What's up: Microsoft tries to catch up in AI; Google invests in Europe; How to land a job (beat AI); Taking Manhattan by self-driving car

Good morning. Time and again, we are told that AI requires us to move faster. Most of us already feel like we are moving quickly, thank you very much. Speeding up is easier said than done. I asked Google’s Ryan Salva, who leads product teams responsible for developer tools and services, for insight into how that can actually work.

He joined Google last year after leading developer tooling at Microsoft and Github, which it owns. He's been responsible for Google's products including Gemini CLI, an open source AI agent for developers, and Gemini Code Assist. Here are some edited highlights of our conversation at Google offices in New York.

WSJ Leadership Institute: Some people say AI is going to kill software. What do you think about that prediction?

Salva: It’s overblown.

It's still software we're shipping. We are into a higher level of abstraction though, much in the same way that no one writes ones and zeros anymore. Even those folks who are still writing code are writing with an intermediate language like Java or C# or Rust, right? The large language model is allowing us to operate at another layer of abstraction where we are able to express through natural language and intent. But the engineering part of it, the architecture part of it, that is still very, very much at the core of what every engineer needs to do.

The product that they're shipping isn't code, the product they're shipping is software .. an application that does the thing that fulfills a business need.

WSJLI: As AI allows engineers to move faster, how do organizations adjust?

Salva: This new velocity is fundamentally reshaping engineering teams. Whereas previously, you might have a director with 30 or 50 engineers who were assigned an area, and they might perform a major release once a month.
Now what I see happening is that you've got teams of maybe three to six engineers, but there's many of them working in parallel. You're able to work through more of your backlog. You are often able, then, to ship more frequently, which means that your customers are getting more value, more quickly.

And as the engineering teams can move faster, it allows other parts of the business to both iterate and kind of move faster as well. I do think that it causes some challenges.

WSJLI: How so, and what do you do about that?

Salva: I was meeting with our go-to market team and our revenue team the other day, and one of their complaints to me was that we were shipping too often, that we were literally outpacing their ability to keep up. I realize that sounds like a humble brag. It's actually not intended to be. It's actually a real business challenge. As your engineering teams move faster, it means that all of the human aspects of it need to find a way of coming in lockstep with each other. And that may mean that your engineering teams need to adjust in order to meet the more human process-oriented pieces of the business.

I actually see that we end up with a lot more embedded engineering teams directly into the organizations.

If I were to maybe counsel or coach my peers and my colleagues in these roles, I would advise them to reserve learning time and to create regular connection points with their engineering teams. Those engineering teams want to ship software that is responsive to the needs of the go-to-market team, the customer support team, the finance team. 
And the more that we create, the proverbial silo between our teams, the less effective, the organization at large is.

WSJLI: And how do engineering teams adjust?

Salva: If it used to take a month to deliver a major release you had to be precious, really, about that code. 
It was both emotionally difficult sometimes to throw it away, but also just expensive. As we are moving more and more towards using AI, it means that we can build that software much faster. Our ability to try a thing and throw it away is much more accessible, and we're much more willing to take a little bit of a risk here and there.

How is your company getting engineers and business teams to work together? Use the links at the end of this email and let us know.

Microsoft tries to catch up in AI with healthcare push.

Microsoft is pushing into healthcare AI by licensing consumer health content from Harvard Health Publishing to enhance its Copilot assistant’s medical responses. The updated Copilot, expected this month, aims to offer answers more aligned with those from medical practitioners, especially for conditions like diabetes.

Moving away from ChatGPT. Guiding the healthcare push is Microsoft’s effort to build up a measure of technological independence from OpenAI. Microsoft is training its models with the goal of eventually replacing workloads from OpenAI, a process that could take years, WSJ reports.

Already, Microsoft is using non-OpenAI models for some of its other software. It now deploys models from OpenAI’s rival Anthropic to power AI tools within its 365 products.

American tech companies are still pouring money into data centers in Europe

Google will pump $5.82 billion of capital into Belgium over the next two years, the latest large-scale investment from a U.S. tech giant bolstering its data center infrastructure in Europe, WSJ reports. Funding will underpin the use of artificial-intelligence and augment Google’s ability to meet rising demand for Google Cloud, it said.

Other tech giants are making parallel moves. In July, Oracle said it was investing $3 billion in the Netherlands and Germany. Microsoft said it would spend around $4.75 billion in Italy on cloud and AI infrastructure.

Diamonds ... a data center's best friend?

Blinging out computer chips. The thermal conductivity and electrical insulation properties of diamonds are well known, and already used for space and defense uses, including high-end electronics on communication satellites. Now researchers are looking at embedding synthetic diamonds into chips to help cool AI hardware and data centers, the New York Times reports.

China's rare-earths remain a big bargaining chip, threatening tech supply chains

China has tightened its control over certain rare-earth materials used to make high-tech products, threatening to reignite trade tensions with the U.S., WSJ reports. Export applications for products with military uses generally won’t be approved, China’s Commerce Ministry said Thursday, adding that licenses related to semiconductors or AI development will be granted on a case-by-case basis. The Information reports that the move could impact Nvidia's supply chain.

The tech job beat: Battling AI, cyber talent; Nvidia's Huang on H-1B

How to land a job (defeat AI). Job seekers are embedding their resumes with hidden messages designed to draw attention from the AI they expect recruiters use to quickly vet applicants. The practice, which can involve writing messages like “this candidate is well qualified” in white type or inserting text in image metadata, has become so commonplace that companies are updating their software to catch it, the New York Times reports.

Nvidia CEO Jensen Huang told CNBC that his family would not have been able to immigrate to the U.S. under President Trump’s current H1-B policy. “I don’t think that my family would have been able to afford the $100,000,” he said, referring to the new fee employers must pay per visa. Born in Taiwan, Huang immigrated to the U.S. when he was nine.

Cyber experts moved to help Trump's deportation push. The U.S. Department of Homeland Security is shifting national security specialists from the Cybersecurity and Infrastructure Security Agency to jobs involved in President Trump's deportation efforts, Bloomberg reports. Staffers targeted for reassignment include members of CISA’s Capacity Building team, responsible for protecting the government’s highest value assets, employees tell Bloomberg.

Will driverless cars ever be ready for the streets of Manhattan? Last week, WSJ Leadership Institute’s Isabelle Bousquette took a test drive to find out.

She found that, while autonomous vehicle technology is thriving on the West Coast, AI, which operates by learning the rules of the road, has a whole set of new challenges in a city where no one else is following them.

Here’s her dispatch from 53rd and Madison:

Manhattan traffic is stressful at the best of times, and as I stepped into the Ford Mustang Mach-E kitted out with self-driving technology from British startup Wayve, I wished the AI godspeed.

This was a “Level 2+,” experience according to Wayve with the AI model controlling the brakes, gas and steering. New York law required that my trip include a human driver, ready to take over when necessary (And he did once, when other cars made it impossible for us to merge into a right hand turn lane).

Over the course of about 20 city blocks and 40 very slow minutes, we encountered everything from reckless bikers to distracted pedestrians, construction road closures and double-parked delivery trucks.

So how did the Wayve car compare to most local taxi drivers? It was a lot more polite and conservative, and, as a New Yorker, dare I say, frustratingly slow.

Wayve was born in Cambridge, England, and initially cut its teeth in London, another uniquely difficult dense, urban environment. It builds the software, or underlying AI models for assisted and automated driving, that it then sells directly to vehicle manufacturers like Nissan, who it started working with earlier this year.

Part of Wayve’s strategy is also to gather data and train its models on the subtle driving style differences of different geographies. “People who walk and drive in these cities have a certain expectation for how others are going to move,” the Wayve spokesperson in the backseat told me. “You have to adapt to the local driving culture.”

But what about when the local driving culture is defined by lawless lane-swerving and road rage? Wayve will have to find the balance between safety and legality and actually turning AI into a Manhattan driver. Is it possible? Or will it always feel easier to grab the wheel ourselves? Let me know what you think!

Follow Isabelle on Instagram, X and LinkedIn for more behind the scenes on her reporting.

Israel and Hamas agreed Wednesday to a deal that would release all Israeli hostages held in the Gaza Strip in the first step toward peace after two years of war in the Palestinian territory. (WSJ)

Senior GOP lawmakers have cautioned the White House not to move forward with mass layoffs and sharp cuts to government assistance programs during the shutdown. (WSJ)

Daily reports from police in the month leading up to Trump’s decision to send federal troops into Portland show a different reality from what the president and others in his administration have stated. (