I settled into a chair to prep for the day. Sixteen stories below, the city moved with intent.
A few blocks surrounding Almaden Boulevard had been closed off for what would be just about a week.
I knew that not from a map, but from my late arrival the night before hauling a much larger-than-practical green suitcase through streets already closing in on themselves.
Green, but not neon green. Not quite on brand.
On every corner, neon green branding filled the streets.
It was clear GTC was already in motion.
Not just as an event, but as a system taking over the city.
By morning, the city had shifted.
The sun that day felt a little more brutal than normal.
Around me, the kind of hustle and bustle familiar only to those who live here.
For me? Enigmatic.
Having arrived late the night before, I should not have been awake that early.
And yet, I was far too excited to sleep.
Now, as I write this with a bit more distance from the week, somewhere between the comedown and the clarity that follows I realise it wasn’t just the scale of GTC that stayed with me, but what it meant.
Not the headlines. The undercurrents.
The signals you only really catch when you’re paying attention.
If I had to distill it, it comes down to five things that haven’t left me:
1. Experimentation is now embedded into the operating model.
What stood out most wasn’t just the volume of innovation, but how deeply experimentation is woven into the way companies now function. It’s no longer a discrete phase or department; it’s continuous, expected, and systematised. And yet, as this becomes the norm, there’s a quieter question underneath it: whether this level of constant iteration is translating into proportional economic value, or simply becoming the cost of staying relevant. The system is optimised for testing but still learning how to measure what truly matters.
2. Intelligence is being produced at industrial scale.
The shift toward AI factories makes something unmistakable: intelligence is no longer crafted, it’s manufactured. Entire infrastructures are being built to generate, refine, and deploy it continuously. There’s a sense of entering a new industrial era, one defined not by physical goods, but by cognitive output. At the same time, it raises a more philosophical tension: whether scaling intelligence inherently leads to progress, or whether it risks amplifying systems that haven’t fundamentally evolved. The capacity is undeniable. The direction is still unfolding.
3. Orchestration, not compute is becoming the real advantage.
The conversation has moved beyond access. Models, tools, even compute at a high level are increasingly available. What differentiates now is the ability to coordinate across agents, systems, teams, and institutions. The challenge is less about building capability, and more about making it work together coherently. At the same time, this introduces a new layer of complexity: governance. As autonomy increases, so does the need for control, oversight, and alignment. The systems are becoming more powerful and more difficult to manage.
4. Narrative has become part of the infrastructure.
What’s being built is not just technology, but understanding. The way companies communicate on stage, in demos, across platforms feels deliberate, structured, and tightly integrated with the product itself. But as expectations rise, so does the need to sustain belief. Breakthroughs are no longer surprising; they are anticipated. And in that environment, narrative begins to carry weight equal to performance. It’s not always clear where one ends and the other begins but both are shaping the trajectory.
5. The real question is no longer whether we can build but what we are building toward.
Across the week, there was a subtle but persistent tension in how this moment is being interpreted. For some, the answer is acceleration, build faster, scale further, move without hesitation. For others, there’s a sense that speed alone doesn’t resolve deeper structural questions about progress. Both perspectives seem to coexist. And perhaps that’s the point: the future isn’t constrained by our ability to build anymore, but by our clarity on what that building is ultimately in service of.
What the week revealed is not just how quickly we can build
but how much room there is for those willing to think, and act, at scale.
Ambition, in this context, is not excessive.
It is required.
It shows up across proprietary and open models alike
not as a choice between them, but as a “yes, and.”
