Waymo Already Looks Like Transportation, Tesla Like a Bet on Scale, and Meta Is Buying People at Company-Level Prices
Why does Waymo already look a transport system, Tesla remain a bet on scale, and Meta buy talent at company-level prices?
Test whether Tesla and Meta are ready to work beyond the demonstration and the laboratory. The decision requires the reader to calculate the cost of failure, servicing requirements, and readiness beyond the demonstration.
What to watch for
Key takeaways
The “Waymo Already Looks Transportation, Tesla a Bet on Scale, and Meta Is Buying People at Company-Level” issue should be assessed with one constraint in mind: robotaxis, a leak of billions of passwords, Meta and OpenAI fighting over engineers, the failure of Apple Intelligence, and the Anthropic court ruling are one market — technology is moving from demos into infrastructure, and the price of error, talent, and data is rising sharply.
The boundary of the “Xiaomi's self-driving car: a new Tesla rival?” case is defined by this point: Xiaomi, once seen as just a Chinese phone brand, has already shipped its own car and joined Tesla's rivals — an announcement becomes meaningful when a product that actually drives stands behind it, not just a loud name.
The “Silicon Valley vs Europe: the technology gap” scene leads to a working conclusion: the gap is striking — European partners still drive themselves and have no comparable know-how, while the Valley is having the very debates about self-driving and hundred-million-dollar engineer bonuses; what the sides treat as strategic is access to talent and a seat at the center of the race.
The boundary of the “Who wins the robotaxi race?” case is defined by this point: a ride in Waymo shows the difference between a striking demo and a product people already trust — Waymo spent years polishing a limited territory and set of scenarios, while Tesla is scaling faster across an enormous fleet, but its Austin robotaxi launch came with videos of errors and questions from regulators.
The decision in “Cadillac Escalade: Best in the US?” depends on one criterion: “autopilot” is not one technology but different levels of permission and liability — Cadillac allows long hands-free highway stretches, Mercedes accepts responsibility under certain conditions, and Tesla still requires constant supervision; the button looks similar, but the risks behind it are entirely different.
The decision in “16 billion passwords leaked: What happened?” depends on one criterion: a leak of billions of login-password pairs is a reminder that infrastructure is dangerous for reasons beyond models — in the US banks proactively analyze compromised credentials, while many outside that ecosystem learn of a leak only after a problem, and the more actions we hand to automation, the more basic access security matters.
The working conclusion from “How to protect yourself from password breaches in 2025” is that protection comes down to simple steps: use a password manager that tracks compromised and weak passwords for you, generate strong passwords unrelated to your personal data, and turn on two-factor authentication at least for the important accounts.
The boundary of the “Meta v OpenAI: theft of best engineers, $100 million bonuses!” case is defined by this point: Meta is trying to buy time by offering OpenAI's leading researchers packages the size of a whole startup — this is more than a salary war: a few people can set a model's direction, and a company's gap cannot be closed with GPUs alone.
The decision in “Apple Intelligence: Full failure?” depends on one criterion: against the race Apple looks especially slow — Apple Intelligence may formally appear in the interface, yet a simple question to Siri can still end in a useless redirect; a feature in the menu and a working product are not the same thing.
The “Anthropic: AI teaching in books is now legal” issue should be assessed with one constraint in mind: the Anthropic ruling shows the industry wants to learn from enormous bodies of data while ownership rules keep forming after the technology has launched — and the market is testing at once whether a system can be trusted with a physical action, how well data is protected, and who owns the intelligence it trained on.
What this episode is about
Robotaxis, a leak of billions of passwords, Meta and OpenAI fighting over engineers, the failure of Apple Intelligence, and a court ruling on Anthropic’s training may look like unrelated stories. They are actually one market: technology is moving from demos into infrastructure, while the price of error, talent, and data is rising sharply.
When a Tesla drives itself to its owner, it looks like the promised future. A ride in Waymo, however, shows the difference between a striking demonstration and a product people already trust.
Waymo spent years polishing a limited territory and set of scenarios. Tesla is trying to scale more quickly across an enormous vehicle fleet, but its robotaxi launch in Austin has been accompanied by videos of errors and questions from regulators.
The boundary is visible even inside expensive cars. Cadillac allows long stretches of highway driving without hands on the wheel, Mercedes accepts responsibility under certain conditions, and Tesla still requires constant supervision. “Autopilot” is not one technology but a set of very different permission levels and legal responsibilities.
The user sees a similar button, but the risks behind it are entirely different.
At the same time, the leak of billions of login-and-password pairs is a reminder that digital infrastructure is becoming more dangerous for reasons beyond models. Banks and services in the United States proactively analyze compromised credentials, while many users outside that ecosystem learn about a leak only after a problem occurs. The more actions we delegate to automation, the more important basic access security becomes.
Inside the AI race, Meta is trying to buy time by offering leading OpenAI researchers packages that once looked like the price of a startup. This is more than a salary war.
A few people can determine a model’s direction, and a company cannot quickly close a gap simply by adding more GPUs. Apple looks especially slow in this context: Apple Intelligence may formally appear in the interface, while a simple question to Siri still ends in a useless redirect.
The Anthropic ruling on books adds the final layer: the industry wants to learn from enormous bodies of data, while rules of ownership continue to take shape after the technology has already launched. The market is therefore testing three things at once—whether a system can be trusted with a physical action, how well data is protected, and who owns the intelligence on which it was trained.
Whether a system can be trusted with a physical action, how well data is protected, and who owns the intelligence on which it was trained; the market is therefore testing three things at once.
Episode transcript
The episode is in Russian; below is an English reading guide to the transcript (the full EN transcript is a machine translation). Voice matching applied to 107 segments: 76 identified, 2 mixed, 13 probable, and 16 unresolved.
Loading…