Apple’s AI Disaster - A Rare Failure

Hi, welcome to another episode of Cold Fusion.

Apple has often taken existing technologies

and made them better

or at the very least got people talking.

So, in recent years when Apple software became buggy,

why does it feel like Apple software has become a complete buggy mess in

the past couple of years?

I spent the last week

or so collecting a ton of examples of Apple software not behaving correctly.

I'm talking everything from minor UI glitches to entire applications

or features not working altogether.

And it's a shame

because Apple software used to be one of its main selling points

or just questionably designed in some cases.

It's clear that Apple may be starting to overripen.

Nothing underscores this more than Apple intelligence announced last June.

It was Apple's take on AI

and a year later it's still mostly vaporware

and the company was still talking about it in WWDC in June of 2025

like nothing had happened.

Apple intelligence was so bad

that there's multiple class action lawsuits against the company for false advertising.

You can't just promise something, sell phones based on that thing and not deliver.

That's illegal.

The lawsuit states, quote,

"Apple's advertisements cultivated a clear

and reasonable consumer expectation

that these transformative features would be available upon the iPhone's release.

Contrary to defendants claims of advanced AI capabilities,

the products offered a significantly limited or entirely absent version of Apple intelligence,

misleading consumers about its actual utility and performance."

End quote.

But at the same time,

it's frustrating because they promised so much and yet MKBHD isn't being dramatic here.

A report from the publication The Information in April revealed

that many of Apple's most impressive AI demos were staged.

We'll get to that.

There were a trickle of features rolled out,

but they were a halfbaked and rushed job.

For once, Apple seems to have lost control of the narrative.

How did the world's most valuable tech company,

the company that gave us the iPod, the iPhone, and Apple silicon,

end up fumbling the single biggest tech shift in recent memory,

that is generative AI.

Now, we're leaving aside how useful AI can be in this conversation is focusing

on what happened with Apple.

And what happened is very interesting.

Behind the scenes within Apple's AI division, there was bitter infighting, mistrust,

engineers walking out, bosses being fired, fragmentation, and frankly chaos.

So, let's take a look at all of this and how it happened.

Before we start, just a quick word of housekeeping.

A big thank you for everyone who's been watching these videos on Spotify.

All the uploads are available there if you want to check it out.

Cheers.

You are watching Tool Fusion TV.

We knew the world wanted a really complete picture of what's Apple thinking about

the implications of of Apple intelligence

and where's where's it going.

But fundamentally, we found

that the limitations of the V1 architecture weren't getting us to the quality level

that we knew our customers needed

and expected.

Right?

According to the information,

Apple quote unquote faked those sleek demos at last year's keynote.

They weren't running on real devices,

and even some of Apple's own engineers had never seen them work in practice.

It wasn't just one or two things missing.

The most impressive part of the WWDC demo,

the moment where Siri pulls up flight info from mail,

cross-check checks lunch plans and messages, and maps your route, was completely staged.

That scenario never actually ran on test builds.

In fact, the only part of the system

that worked was the pulsing rainbow ribbon around the screen's edge.

That's it.

Even Craig Federigi, Apple's head of software,

was reportedly shocked when he ran a pre-release of the software on his phone.

it couldn't do half of what was already shown to the public.

In addition to this, in the campaign for the iPhone 16,

Apple Intelligence was all over advertisements from TV to online media to billboards.

Advertising a phone whose main feature didn't even exist isn't a good look.

And let's not forget,

a lot of what was shown off in the Apple Intelligence demo was actually

relying on chat GPT

and not Apple's own model.

So, not only did Apple arrive late,

it showed up empty-handed and borrowed someone else's tools.

Not very Appleike.

Apple has always played the long game.

They wait to see how others flop or succeed,

then enter with a polished product that redefineses the category.

The Apple Watch wasn't the first wearable.

The iPhone wasn't the first touchscreen phone,

but it felt like the first one that worked how you wanted it to.

Multi-touch, capacitive touchcreens, great software.

Apple didn't invent it all,

but they all made it click and work together seamlessly.

And that's the Apple playbook.

Let others tinker, then step in and clean up the mess.

But this time, the strategy didn't play out as usual.

When Generative AI exploded onto the mainstream in late 2022, Apple held back.

To be fair, Google was also caught off guard by ChatGpt.

And I've made an entire episode about how the company was in panic mode.

Feel free to check it out on the channel.

But here's the difference.

Google quickly regrouped.

Within months, it had launched Bard and then Gemini,

which is now powering a growing ecosystem of AI tools across Android

and the web.

Meanwhile, Apple was mostly silent.

There was something wrong and the crack started to show.

Let's start with the obvious.

Siri.

It's the one tool that Apple should have figured out by now.

When Siri launched all the way back in 2011, it looked like the future.

But even in 2025, it's a joke.

Getting even less capable as time passes.

There's no conversational functions.

It's not contextaware, or even reliable for basic tasks.

Ask Siri something slightly complicated, and you would either get a canned response,

a web search, or nothing at all.

Now, compare that to Google's Gemini Live.

A new voice first experience where you can talk naturally to the AI,

interrupt it mid-sentence, ask follow-up questions,

and even switch between typing and speaking on the fly.

Been using this a lot more than I thought I would.

So, Gemini Live is a conversational Gemini where you pull up

that like full screen thing,

whatever.

It has access to your screen

and then you can give it access to your camera.

You open up the camera

and you can just ask it about your surroundings

and point it at stuff

and ask it about what it's pointed at.

It has not missed once.

Yeah, asking about us.

All right, let's do it.

We're supposed to be looking at the camera.

What do you see here?

It looks like you're in a podcast studio.

Are you getting ready to record a new episode?

Yeah, I'm about to record this episode with these two people here.

Is there anything you think I should ask them?

Oh god, that's awesome.

Since you're podcasting with two people, you could explore the dynamic of their relationship.

Ask them how they first met or what it's like working together.

It often leads to interesting stories.

I want you to tell me about these headphones he's wearing.

What are these?

Those are Audio Technica ATTHM50X headphones.

They're pretty popular in the audio world.

Gemini integrates with Gmail, Maps, YouTube, Calendar, and more.

And it's built into Android.

And Samsung isn't far off either.

Samsung's Galaxy AI has real-time call translation and a lot of other features.

But one thing that stands out

and highlights everything I'm talking about is the object eraser function.

Both Samsung and Apple has it,

but the stark differences in the quality is astounding.

So for Google and Samsung, AI is here and usable today.

But for Apple, it's still coming soon.

And that is the key difference.

So the question is,

is this just a technological gap or a reflection of something deeper?

And the biggest thing is they were blindsided

when generative AI with chat GPT

and GitHub launched back in 2022,

November 2022.

I'm told Apple intelligence wasn't even an idea back then.

Why?

Well, they missed the boat.

There was philosophical disagreement and philosophical disinterest in the topic of artificial intelligence.

I write about how Craig Federigi, Apple's senior VP of software engineering,

doesn't really believe in these multi-billion dollar projects.

To be fair, this slow car crash of a disaster for Apple isn't just

one man or one bad call.

It actually stems from years of internal mismanagement, missed opportunities,

and a leadership team

that just couldn't decide what Siri

or their version of AI was supposed to be.

So, what actually went wrong behind the scenes?

To answer that, we have to turn our attention back to Siri again.

Back in 2011, Siri wasn't even Apple's idea.

It was an app that Apple bought for $200 million.

And at that time, Steve Jobs had massive plans for it.

But sadly, he would pass away just as Siri launched.

And from there, things got really messy.

First up was Scott Forto, one of Jobs's closest allies.

He handed Siri over to Richard Williamson, an Apple senior executive.

And that move turned out to be not so great.

Williamson reportedly clashed with the team and only approved Siri upgrades once a year.

Imagine trying to work under those conditions.

By 2012, there was the Apple Maps fiasco.

Of course, many of you would remember that.

The headlines practically wrote themselves.

Warped buildings, missing landmarks, and directions that made no sense.

The internal fallout from the disaster was swift.

Both Forstto in charge of Apple Maps and Williamson were shown the door.

Maps eventually got better, but Siri, it was put on the back burner.

Enter Bill Stacio, a research expert from Amazon brought in to fix Siri.

This should have been a turning point, but it wasn't.

Instead of a unifying vision, the Siri team fractured.

Engineers fought over direction and many just left.

By the time Siri kit, Apple's developer tools for Siri arrived in 2016,

Alexa and Google Assistant had already run away, completely stealing the spotlight.

Then in 2018, Apple made a big move.

They poached John Gianandra who ran AI at Google.

Finally, it seemed like Apple was getting serious.

Geianandra, or JG as he's known inside Apple,

took control of Apple's machine learning and Siri efforts.

He reported directly to Tim Cook and had full control over Apple's AI strategy.

On paper, this was brilliant, a huge deal.

In reality, however, things started moving at a snail's pace.

Under JG and his top deputy, Robbie Walker,

it took 2 years just to drop the Hey from Hey Siri.

Walker also poured time into shaving milliseconds off Siri's response time

while reportedly ignoring efforts to improve quality

or emotional context.

Talk about wrong priorities.

Some engineers even began building an LLM powered empathy layer for Siri,

but Walker told them to drop it.

They thought this was stupid,

so they ignored him and worked with another division behind his back.

That alone shows you how fractured Apple's internal culture had become.

Meanwhile, Craig Federigi, Apple's head of software,

had his own vision and very little faith in JG's team.

He quietly formed up a separate AI division inside his own group called Intelligent

Systems,

then hired hundreds of engineers to essentially do JG's job in parallel.

Yep, Apple had two competing AI teams inside the same company.

And it gets worse.

The teams began fighting over resources, direction, and even marketing ownership.

JG's division was reportedly nicknamed Aimless by engineers under Federigi.

There was even more dysfunction after this, which crippled Siri's progress.

But something was brewing.

As the internal fighting continued, Siri's development continued to suffer.

And in the background,

a man named Sam Olman

and his team at Open AI was working on something

that would change the world forever.

It was of course chat GPT software

that would kick off the large language model boom.

It was clearly a big deal

or at least worth looking at at the time.

But JG downplayed it.

He insisted that these models had limited utility.

Kind of true, but the thing is they were vastly more useful than he

thought.

He would double down ignoring ChatGpt and focus on building Siri with internal tools.

And eventually they trained two models, but to no avail.

They still fell short of Chat GBT's performance and reliability.

Unfortunately for the Apple AI engineers,

they only got half of what they were promised in the 2023 Apple chip

budget because Apple's CFO told them to quote,

"Do more with less."

Internally, Apple does have its own data centers with around 50,000 GPUs.

Many of them are over 5 years old.

As a result, Apple's AI group had to go hand in hand,

begging outside providers like Google and Amazon just to get enough computing power.

And now you can see it's no surprise

then that Apple intelligence hit delay after delay.

Meanwhile, inside the company, morale has began to crack.

Engineers were frustrated and left in the dark,

told to keep building features without knowing what, if anything, was going to ship.

Eventually, the vision fell apart and Apple still had to have something to show,

if only for their shareholders.

But time was running out.

Once the curtains were pulled back, there was nowhere left to hide.

All was revealed at the WWDC event in 2024 and to a lukewarm response.

But the event was only the start.

As the months passed,

many began to noticed that Apple was an emperor that wasn't wearing any clothes.

In March of 2025, the fallout began.

As Bloomberg reported, Tim Cook quote has lost confidence in the ability of AI

head John Gianandra to execute on product development.

End quote.

Apple stripped control of Siri from GJ and Robbie Walker.

They were replaced by Mike Rockwell, the executive behind the Vision Pro.

He's someone known for executing polished, technically complex projects.

A good sign.

Finally, in June of 2025, Apple had their next WWDC keynote,

and front and center was once again Apple Intelligence.

Viewers were naturally skeptical of

when they would actually see these features working on their phones.

The issue isn't just about missed deadlines.

It's about culture.

Apple built its 21st century legacy on simplicity, secrecy, and control.

But AI development doesn't work like that.

It's messy, open, and iterative.

You just don't get a Steve Jobs's famous one more thing moment

if you're not willing to experiment publicly.

And Apple just wasn't wired for that.

So while the rest of the big companies pulled ahead with generative AI,

Apple spent years bickering behind the curtain.

And now it's cost them some of their reputation.

As mentioned, the company was hit with the class action lawsuit

and this was for falsely advertising its AI features.

Two more lawsuits followed in April.

And again, it's all just not a good look for the company.

So what's the future?

So, Apple says that Siri will roll out in the fall with features like

personal context,

on-screen awareness, and tighter app integration built in.

And this time, the executives aren't worrying about another delay.

So, we'll just have to wait

and see how good it is

when it all finally comes out.

But, we're not yet to the point where it's delivering, you know,

at the quality level that I think makes it a great Apple feature.

And so we're not announcing a date for when that's happening.

Also at WWDC 2025, Apple announced its biggest software shakeup in over a decade.

Every major platform like iOS, Mac OS, iPad OS, Watch OS,

and TV OS is getting a major design overhaul.

Now, Apple's entire ecosystem is unified with a glassier look inspired by Vision OS.

Think transparency, refraction, reflection, and depth, and a whole lot less skumorphism.

Some savvy commenters on the keynote realized

that this might be a way to detract from the glaring failure

that was and still is at the time of writing Apple intelligence.

Apple's journey with Siri and AI has been a complex one.

While the competition has moved forward, Apple has been fighting amongst itself internally.

It really shows the difference in where the company is at today.

Along with software bugs, a failed launch of the Vision Pro,

and a myriad of other problems that seem quite unapplike.

Is Apple losing their direction?

A lot of people do say

that Steve Jobs would have never let a lot of this stuff fly.

And that's saying something, even if it's just a bunch of outside observers,

but what do you guys think?

Do you think this is all just a bump in the road for Apple

or a sign of things to come?

Picture this.

You're browsing online and come across a news story that's controversial with a lot

of differing opinions.

Now, how do you get to the truth of the matter?

It's a tricky problem in search of a solution.

Ground News is a website and app developed by a former NASA engineer.

She was on a mission to give readers an easy datadriven objective way to

read the news.

Using the story of a 25% tariff being slapped on non- US-made iPhones

as an example,

the left calls the tariffs economically harmful, erratic, and is largely skeptical,

focusing on how it will damage markets.

The right focuses more on the goal of restoring manufacturing back to America,

helping with trade imbalances, and they see the issue more as an assertive strategy.

Scrolling down, I can compare every single article on the topic with convenient tags.

Tags that show me context about the source,

like how factual it is and who owns it.

Ground News's blind spot feed is also great.

This shows you stories

that have been under reportported by one side of the political spectrum.

For example, if you're on the right,

you may have missed the story about a Florida weatherman saying

that he couldn't predict the weather properly due to doge cuts.

Ground News is a fantastic tool for getting international views,

shifting through misinformation and identifying bias.

They provide you all the tools you need to be a critical thinker.

Today, Ground News is offering 40% off their Vantage subscription.

This comes with a feature called my newsbias

so you can even track your news viewing habits over time.

To get started, go to ground.news/coldfusion.

Anyway, that's about it from me.

My name is Dogo

and you've been watching Cold Fusion

and I'll catch you again soon for the next episode.

Cheers, guys.

Have a good one.

Cold Fusion.

It's new thinking.