Energy storage is scaling faster than solar or wind ever did!

Four companies now spend at a staggering scale. Hyperscaler AI capex is on track for ~$733B in 2026, about 87% of the base U.S. defense budget.

The cost center of energy has become more about delivery than production.

We’re excited to see Base Power launch Base Core today and to keep backing Base Power and others reindustrializing the U.S. grid.

AI is changing the way we use Search.

As AI Overviews and AI Mode scale, users are asking longer, more conversational questions and spending materially more time in the Google app.

Better answers seem to be creating more questions.

The next scaling law: learning velocity "on the job".

AI has been defined by pre-training: spend more upfront, get a smarter model. But a shift is underway, from pre-training to continual, on-the-job learning.

A new study measured whether models improve when given time, feedback, and room to work. The data shows the newest models learn on the job almost twice as fast as models from just three months earlier, and the gains compound with iteration.

It turns out models may not be so different from us: given room to practice, they get better.

For more on this C:\Take, watch Max:

Major cloud providers are re-accelerating as AI adoption increases!

CPU is back from the dead.

Before AI, a human would think about what to do and then take action on their computer screen. Today the GPU plans the action by writing tokens, and the CPU executes it by writing code.

The workflow a lot of people can relate to is pulling data, analyzing it, and displaying a chart. Every single one of those steps can now be done by an agent using a computer driven by a CPU. That is massive CPU demand, because agents can move faster, longer, and deeper than a human possibly could.

Before, you had to click, read, and plan a task yourself. Now an agent runs that same sequence in a loop. The GPU decides the next move, the CPU carries it out, and the cycle repeats. Every cycle is a GPU-to-CPU round trip.

For more on this C:\Take, watch Frank and Nick:

Codex adoption has broadened from engineering into nearly every function across OpenAI. Is this a predictor for the rest of the enterprise world?

We spend a lot of time thinking about the big waves of technology, the move from mainframe to PC, the networking of those PCs, web 1.0 and web 2.0, SaaS, and now AI.

What is easy to miss is that inside every one of those waves are smaller waves. We call them handoffs, or shakeouts, and they tend to follow a recognizable pattern: new IPOs start appearing, more private companies form, drawdowns get sharper, and the whole market turns more volatile. By the time it settles, the companies sitting on top are rarely the same ones that led going in.

That raises the obvious question of whether software is in one of these handoffs right now.

If you look at quarterly net new ARR, the AI labs are starting to out-add both the hyperscalers and the entire public software universe. When a new set of entrants is putting on more revenue than the incumbents that have defined the category for years, that is usually what the early stage of a handoff looks like.

History suggests the leadership board gets redrawn each time this happens. When software moved from the enterprise era into SaaS, only a handful of the largest names carried over into the next list, and the companies now leading the AI era, OpenAI and Anthropic among them, did not even exist during the last shakeout.

For more on this C:\Take, watch Max:

Innovation waves get credited to new technology. The real trigger is when great talent becomes available to build.

The Talent Mobility Index tracks how freely top tech talent can move in a given year, blending hiring, job change, and AI funding data into a single score.

The AI labs concentrated the best builders in the world and created extraordinary success. Now, as that success matures, some of those builders are stepping out to start what's next.

Data collaboration with Outcast Ventures and Coatue.