AI News Daily한국어

Code got cheap, verification got hard: Satya Nadella says his "dream next job" would be joining one of those teams to learn

Nov 17, 2026Original The Pragmatic Engineer Podcast(게르게이 오로스 × 사티아 나델라) · Fox 13 Seattle · The Economic Times · LeadDev ↗

In 3 lines

  • With code cheap, the bottleneck moved to verification, and 'factories' that turn every failure into a test emerged
  • Nadella says skills aren't devalued but get new leverage, calling the 60,000-person buyout a matter of agency
  • He says the junior hiring slowdown may be temporary, yet young developer employment fell ~20% from its peak
Code got cheap, verification got hard: Satya Nadella says his "dream next job" would be joining one of those teams to learn

Microsoft CEO Satya Nadella appeared on Gergely Orosz's podcast The Pragmatic Engineer, released on November 16. One of the few Big Tech CEOs who started out as a software engineer, he said that in a world where code generation has become cheap, what has become hard is verification. If he could pick his next job, he added, he would join a team whose code is mostly output and learn firsthand. We read the 64-minute conversation, which passed 60,000 views in a day, alongside reporting on the buyout offer to 60,000 employees, an open letter from staff, and research on early-career hiring.

Code became output, and verification became the hard part

"In a world where code generation has become cheap, what has become hard is verification," Nadella said. As code went from something people put in to something models put out, specs and intent matter again. But code as output "has a particular order of operations," he stressed; it does not mean everyone generating random code. Teams are building up a new kind of tribal knowledge, such as taking every failure and building back the test case to verify the fix, and one team even had to revert a Rust migration.

The IDE faded in about six months

Orosz asked why Microsoft, early with Copilot, had no immediate response when Claude Code arrived out of nowhere in 2025. Nadella credited Anthropic's model with making the CLI a form factor and agreed that in six months or so IDEs kind of faded away. "When model jumps happen nonlinearly, the form factor has to change quickly," he said, adding that "AI first product building in that sense is unforgiving" and "don't be in love with what you built." He still sees a possible rebirth of the IDE as 'decision engines' embed intelligence into software like functions.

One person runs a runtime, about ten run a factory

For examples of a 'software factory,' Nadella pointed to blogs from two internal teams. On the GitHub Copilot side, a single developer maintained an entire runtime, and the Cowork team is "nine or 12 or whatever number of engineers." Azure's networking team, which used to manage some 500 fiber operators worldwide through emails and ad hoc processes, built an agentic ops system connected to its repos. He sees technical teams shifting from doing the job to building factories that do the job.

The real work is in the brownfield

Existing systems are the bigger challenge than new projects, he said. With AI models now finding Windows vulnerabilities (CVEs), teams are building a new pipeline from crash traces to fixes, while the Windows build system is still the Windows build system. In brownfield projects, he said, the hardest parts are builds and CI/CD. And as coding agents showed up to work at GitHub and scale went through the roof, Microsoft is reworking the Git protocol and the storage system.

"My dream next job would be that"

Nadella said he has never worked on a team where code is mostly output and still has to learn, calling a spot on such a team his dream next job. He said he now understands why CTOs and VPs of engineering step down to work on small teams. Onboarding onto a codebase has actually become easier for early-career people, he noted, because teams running a software factory document everything and you can ask the agent instead of hunting for tribal knowledge. He pointed to the 'preceptorship' program Mark Russinovich and Scott Hanselman built inside Microsoft as one way for new engineers to learn on these teams.

Same skills, new leverage

Asked how engineers who spent decades mastering coding should reconcile the change, Nadella said "the same skills are still required." What changed is the order of operations and the tool chain; it is not about devaluing expertise but exercising it "with new leverage." Anyone at Microsoft with great depth who has curiosity and is learning the new ways to work is "absolutely flourishing," he said. He also recalled meeting students at the last GitHub Universe who were the first batch never to know an IDE before Copilot.

The 60,000-person buyout, framed as agency

Orosz asked directly about speculation that the offer is a prelude to layoffs. On October 26 Microsoft offered voluntary exits to about 60,000 employees at level 70 and below in the US, UK, Canada and India, out of roughly 223,000 full-time staff. Nadella said the shape of opportunity is changing and the offer "was mostly about let's give everybody more agency." Wages for systems engineers and AI training researchers are "through the roof," one of the fastest growing jobs is FDE (forward deployed engineer), and anyone stuck on a team that isn't transforming can vote with their feet, even inside Microsoft. "I wish I could just sit here and say it'll all be the same. I cannot," he said.

Satya Nadella · CEO, Microsoft · YouTube · The Pragmatic Engineer · Fox 13 Seattle ↗

Counterpoint: employees call it 'joyless'

The view from inside looks different. An open letter from 'Concerned Microsoft Employees,' circulating before the buyout announcement, called the company "joyless and demoralizing." According to The Economic Times, Microsoft's latest quarterly profit was $36.6 billion, up 32% from a year earlier, while employees complained of raises below 1%. The letter says staff are pushed to use AI tools in nearly every task without enough training and end up fixing wrong AI output.

Concerned Microsoft Employees (open letter) · The Economic Times ↗

Is the junior hiring slowdown temporary?

Nadella said it is too early to read the slowdown in new-grad hiring as a long-term trend. First people get reorganized into configurations that make sense, then new college hires are added. An AI-native company starts "with a token budget and a human," adding the next human once that person maxes out the budget, and reconfiguring every team that way will take time. By contrast, Stanford Digital Economy Lab research found employment of software developers aged 22 to 25 was down nearly 20% from its late-2022 peak as of July 2025. In a February 2026 follow-up, the same authors said that with stricter controls the decline is significant only after 2024.

Satya Nadella · Stanford Digital Economy Lab · YouTube · The Pragmatic Engineer · LeadDev ↗

Can you swap any model? And back to school

Nadella urged keeping the harness, context, memory and action space separate from any one model. His test: "you should be able to replace any model and still hill climb on an eval that's your private eval." Because models are getting more capable but not more reliable, he said, agents should be treated as an insider risk, with authority separated from intelligence. Asked what he would do to become a standout engineer in a year or two, he said to skip the headlines and learn hands-on. "If I could change the ratio of my talk to do I would do it," he said, closing with: "When am I at my best? It's when I'm in the school mode. So let's get back to school."

Comments 0

Select a sentence in the story (long-press on phones) to comment on it.

OriginalThe Pragmatic Engineer Podcast(게르게이 오로스 × 사티아 나델라) · Fox 13 Seattle · The Economic Times · LeadDev↗

More insights