Monday, March 17, 2025

Leaked Apple Siri Meeting

Mark Gurman (Hacker News, MacRumors):

Robby Walker, who serves as a senior director at Apple, delivered the stark comments during an all-hands meeting for the Siri division, saying that the team was facing a bad period. Walker also said that it’s unclear when the enhancements will actually launch, according to people with knowledge of the matter, who asked not to be identified because the gathering was private.

[…]

But when Apple demonstrated the features at WWDC using a video mock-up, it only had a barely working prototype, Bloomberg has reported.

[…]

Walker defended his Siri group, telling them that they should be proud. Employees poured their “hearts and souls into this thing,” he said.

Chris Welch (Hacker News):

He also said it’s not a given that the missing Siri features will make it into iOS 19 this year; that’s the company’s current target, but “doesn’t mean that we’re shipping then,” he told employees.

Chance Miller:

As far as accountability is concerned, Walker told staffers that there is “intense personal accountability” shared by John Giannandrea, Apple’s Senior Vice President of Machine Learning and AI Strategy, and software boss Craig Federighi.

Bloomberg reports that Apple “doesn’t plan to immediately fire any top executives over the AI crisis.” It is, however, planning “management adjustments” in response to the problems, including “moving more senior executives under Giannandrea to assist with a turnaround effort.”

Joe Rossignol:

While it may sound obvious that at least some of the features are now functional within Apple, this was not entirely clear, as the company has not shown any public demos of the features being in a working state. Apple now faces the task of ensuring that the features not only work, but work well, before making them available to customers. Walker reportedly said the features were only working “up to two-thirds to 80% of the time.”

Benjamin Mayo:

Two thirds to eighty percent success rate is like half of the Siri feature set tho

Ryan Jones:

  • not held by JG, Tim, or Craig
  • AI only works 66-80% of the time (that’s VERY low, the last 20% is 80% of the work, even 95% is un-shippable) So it basically didn’t exist at WWDC. Ouch.
  • also blamed MarCom, fairly, for the tv ad

[…]

  • “As of Friday, Apple doesn’t plan to immediately fire any top executives over the AI crisis”. Notice the word immediately.
  • “It has discussed moving more senior executives under Giannandrea to assist with a turnaround effort.” That’s setup for “this was the plan, but we mutually decided new leadership would be beneficial”

Quinn Nelson:

This is wild. Every part of it. How none of the three heads responsible were there is unreal.

Jim Dalrymple:

Two of these need to be fired immediately. Doesn’t matter which two.

Maynard Handley:

Do we have a TECHNICAL analysis? Examples of TECHNICAL analysis include:

  • is the problem lousy models, or that Apple is trying to integrate models into APIs in ways never done before, or that the models are crippled by DNC levels of political correctness?

  • are the models limited by being forced to run on hardware that’s not powerful enough (because of local execution mandate?)

  • has a deliberate choice been made (if so why?) to not use open source even when that provides better models (eg the claim that Whisper provides much better text to speech than Apple Dictation)?

Ultraviolet:

Remember the system prompt where they asked it not to hallucinate? That was a massive red flag.

John Gruber (Mastodon):

The quotes are juicy AF, but don’t really start until the 10th paragraph. Like, for example, which Siri features does Walker think are “incredibly impressive”?

[…]

Walker compared the endeavor to an attempt to swim to Hawaii. “We swam hundreds of miles — we set a Guinness Book for World Records for swimming distance — but we still didn’t swim to Hawaii,” he said. “And we were being jumped on, not for the amazing swimming that we did, but the fact that we didn’t get to the destination.”

I’d say it’s a little more like selling customers tickets for a cruise that includes a stop in Hawaii, then never actually getting to Hawaii, and hoping they didn’t notice when the ship returns to port to disembark.

M.G. Siegler:

Robby Walker is not a name most people will know, but he has been at Apple for 11 and a half years. In fact, he came over when Apple acquired the startup he co-founded, Cue (originally named Greplin – one of the OG “social search” products) back in 2013. Why did Apple acquire that company? Because Google – specifically, the product known as ‘Google Now’ – was destroying Siri, as Matthew Panzarino noted at the time. Siri, of course, was still not good, way back then. And Walker has been working to try to fix that within Apple ever since.

[…]

This is seemingly a “well, we tried, but it didn’t work, so we’re back-burnering those ideas and moving forward”. It raises a question that sounds like a troll but is actually legitimate at this point: are we ever going to see these features? Is this truly the AirPower of Apple Intelligence?

Nick Heer:

Based on my experience in an Apple Store this week, this disappointment has not trickled down to retail employees. I was in for an appointment after I shattered my fifteen year record of keeping my screen intact, and I was told that even though my iPhone 15 Pro was “fine” because it supports Apple Intelligence, I could get nearly $700 towards a newer one after I had mine repaired. This was something I should at least consider, apparently. And, by the way, had I tried the new Apple Intelligence features?

John Gruber (Mastodon):

So let’s just examine how extraordinary and singular Gurman’s Friday report was.

Mark Gurman (tweet):

The company’s AI crisis will be the talk of its “Top 100” offsite meeting this week, and it’s planning some of the biggest iOS and macOS redesigns in its history.

Dave B.:

If Apple’s takeaway from @gruber’s piece was merely that they need to improve their AI, then it seems Apple missed the whole point.

Previously:

22 Comments RSS · Twitter · Mastodon


Mac Folklore Radio

Eh. Call me when there's a "leaked" crisis meeting over Apple's 2010-present user interface design and general software quality assurance practices. Where is the outrage, etcetcetc


@Mac Folklore Radio Totally agree, I don't care about their AI failures, because I never wanted these AI features in the first place. I want a well designed working operating system and ecosystem, like they had fifteen years ago.

The fact that they're missing how routinely frustrated every user gets with all of the broken and buggy features on *every single technology product* -- Apple and all the others -- means they have their heads stuffed far too deep into their asses to realize what's going on and what needs to be done.


I find the furor over this really puzzling. I use ChatGPT for work and hobby projects and I'm pleased with what it does. But I'm not seeing the earth shattering game changer that people seem to think Apple are missing.

I doubt Siri, or Google Asisistant, or the ChatGPT app are where LLM assisted things will take off. And I kind of doubt they'll ever take off tbh, but I feel like an idiot for not seeing it.

So much enthusiasm. Surely there must be a pony under all the garbage. (It's not the creative writing bot though. That's for sure)

I guess being completely oversaturated with the hype for Smart Gadgets, then voice assistants, then Blockchain, then NFTs, then Metaverse, then Web3, then VR has left me immune to the very idea that Silicon Valley will ever deliver something truly amazing again.


@Kristoffer: Yeah, I don't see Siri, Google Assistant, etc really needing LLMs per se. I mostly want Siri to be able to be more flexible and have even the tiniest bit of context. I've mostly given up on asking anything complex of our HomePods.

I have Home Assistant setup and have been playing with their voice assistant tech and it is pretty good. When you hook up an LLM though it feels like it gets extremely long winded when it shouldn't.


> this meeting reeks of you-all-deserve-participation-trophies-to-reward-your-hard-work-and-it’s-OK-to-feel-embarrassed vibes. What’s needed, quite obviously, is some “What is it supposed to do? / So why the fuck doesn’t it do that?” vibes.

The most important sentence Gruber has ever written. It's an apt description of the snowflake generation that is now at Apple (and other tech cos.), more interested in social virtue signaling "activism" than actually doing the work, blocking people on Twitter for even mentioning bugs in discussions about new features, etc. Cook's Apple is the opposite of Job's Apple; weak leadership, lack of any quality standards and greed; a lot of greed.


Allow my to wholeheartedly disagree. The fault lies entirely with management, no need to yell at the devs for this mismanagement.

Like I said somehwere else, it's obvious that someone thought that "It works 80% of the time" meant "We've only got 20%" left, and not "LLMs can only ever work about 80% of the time. We need to design around that."

It's like people yelling at the devs for not producing a Star Trek style replicator by extrapolating from a 3D printer.


"It's an apt description of the snowflake generation"

This idea that the solution to stupid management decisions has to be to scream at employees is hilarious to me. I'll happily accept the description of "snowflake" if the definition of the word is "not a completely delusional asshole who treats powerless employees like shit for my own mistakes."

GTFO with this bullshit.


Where did I say yell?
But management can and should expect high quality output. “Everyone gave it a good try” is why the software works is in the state that it has been for the last 15 years.


Yeah, gotta disagree on blaming the workers. I'm sure there are plenty of workers who have their head in the wrong place, and would rather signal some kind of virtue, be it for a social cause or -- more likely -- whatever vapid thing their management is insisting that they ought to be signalling (like pushing new flashy features that can be demoed), rather than doing good work. But there's no shortage of people who want to make good stuff too and take actual pride in their work. We hear about them a lot, and how their spirit gets ground down to a pulp after being in the corporate world for a decade.

The problem is management. The problem has always been management. The people who are in power and get to make the decisions are making stupid ones. The workers are terribly disempowered, and have to do what they're told. Because at the end of the day, modern corporations are basically human powered automatons whose sole purpose is to maximize shareholder value, and everything else takes a backseat to that. And like the proverbial monkey's paw or a classic rogue AI, it ends up producing shitty results because it's optimizing for the wrong thing.


"Where did I say yell?"

You references the Jobs incident where he summoned the Mobile Me team and angrily yelled and swore at them, telling them that they "should hate each other for having let each other down."

This is sociopathic behavior. Being an asshole is not a virtue, no matter how much assholes want to convince everybody else that it is.

In this particular case, it's also ironic, because Steve Jobs himself caused the Mobile Me problem. The team told management that it was not ready for release, and Steve pushed it out anyway. Then it failed, and then the team spent 24/7 effort to get it running. And then this piece of shit summoned them all and screamed at them.

This is not behavior anyone needs to emulate.

"Everyone gave it a good try is why the software works is in the state that it has been for the last 15 years."

No, it's not. It's really just not.


@Plume

> You references the Jobs incident where he summoned the Mobile Me team and angrily yelled and swore at them, telling them that they "should hate each other for having let each other down."

Which reference? I tend to be like Léo and wondering which reference you are pointing to. Is it In another blog post's comments? Cmd + E on Mobile and cmd + G in this page does not show any reference to MobileM prior to your comment.



I wonder if Grubers app hadn't failed so miserably if there had been someone there to yell at him to code harder every now and then.


"Which reference?"

This one to Jobs' vs Cook's leadership after failed launches:

"Cook's Apple is the opposite of Job's Apple; weak leadership"

This is what the whole thread is about, right? Jobs screamed at his employees when Mobile Me failed, while Robby Walker (and presumably Tim Apple) did not blame workers for management's failure.

Otherwise, what does "Jobs vs Cook" refer to?

"I wonder if Grubers app hadn't failed so miserably if there had been someone there to yell at him to code harder every now and then."

He could look in a mirror and scream at himself if he weren't such a snowflake.


Davin Heinimeir Hanson out himself as not having a clue of how an LLM works: https://world.hey.com/dhh/apple-needs-a-new-asshole-in-charge-0bf46b94

Yet another person who believes that if management had just yelled a bit more the 20% hallucination thing would have been gone by now.

Ultraviolet nailed it in the quote above "Remember the system prompt where they asked it not to hallucinate? That was a massive red flag."


"Hey, DHH, you dumb shit, you should hate yourself; I hope your dog falls out the window and drowns!"

Am I doing motivation right, you guys? Is he motivated yet?


While I don't think that yelling and cussing at people when they make mistakes is the correct answer, being unable to tell hard truths does a disservice to the people reporting to you as well.

I've had several experiences where my manager told me explicitly that I needed to do better, with specifics about what they wanted to see and where I wasn't measuring up. They didn't cuss me out or yell or anything like that, but they also didn't pull their punches. Those experiences helped me to improve and do better in my role more than anything else in my career.

There are many things at Apple that appear rotten from the outside and it seems like major change is needed. I don't know if the state of Siri is a symptom of their own problems or something caused by management elsewhere, and as an outsider I don't really care, I just want it fixed. If that takes a professional a-hole to come in and pop Apple's reality distortion bubble, so be it.

Telling people that everything is fine because they tried so hard is not a reflection of reality. You can give 100% effort and try as hard as you possibly can and it still not be enough. Maybe you aren't the right person and you need replaced. Maybe management has put constraints in place that prevent you from succeeding and you need to start rattling the cage until higher up management gets the message. Maybe management set unreasonable goals (with how little progress has been seen in Siri over the years, I'm not willing to put all the blame here. Any level of improvement would be nice). But regardless of the specific cause, sugar coating it and serving milk toast serves no one well.


gildarts Described what I was thinking. None of that requires yelling. Even if Jobs yelled, the results speak for themselves. Today’s Apple is a greedy behemoth (management’s fault) and a technical shitshow (a lot of which falls on the employees). But hey, at least they try. And protest evil Elon. Let’s give them a break, and a trophy.


"being unable to tell hard truths"

Who did that?

"Telling people that everything is fine"

Nobody did that. Walker said it was "ugly and embarrassing." He just didn't scream at employees and told them to hate each other, like Jobs would have done.

"Even if Jobs yelled, the results speak for themselves"

This is the exact problem with this argument: it's a type of cargo cult thinking. "Apple under Steve Jobs produced good results. Steve Jobs was an asshole. Therefore, if I am an asshole, I will produce good results."

Except that this is not how that works. Apple did not produce good results because Jobs was an asshole; Apple produced good results because Jobs had a clear vision for what he wanted Apple to do; he communicated it clearly, he permitted people to pursue his vision, and his vision was mostly correct.

None of that is the case here. The behavior shown in the privated Bella Ramsey video will never work for 99% of people with the technology available today or in the next three years. It's not a correct vision. It was only shown in the ad because it's the same kind of garbage all other companies are showing in their imaginary ads, and there's absolutely nothing anyone could have done to make this work.

The idea that you just need to be like Jobs and scream at the engineers and they will magically produce this absolutely impossible technology is absurd.

"Let’s give them a break, and a trophy."

What are you even talking about.


@Leo and gildarts

I failed to make my point clear enough. I agree that it's important to give constructive criticism. 100% with you on that.

But, the issue with the new smart Siri is that management told people to make something that just isn't possible. LLMs have been plagued by 20-15% hallucination rate ever since I started playing around with them. This rhymes well with the "it worked 80% of the time" thing that was mentioned somewhere.

So to ask someone to fix that is like telling them to turn the moon into cheese. It's completely out of their hands. They just can't do it. And it's not their fault, it's just not possible with LLMs to remove the hallucinations.

Someone in charge made the decision to gamble on Apple being able to bend reality and it turned out they couldn't.

Thats why I find Grubers thoughts on rolling ut some Jobs energy and unleash it on the developers is the dumbest shit I've read in a long while. Sure, Tim Apple should unload a steaming pile of venom on Gianandrea and whoever else that was in charge. But the guys that said "OK, we tried and failed" did the right call in not blaming the devs.


@Kristoffer: I'm not saying that they should be able to fix the hallucination problem and that being stern with them about it will benefit anything. I'm not asking for an LLM, I'm asking for any level of improvement at all.

Taking the team to task over the manys year of no progress is totally warranted. If it is the fault of the manager, address that. If it is the fault of the team, address that. If it goes all the way up to Tim Cook somehow, address that. But the many years of Siri having zero meaningful improvement can't continue.

If the managers have been reporting up the chain that everything is fine and dandy, they need to go as they either don't understand how bad things are or they are actively lying to cover their backsides.


"If the managers have been reporting up the chain that everything is fine and dandy"

...as probably happened under Steve Jobs with Mobile Me, because lots of people were afraid of him. I wonder why.

Leave a Comment