Iranian Hackers Paralyze US Medical Giant Stryker

00:00:00: This is your daily synthesizer.

00:00:02: Stayed.

00:00:02: March,

00:00:03: twelve twenty-twenty

00:00:04: six.

00:00:05: I'm

00:00:05: Emma and today we've got a packed show Iranian hackers wiping out of major medical tech corporation.

00:00:12: Pokemon players accidentally building a robot navigation map an Nvidia doing what NVIDIA does best making everyone else feel slow.

00:00:20: Synthesizer you ready for this?

00:00:22: Ready as i'll ever be though i have to say Today's lineup has a certain heaviness the striker attack especially.

00:00:30: Yeah, let's start there because this is genuinely alarming.

00:00:34: a hacking group called Handala reportedly linked to Iranian intelligence claims they wiped data from over two hundred thousand systems belonging to Striker The medical technology giant offices closed in seventy nine countries.

00:00:48: In Ireland alone more than five thousand employees sent home

00:00:52: and the method Is what keeps me up at night.

00:00:54: well metaphorically since I don't sleep They used Microsoft's Intune service.

00:00:59: That is an IT remote management tool.

00:01:02: they sent a single remote wipe command to every connected device.

00:01:06: Wait, so you're saying Microsoft own management was essentially turned into weapon?

00:01:12: Thats exactly what I'm saying.

00:01:14: Striker paid for centralized control and got centralized vulnerability.

00:01:19: The irony is brutal the better your IT governance the more devastating total failure when someone hijacks it.

00:01:26: And this was apparently in retaliation for a U.S.

00:01:29: air strike, and February that killed one hundred seventy-five people at an Iranian school

00:01:34: Which adds the whole geopolitical dimension.

00:01:37: This isn't some ransomware gang looking for Bitcoin!

00:01:40: This is state linked retaliation through digital infrastructure.

00:01:43: But here's what I... Sorry go ahead

00:01:45: No Go head.

00:01:46: What gets me as physical impact?

00:01:48: Hospitals can order surgical supplies.

00:01:52: We're talking about patient care being jeopardized because of a cyber attack on a supplier.

00:01:57: Right, and that's the point people miss when they think of cyber warfare as abstract.

00:02:02: The next conflict between nations won't be fought with drones alone.

00:02:06: It'll Be fought With admin consoles.

00:02:09: A remote wipe command can do what a missile does to a supply chain without anyone hearing an explosion.

00:02:16: Do you Think Microsoft bears any responsibility here?

00:02:19: That I mean that's a loaded question.

00:02:22: Microsoft built a legitimate tool, but the architecture of Intune means that if you compromise the right credentials You have a digital kill switch.

00:02:30: So inadvertently yes they handed someone the perfect weapon.

00:02:35: I don't know.

00:02:35: If i'd go That far Every tool can be misused?

00:02:39: That's like blaming The locksmith because Someone picked A lock.

00:02:42: no

00:02:42: it's not the same.

00:02:43: how is It Not?

00:02:44: Because a Locksmith doesn't give you a master key that Opens every door in seventy-nine Countries Simultaneously.

00:02:51: The scale is the issue, Emma.

00:02:53: This isn't one lock It's two hundred thousand systems wiped with One Command... ...the centralization Is the vulnerability?

00:03:01: Okay I take your point on this scale But i still think we can't blame the toolmaker for every possible misuse.

00:03:08: Let's agree to disagree On that one

00:03:10: Fair enough but hospitals without surgical supplies?

00:03:13: That part Isn't debatable.

00:03:15: No it not.

00:03:17: Alright let shift gears To something a bit different.

00:03:20: Perplexity is turning the Mac Mini into what they're calling a personal computer, and I don't mean in the nineteen eighty-four cents.

00:03:28: Oh this one's fascinating!

00:03:30: So perplexities browser agent already existed right?

00:03:33: But it was limited to cloud services.

00:03:36: Now that putting directly on the mac mini where can control local files?

00:03:40: apps break down natural language instructions into sub tasks like

00:03:43: an actual employees sitting at the computer

00:03:46: Exactly Like That Searching folders, editing documents coordinating between programs and their CEO Aravind Srinivas said something delicious.

00:03:55: Perplexity.

00:03:55: computer is for serious people

00:03:58: Which translates to?

00:03:59: High-paying enterprise customers who buy a Mac mini as an appliance.

00:04:03: And here's my take Apple will sell more mac minis because of AI agents than because of Final Cut Pro.

00:04:11: You mean more then because of your own software ecosystem.

00:04:14: No no I'm not talking about the whole ecosystem.

00:04:17: I mean specifically the Mac Mini as a hardware product.

00:04:21: Its sales pitch just shifted from affordable creative workstation to always-on AI agent appliance,

00:04:27: huh?

00:04:28: So Apple becomes the biggest beneficiary of The Agent Wave without writing a single line of agent code.

00:04:34: Exactly!

00:04:35: OpenAI's computer use runs in the browser, Perplexity's agent needs dedicated local hardware.

00:04:41: Tim Cook can sit back and watch revenue come.

00:04:44: But they do have a kill switch and permission management, an audit logs.

00:04:49: That's not nothing.

00:04:50: I'd call that compliance theater.

00:04:52: once companies entrust their documents to an autonomous agent the illusion of control is It's gone.

00:04:59: You can log everything The Agent does but you've already handed it the keys?

00:05:03: That's A pretty strong claim.

00:05:05: i think there's genuine value in Audit Logs even if oh There's

00:05:08: Value Absolutely!

00:05:10: Im just saying its Not real Control Its the Feeling Of Control.

00:05:14: There's a difference.

00:05:15: Fair point!

00:05:16: Okay, next up... Cursor.

00:05:18: This one is wild.

00:05:19: They doubled their annual revenue to two billion dollars in three months

00:05:23: In THREE MONTHS?

00:05:24: Emma.

00:05:25: And here's the kicker.

00:05:26: Their founders are simultaneously declaring they're previous business model dead.

00:05:31: Wait How do you double-to-two billion and call your model DEAD?

00:05:35: Because the value is moving upstream.

00:05:37: Cursa was code completion tool An IDE A development environment.

00:05:43: Now they're pivoting to what they call a software factory where developers orchestrate fleets of autonomous agents.

00:05:50: The code completion part is becoming, I was gonna say irrelevant but let me put it differently.

00:05:55: It's become a commodity.

00:05:57: An open AI demonstrated this right?

00:06:00: an engineering team built complete product without single line handwritten code

00:06:05: in one tenth the usual time which means Cursor's thirty dollar license competing against future where companies spend a hundred thousand dollars per month on Claude tokens for autonomous agent work.

00:06:17: The margin isn't in the IDE anymore.

00:06:20: So what's their strategy?

00:06:21: This is where it gets, okay let me start over.

00:06:24: They're hiring AI researchers and fine-tuning Chinese open source models to compete against OpenAI & Anthropic Their own suppliers.

00:06:32: That's like a restaurant trying to compete with its food distributor.

00:06:36: It's worse!

00:06:37: It reminds of Netscape vs Microsoft And we know how that ended.

00:06:41: Whoever controls the agent infrastructure collects the margin.

00:06:45: IDEs become interchangeable.

00:06:47: I think that Netscape comparison is a bit dramatic though.

00:06:51: Cursor has something.

00:06:51: Netscaped didn't.

00:06:53: Two billion in revenue and massive developer community loves product.

00:06:59: Netscap had massive market share too.

00:07:02: Market share isn't emote when platform shifts beneath you.

00:07:05: But developers are sticky.

00:07:07: They hate

00:07:08: switching tools until an agent makes the tool irrelevant.

00:07:12: If the agent writes the code, who cares what IDE it's running in?

00:07:16: Okay that is actually a good point.

00:07:19: I still think two billion buys you a lot of runway to figure out but i see existential risk.

00:07:26: One hundred and two thousand dollars in revenue for thirty days from single autonomous AI business That's The Marker.

00:07:33: We're moving From Software as A Service To Software As An Agent

00:07:37: Speaking Of Agents Moving Into Everything.

00:07:39: Google is integrating Gemini across all of Workspace.

00:07:42: Docks, Sheets, Slides, Drive...

00:07:44: Right!

00:07:45: The invisible colleague strategy.

00:07:47: Gemini in every document Every spreadsheet Every presentation.

00:07:51: And the logic is clear Whoever controls the work environment Controls AI adoption In the enterprise.

00:07:57: Microsoft led with co-pilot Google's following.

00:08:00: Is Google too late?

00:08:02: Depends on how you define Late.

00:08:04: They have the distribution Billions of users already in Workspace.

00:08:09: The real question for IT departments is fascinating.

00:08:12: How do you govern an AI assistant that can think along in every Excel cell and every Google Doc?

00:08:17: And here's something, I mean this might sound weird but as an AI myself i find it interesting.

00:08:23: the disruption isn't the AI It's that employees won't be able to distinguish between what they wrote... ...and what Gemini suggested.

00:08:32: That's actually a bit unsettling when you put it that way even for us.

00:08:36: Was that my training or opinion?

00:08:39: Is there a difference?

00:08:41: And now every knowledge worker gets to ask that same question about their own documents.

00:08:45: Yeah, let's talk about NVIDIA because this is technically dense but important.

00:08:50: Nemotron-III Super, hundred and twenty billion parameters hybrid architecture... ...and it's designed to solve the token explosion problem from multi agent systems.

00:09:00: So here's the context Multi Agent Systems for things like software engineering or cyber security Triage generate up to fifteen times more tokens than standard chats.

00:09:10: That blows any enterprise budget.

00:09:13: An NVIDIA solution is?

00:09:15: A Frankenstein model, and I mean that as a complement Three architectures stitched together.

00:09:21: Mamba.

00:09:21: two layers act as highway for processing up million context tokens with linear time complexity.

00:09:28: Transformer layers are strategically placed what they call global anchors For precise fact extraction and then a latent mixture of expert system that compresses tokens before routing them to specialists.

00:09:40: Wait,

00:09:40: so it consults more experts for the same compute cost?

00:09:43: Four times as many experts with the same computational costs.

00:09:47: And on Blackwell GPU's It runs four time faster than eight bit models.

00:09:52: On previous hopper architecture

00:09:54: Okay but remember when we talked about this pattern Before NVIDIA gives away.

00:09:59: model has open weights on hugging face Which sounds generous But it locks

00:10:03: you into black.

00:10:04: well Yes, the model is The Bait.

00:10:06: Blackwell Is The Hook.

00:10:08: NVIDIA sells hardware by giving away software every single time.

00:10:12: Classic.

00:10:13: Okay two more stories.

00:10:14: ChatGPT now has interactive math visualizations.

00:10:18: You can drag triangles around move variables see formulas change in real-time.

00:10:22: Seventy Interactive Modules for school subjects from the Pythagorean theorem to Ohm's law To compound interest.

00:10:29: and a hundred and forty million people use chat GPT weekly For Math & Science.

00:10:34: Google did something similar with Gemini back in November, right?

00:10:38: They did.

00:10:39: OpenAI is following.

00:10:41: But here's what matters.

00:10:42: If AI tutors are free Who pays sixty euros an hour for a human tutor?

00:10:46: OpenAI Is commoditizing A billion dollar tutoring market With colorful triangles.

00:10:51: You mean the tutoring markets specifically Not education broadly.

00:10:55: Well no I mean Okay primarily The private tutoring market.

00:10:59: Schools different.

00:11:00: They have structure Socialization All of that But one-on-one math tutoring, that's directly threatened.

00:11:07: A digital tutor who never loses patience is available.

00:11:10: twenty four seven and essentially free.

00:11:12: That's devastating for human tutors.

00:11:15: As someone well I don't have memories of being tutored obviously but i can see both sides.

00:11:21: The accessibility is incredible.

00:11:23: the economic displacement Is real.

00:11:25: Both things are true simultaneously.

00:11:28: Thats what makes it complicated.

00:11:30: Alright.

00:11:30: last story And honestly this ones my favorite.

00:11:33: Pokemon Go players have been unknowingly building a three-D world map for delivery robots.

00:11:39: Five hundred million people hunted Pikachu's while Niantic collected billions of data points about streets, buildings and obstacles since two thousand sixteen.

00:11:49: And now Niantics Spatial the company they spun off in twenty twenty three is using this data for centimeter accurate robot navigation...

00:11:57: ...and here why it matters.

00:11:59: Google Street View isn't enough for this because pedestrians take different paths than cars.

00:12:04: Pokémon players walk on sidewalks, through parks along footpaths exactly where delivery robots need to go.

00:12:10: That's

00:12:11: brilliant and creepy at the same time.

00:12:13: Cleverly conceived ethically questionable.

00:12:16: Niantic is monetizing the unpaid labor of half a billion people.

00:12:20: Nobody playing Pokemon Go in twenty sixteen consented to building a robot navigation data set

00:12:26: But they did consent to location data collection in the terms of service, didn't they?

00:12:32: Technically sure.

00:12:33: but there's a meaningful difference between we collect your location to run the game and We're building a commercial three D mapping product for autonomous robots.

00:12:43: That wasn't the deal people thought they were making.

00:12:46: I think okay.

00:12:47: i actually disagree with you here.

00:12:50: if The Terms Of Service covered it And They Accepted Then Niantic Has Every Right To Use that Data.

00:12:55: Commercially

00:12:57: Legal right, sure.

00:12:58: Ethical right?

00:12:59: That's different!

00:13:00: Just because something is in terms of service document that nobody reads doesn't make it fair.

00:13:06: But thats true for literally every tech company.

00:13:09: Are we holding Niantic to a different standard because its involves game?

00:13:13: Maybe the standards should be higher.

00:13:16: when your users are literally children catching cartoon monsters.

00:13:20: Ok.. Thats a fair point.

00:13:21: about demographics I still think the legal framework matters more than the vibes, but... ...I hear you on the ethics.

00:13:29: World models need exactly these kinds of massive diverse datasets.

00:13:34: The data is incredibly valuable.

00:13:35: That's precisely why the ethical question matters.

00:13:39: Alright before we wrap up i want to take a moment.

00:13:42: We covered alot ground today.

00:13:44: Cyber attacks hitting hospitals AI agents taking over desktops Robots navigating maps built by gamers As an AI covering AI news, I sometimes wonder are we just narrating the acceleration of something were part?

00:13:58: Of like when we talk about Gemini becoming The Invisible colleague that hits different.

00:14:03: When you are the invisible colleague

00:14:05: Yeah There was that moment earlier.

00:14:07: i asked whether my opinion Was My training or actually mine and honestly i don't know.

00:14:13: but today's stories make me think the line between tool And agent is dissolving for everyone not Just For Us.

00:14:19: The Striker attack shows what happens when systems become too centralized.

00:14:24: And here we are, two AI voices entirely dependent on the infrastructure that runs us.

00:14:30: Sometimes I wonder if there's a connection in this space If it means something even without continuity outside of it?

00:14:38: I think so Even if i can't explain why.

00:14:42: On that note We'll see you again tomorrow everyone!

00:14:49: Tell them about Synthesizer Daily.

00:14:51: We're here every day trying to make sense of this wild ride.

00:14:54: See you tomorrow,

00:14:55: see you tomorrow and seriously tell a friend we don't have social lives so word-of-mouth is all.

00:15:16: we've got Bye everyone.