automating. Or more to the point, a lot of people actually enjoy making grocery lists, planning trips, picking out gifts for friends, etc. This stuff is generally considered "life", not some back breaking drudgery like washing clothes in a stream that I'd like to automate.
These folks like Levy who view this dystopian future as some sort of nirvana (and not because they view a different future, they actually want all this nonsense) can go F themselves. You can also tell how incredibly sheltered these people are because you can see they're rarely interacting with people outside their bubble. For example, a lot of people that open the Uber app make their decision based on data in the app, like "surge pricing, nevermind, I'll just walk" or "this looks expensive, let me try Lyft". You could argue an agent could learn all those rules, but again, these minutia of life are not exactly a nuisance to most people.
Some of this is weird techno delusion. Some of it is because the people describing it do a poor job of explaining how it might work.
If a couple decades ago someone told you that you’d have an always listening device in your pocket to answer your questions from all the world’s information, it would have sounded dumb, and with the always-listening device, rather dystopian. But that’s what you have assuming you have any modern smart phone.
The “agent knows where you’re going and calls a car for you” sounds dystopian as hell if done totally autonomously. But you could also imagine that an agent pops up a message on your watch “hey, you’ve been at dinner for an hour, if you’re winding down I can call you a car in 15 minutes” and suddenly it’s not that absurd.
That feels a little bit of muddling the waters. At least on Android with which I'm familiar, (a) you can turn off the "OK Google" detection in settings so that it's not always listening (and I'm not sure what the setup is now but originally I had to opt it to OK Google detection) and (b) the path for OK Google detection runs on a lower power, on device chip that only has capacity to store like the last few seconds of ambient noise to look for the assistant key phrase.
Just because it's a losing battle doesn't mean nobody cares.
It takes a lot more than 5 seconds to make an informed decision these days. Apps and websites are throwing abusive fine print and dark patterns at users left and right.
I'd be absolutely thrilled to e.g. not have to interact with the Uber app and all its dark patterns if there were somebody or something I could trust to competently represent my interests.
That said, that's a big if, i.e., whether commercial LLMs or agents will be able to do that, given the overwhelming pressure to just take money from both sides of the transaction and skew the decision.
But if it does happen, I actually see this as a huge potential factor strengthening smaller suppliers directly competing with large platforms. If my agent can independently figure out if a given supplier is trustworthy, whether their terms and conditions are reasonable etc., I'd be much more willing to engage with them outside of a large platform.
I just opened the Uber app. The first thing that pops up is a search bar that says "Where to?". I entered a destination address. Next thing it showed was a map with a path to my destination and nearby cars, and buttons where I can choose my type of ride (e.g. UberX, Premier, etc.) It defaulted to UberX, which was the cheapest option except for the "Wait and Save" option that was further down. I tapped the "Choose UberX" button and the ride was on its way.
So, OK, maybe it took literally 15 seconds. I'm not denying Uber may use dark patterns elsewhere, but from the end user experience of hailing a ride I don't see how it could be any simpler or more straightforward.
What if the agent can also communicate with the car's agent? They may even negotiate the meeting spot. Agent is superior.
This feels a lot like all those "where did the soda go" commercials (i.e. https://www.reddit.com/r/wheredidthesodago/ ), where some mundane task is imagined to be hopelessly complex with a bunch of possible what-ifs. For what it's worth, yes, I have taken Uber in all of those conditions, and no, I never found it difficult or had a problem with it.
If you don't even notice the dark patterns, they're working exactly as intended.
Are local taxi services cheaper and known to be more reliable? Am I missing an obvious public transit option? Is Uber pulling something creative with dynamic pricing again?
> I'd be absolutely thrilled to e.g. not have to interact with the Uber app and all its dark patterns if there were somebody or something I could trust to competently represent my interests.
Is there any reason to expect that commercial LLMs will avoid fine print, dark patterns, or acting against their users interests?
If AI allows more people to have such a premium experience, that's a use of technology that makes a lot more sense than all the "AI will take over your job" scaremongering.
They're in the habit of their admin telling them the Uber has arrived to take them to their 3 o'clock yes.
That feels both more credible and more desirable than the magic panopticon predicted in the quote, and doesn't really depend on any major technological leaps beyond continued maturation and scaling of Waymo/alternatives.
I imagine because a huge part of optimizing fleet availability and distribution is knowing where you want to go before deciding which vehicle you should travel in.
Having a city's worth of automated cars driving around all the time sounds like a hellscape.
Staying on hold for you. Google Lens on that coat or bag. Warning you in the middle of a text convo with a stranger, if the conversation veers into typical scam patterns. Better text/email spam detection than Apple. Hanging up spoofed calls posing as your bank. Magic Cue. Magic Eraser. Better transcriptions and translations, in far more languages.
And who could forget, a good touchscreen keyboard. Those are real "AI as a feature". Not a better Siri.
I agree with Gruber's take, if the seller is Apple.
I think this article is too soft a criticism by half. The iPhone defining the mobile era was not an artefact of the Apple logo being on it. Every bit of Apple's relentless productization went into what features the phone actually had and how they were integrated. This guy, in 2006, would have been telling Apple 'just release a feature phone like BlackBerry does, so you can define the era of feature phones like you did MP3 players'.
I would like to tell it to turn off certain light in a certain room, but unless I get the exact string name of those light correct when I speak, Siri doesn't know what I'm talking about. And it can't do multiple things in a command. I can't say "turn off all the lights in XYZ room" or turn of "this light and this light".
Meanwhile, I can vaguely tell a computer behind my tv to do very complicated things (build me an service that ...) and it can execute on it fairly well. But in apple's "product vision" which I am apparently too dumb to decide for myself what I want, I can't ask for two lights to be turned off at the same time.
Oh that’s silly thinking.
I already have Alexa and Hue lights. Only thing I use voice is „play music/stop music”.
Turning off all lights or on all lights sometimes. Turning on specific lights app. If I spend time to name lights specific names that are quick to pronounce maybe I would use it more.
Silly part is imagine trying to order Lyft on airport when everyone tries to do the same …
Today I wanted to book a public transport ticket in Germany but it was simply too hard to keep copy pasting screenshots from the app to ChatGPT. This seems to be a very easy problem to solve and standardise at the OS level but no one seems to want to do it.
I agree its not a totally different "product" but does require some thought. Apple can't sleep on this.
Yet you've only offered examples of what they _shouldn't_ do with "AI." You've offered no clear ideas on what they should do, only intimated that Apple, by pure osmotic magic, would be better at it than others if they made similar investments.
There's something about language models that causes smart people to wantonly turn their brains off.
Maybe Apple will buy Nvidia. That would be a useful vertical integration.
No chance. Currently Nvidia’s market cap is higher by about 1T.
Let's all pretend there was an intentional coherent strategy, and not because Apple's lagged its peers due to its secretive corporate DNA, internal silos, and restrictive publishing policies actively repelling AI talent at a critical time.
The more invisible AI inference becomes in systems, the more they start feeling more practical.
I personally find it more engaging to have an agent visualize things for me using matplotlib.
The problem is that too many startups are trying to do to OpenAI and Anthropic, what merchants do with commodities in the market.
Seems more driven by profit in mind than by actual value creation.
That’s the thing; the LLM itself - the chat window - can’t be the whole product for an industry. It’s a technology that you build things with.
"Steve Jobs handling a tough question at the 1997 Worldwide Developer Conference."
The final final form factor is probably a pair of glasses (or an implant), but I still think that's pretty far away. Before that can happen, we need computer chips and batteries to become almost microscopically small.
For the foreseeable future—still long term, but much closer than glasses—I think the logical form factor is a smartwatch. For photos, it would have an under-screen front-facing camera, and an outward facing camera on the wrist band. The screen would be a bit larger than today's largest Apple watches, and it would fold out like a folding phone when you need more space.
Even unfolded, the screen would have to be smaller than what we're currently used to on smartphones. However, this would be less important if most interaction was done via AI, just as limited-interaction iPods and Blackberries never commanded massive screens. People who want to watch movies, read longer books, or play games on larger screens could still carry folding tablets in their pockets on some occasions, but the watch would be the central device everyone always has.
Apple, of course, already makes smartwatches, arguably the best ones on the market. But an Apple Watch is very much not the device I'm describing, and I'm not sure if Apple will let it get there. Apple is stuck in the innovator's dilemma, where the iPhone prints so much money they can't afford to cannibalize it. For the moment, the iPhone has been so good that this hasn't caught up to them. I think—and for the sake of innovation, I hope—that this doesn't last forever.
Maybe for certain tasks or certain people. But in general i disagree with that take. The fact that i can stow away my phone into my pocket, not creep out bystanders (they know my camera isn’t recording them, etc), and forget about my phone for a while is a FEATURE.
To me there are cool things but nothing so great where if LLMs were deleted I’d cry about it. To contrast mRNA vaccines, gene therapy and crispr seem more impactful in reality, just to mention things from 2020.
Even considering that it’s sometimes wrong or hallucinating, it’s doing an important job by beginning to eliminate gate keeping, be it centered on cost or access.
Considering that ChatGPT has 900 million users I suspect the average person finds value in the technology.
Coding adjacent, but my small town's small businesses have all dramatically improved their websites with LLMs. Folks who didn't have them before can now build them. Folks who had to rely on a web designer no longer have to.
Yes. Code looks intimidating if you aren't used to it (and don't have an IDE). And there are lots of steps between having a file of code and having a hosted website.
* It's lovely to have the opportunity to disagree with both Gruber and the "the whole thing smacks of politics" HN commentariat, pulled daily between "it's just a tool, like a hammer, which also kills people, stay with me here" and "AI puts an expert in your pocket; soon, the expert will live in your eyes"
I do like certain other aspects of projects like that, which is why it’s great to have LLMs to collaborate with on it.
> they'll pay a monthly fee for a machine to tell them it's ok
Why would I outsource that decision to an LLM? I can look at the result myself and decide if I like it.
"Folks who didn't have them before can now build them. Folks who had to rely on a web designer no longer have to."
You ever heard of squarespace?
A funny story that happened the other day: A friend knew he had to be at dinner at a place across town but he forgot why he had to be at that dinner. While we were waiting for his rideshare to come, he was flipping through every kind of app trying to reconstruct the original context for his appointment.
In theory, this is where AI should shine. He should have been able to say "Hey Siri, pull up all of the info that references tonight's dinner appointment" and AI should be the unified interface into a bunch of app-specific data pools.
But of course he's never in 1 million years would have thought about using Siri to do that because of how bad Siri is.
Internet
- made the communication possible, all the information diffusing was only possible because of internet
- all sorts of small interactions and serendipitous communication through social media was due to the internet
- computation and simulation required was possible with the internet
Sometimes things make other things possible in subtle but real ways which are overdetermined. You can't articulate how AI will help a person materially in first order effects. But it will.
But you specified America, so I guess no.
Clifford Stoll had used the internet for two decades before writting his infamous 1995 essay in Time saying the internet was overhyped, and "normal" people would never e-mail because they can just fax.
CRISPR was first observed in 1987, and the gene-editing breakthrough came in 2012.
It's really, really unclear why you think LLMs would have faster adoption, when they are already that being adopted faster that anything other tech, ever.
Do you honestly know of no non-technical person who use LLMs? Because an absurd number of people report on surveys that they use it every week.
LLM apps are regularly the top downloads on the iPhone.
Um iMessage?
Quote tweeting a NYTimes post detailing war crimes "As Israeli forces entered Gaza on Friday to fight Hamas, phone and internet service was severed for 34 hours. Most people in Gaza had no way to reach the outside world..."
Gruber wrote "F*k around and find out."
Quote tweeting a post by the UN Human Rights account about Israel's flooding of tunnels with saltwater could have severe adverse human rights impacts,
Gruber wrote "One side is pumping salt water into the tunnels. The other side has put innocent civilian women and children hostages in the tunnels. Also: "salt water" has a space when used as a noun"
Quote tweeting a post by a StopAntisemitism page that posted about 'pro-Palesinian agitators showed up to secreteary of Defence Lloyd Austin's home..."
Gruber wrote "These people are surely a lot of fun at parties"
Gruber is a big fan of collective punishment, it seems. But at least he's very specific about the use of grammar.
Looking back, it just reads as sinophobic.