Daily Briefing — May 25, 2026

A veteran engineering leader who has run software teams at Microsoft, Snap, and Google is pushing back on the "AI kills coding jobs" narrative, and his argument is worth taking seriously. His framing: engineers were hired to solve problems, and code was just the tool they used to do it. That changes how you should be thinking about your own position right now.

The piece is honest about not knowing what happens after the next 12 to 18 months. But within that window, the author is clear that the shift is already underway and moving faster than anything the industry has seen before. AI agents can scaffold, generate tests, wire up APIs, and churn out boilerplate faster than any human team. That part of the job is effectively being automated, and the open question is what fills the space that creates.

The implication for you is concrete. If the execution layer gets cheaper and faster, the constraint moves up the stack. Deciding what to build, understanding why it matters, and navigating the messy human context around a product become the scarce resource. The engineers who thrive in this shift are the ones who were already doing that work, or who start doing it now before the window closes.

Nvidia just dropped another quarter that made Wall Street look like it was lowballing on purpose. The company posted 92% year over year growth in its datacenter business alone, hitting a record $75.2 billion for that segment. Jensen Huang basically went on stage and said the AI infrastructure buildout is the largest in human history, and the numbers are not exactly arguing with him.

For context, US tech giants are collectively planning to spend around $750 billion this year on AI infrastructure. A significant chunk of that flows directly to Nvidia because nobody else has the full stack locked down the same way — chips, software, infrastructure, the whole thing. Nvidia has made itself hard to route around, and that took work.

Huang also said he expects Nvidia to grow faster than the capital expenditure of the hyperscalers themselves, which is a bold thing to say and a bolder thing to believe. When the company supplying the picks and shovels is outgrowing the miners, the economics of AI are shifting in a strange way.

The takeaway is that agentic AI, the kind that does actual work inside companies, is no longer a pilot programme. It is getting funded at a scale that changes job functions, team structures, and what skills are actually worth something in the next two years.

Google just dropped Omni Flash, the first model in its new Omni family, and it is now live inside Flow, the company's AI video generation and editing platform. The pitch is ambitious: eventually, any input goes in, anything comes out. Photo, video, text, whatever. For now it is focused on video, and it is a meaningful step up from the previous Veo model. You can feed it an existing video plus a text prompt and it builds from there, with better character consistency and more real world knowledge baked in.

The Verge writer tested this by deepfaking themselves in front of the Eiffel Tower and sending a kid's stuffed animal on a virtual rafting trip. Both worked. Forget the cute examples — what matters is that someone with no specialist skills, no production budget, and no real effort pulled off results that would have taken a video team serious time and money just a few years ago.

Google isn't hiding the ball about how far the "anything to anything" vision still has to go. But the gap between what these tools can do now and what most people think they can do is getting wider every month. If your work touches content, communications, marketing, or client facing materials at all, the production bar has already shifted under you.