The AI Acceleration Gap

The compounding divide between AI early adopters and everyone else

Nathaniel Whittemore on a recent AI Daily Brief episode coined the term “AI acceleration gap” to describe a feeling I’ve had for several months:

A widening gap is emerging between people and organizations that are compounding new capabilities and those moving at a linear pace, and recent advances have made the divide feel suddenly sharper.

Kevin Roose of the New York Times posted something similar on X:

i follow AI adoption pretty closely, and i have never seen such a yawning inside/outside gap.

people in SF are putting multi-agent claudeswarms in charge of their lives, consulting chatbots before every decision, wireheading to a degree only sci-fi writers dared to imagine.

people elsewhere are still trying to get approval to use Copilot in Teams, if they're using AI at all.

The AI epicenter is San Francisco and I fear being on the outside has consequences.

---

I live in New York now, but I spent the first twenty years of this century in San Francisco.

I watched Craigslist, wikis, web 2.0, and social sharing go from weird to normal. Then robust web apps evolved into the SaaS explosion. Then mobile apps, Uber, Airbnb, and self-driving cars. Each wave started as something most of the country didn’t understand and was simultaneously unremarkable if you lived in the Bay Area.

Staying plugged into the latest technology was effortless. The billboards were all tech. Everyone you met worked in tech—engineers, sales, customer success, ops—they all came to SF for this industry. You’d overhear conversations at coffee shops. Your friends would mention what they were working on. It was in the air.

But outside of SF, you have to work to stay current.

I still work in tech, but in my day-to-day life here I almost never encounter anyone paying attention to AI. A few friends and colleagues I geek out with, but everyone else? Nope.

There’s this window where San Francisco is living in the future and everyone else isn’t. But eventually these things diffuse through the rest of society and that early-adopter technology becomes normal.

I’m not sure it plays out this way with AI.

---

I’m afraid that in this case, first mover advantage is insurmountable.

The companies that figure out how to fully leverage AI first with teams of 10x engineers will run circles around everyone else. The individuals who figure this out first, who become fluent in these tools, will be dramatically more valuable than everyone around them.

The pace isn’t slowing down. Just in the last two months we’ve seen examples of Claude Opus 4.5, Claude Code, Conductor, compounding engineering, the Ralph Wiggum loop, and Clawdbot Moltbot demonstrate huge gains in productivity. Every week there’s something new. The models get better. The systems around the models get better. You blink and you’ve missed a generation of progress.

If a company doesn’t have a significant presence in San Francisco or you’re not immersed in online tech culture, the gap that Kevin and Nathaniel describe could have real consequences—for people’s careers, for companies that fall behind, for anyone who isn’t paying attention.

I’ve lived through enough of these waves to know what the early stages feel like.

This one feels different.