When I wrote Changing Culture in August 2022, the strange part was that the post had AI-generated images in it.
DALL-E had just crossed into public consciousness. ChatGPT had not launched yet. The images were beautiful, weird, unstable, and obviously generated: fake text, warped interiors, dream logic, almost-branding, horses and hats and company lobbies that looked like the machine had understood the myth but not the physics.
That was enough.
The point was not production quality. The point was that the machine could now participate in the act of cultural imagination.
Four years later, the image is no longer the interesting unit.
The workflow is.
The image was the demo. The workflow is the product.
On May 19, 2026, OpenAI published a post about pushing provenance through C2PA, SynthID watermarking, and public verification tools.OpenAI, "Advancing content provenance for a safer, more transparent AI ecosystem", May 19, 2026. The important part is not just watermarking. It is the attempt to make generated media carry origin, attribution, and verification metadata through the distribution layer. That shift matters because the hard problem is no longer just whether a model can produce a convincing image. It can. The harder problem is whether generated media can travel through the world with origin, attribution, context, and trust attached to it.
The same thing is happening inside companies.
In 2022, I framed culture as something that cannot be changed by fiat. Culture emerges from repeated mechanisms: rituals, incentives, spaces, slogans, hiring loops, reward systems, and the carriers who transmit the behavior to everyone else.
I still think that is true.
But 2026 makes the blunt version harder to ignore: culture changes when a system discovers its bottleneck and reorganizes around it.
Sometimes that bottleneck is latency. Sometimes it is coordination. Sometimes it is trust. Sometimes it is human attention. Sometimes it is the employees themselves.
Layoffs make this visible.
Layoffs are usually discussed as economic events: margin pressure, overhiring, capital costs, business resets, AI productivity assumptions. Those explanations can be true. But they are incomplete.
A layoff is also cultural surgery.
It removes carriers of the old system. It changes what survivors believe is protected. It rewrites incentives without needing to rewrite the values deck. It teaches people, very quickly, what the organization now considers sacred.
As of this update, Layoffs.fyi is tracking more than 114,000 tech employees laid off across 148 tech companies in 2026.Layoffs.fyi reported 114,173 tech employees laid off across 148 tech companies when this 2026 update was drafted. That number is not just a labor-market statistic. It is a culture-change signal.
The cost is not abstract. It shows up as layoffs, attention collapse, career tournament culture, and a generation trained to either climb faster or disappear. The organization calls it efficiency. The people inside it experience it as weather.
The AI layer sharpens this.
AI gives companies a new story about leverage: fewer people, more output, smaller teams, faster loops, less coordination drag. Sometimes that story is real. Sometimes it is managerial fantasy with a better demo. Either way, it gives leadership permission to ask a colder question:
What are humans for inside this system now?
This is where the old essay's religious-conversion metaphor still holds, but needs an update.
Culture is not only converted through belief. It is converted through constraint.
If the constraints change, the culture follows. If the tooling changes what work is possible, the culture follows. If the company decides human attention is the scarce resource, the culture follows. If AI expands the hardness frontier and makes previously impossible work tractable, the culture follows.
Output is the easy part to measure.
The harder part is what kind of people these systems produce inside the institutions that adopt them.
That is why "know thyself" stops being spiritual wallpaper and becomes systems design. Humans are forever the bottleneck in every system. The choice is whether we adapt blindly to existing systems, or build from scratch with a more honest understanding of our bottleneckness.
The image was the demo.
The workflow is the product.
The layoff was the financial event.
The culture change is what remains after the cut.
References
- Original 2022 essay and DALL-E prompts: Changing Culture.
- Related personal notes: Big Math, mass layoffs and attention destruction, corporations prioritizing resources over humans, and know thyself / humans as bottleneck.