---
title: "Why is nobody talking about this?"
description: "Chip stocks rocketing. Robots rolling off lines. AI starting to improve itself. The biggest story of our time is unfolding in plain sight, and most people haven't noticed."
date: 2026-05-10
image: "/uploads/2026/05/why-is-nobody-talking-about-this.jpg"
categories:
  - "AI"
  - "Future"
---

I had a strange moment this week.

I was reading [a post by Matt Shumer on X](https://x.com/mattshumer_/status/2021256989876109403) about how fast things are moving on the AI frontier, and I caught myself nodding all the way through. New training capacity coming online every month. Nvidia, AMD, Intel and the memory makers all sitting at sky-high valuations. Robots moving from prototype to production at Figure, Tesla, X1 and across a long list of Chinese manufacturers I'd never heard of two years ago.

Then I closed the browser, opened my calendar, and looked at the week ahead. Status meetings. A campaign brief that's been in review for three weeks. A six-figure budget approved for a small website rebuild that I could prompt my way through in a long afternoon. A roadmap session for work an agent could do overnight if anyone bothered to point it at the problem.

That's the gap nobody can quite explain. What's happening on the frontier and what's happening inside the average company are two completely different speeds. And the gap keeps getting bigger.

## What I can actually do today

I built a small internal tool from scratch in two days last week. I'm not a developer. I don't read or write code. I structured the project, told [Claude Code](https://www.anthropic.com/claude-code) what I wanted, watched it plan, build and test its own work. The result was cleaner than most of what I've seen shipped by real teams, in a fraction of the time.

[I've written about this before](/thoughts/two-days-to-build-what-used-to-take-months). What's changed is the consistency. It's not a lucky outcome anymore. Software, for me at least, is more or less solved.

In the same week I let Claude Code take over my computer to set up a Google Ads campaign. I had ChatGPT generate a moodboard for an app concept. I wrote a strategy presentation in PowerPoint with the right branding by describing what I wanted in plain language. None of this is exotic. Anyone with a credit card and a couple of hours can do it. Almost nobody does.

## The frontier is sprinting

While most people haven't tried any of this, the people building the frontier are running flat out. Power plants are being commissioned to feed the next generation of data centres. Chip shortages aren't really shortages anymore, they're allocation decisions about who gets the next wafer.

Robotics is moving faster than I expected. Figure's humanoids are doing real warehouse work. Tesla's Optimus is leaving the factory floor. X1 is shipping. Chinese manufacturers are rolling out machines that walk, see, lift and carry, at a price point that keeps falling.

And AI is starting to improve itself. [I wrote about this loop a few weeks ago](/thoughts/closing-the-loop-how-ai-software-starts-improving-itself). The bottleneck was never raw intelligence. It was the missing feedback loop. Once that loop closes for real, the curve doesn't bend gently. It launches.

## Two populations, drifting apart

There is a small group of people living about a year ahead of everyone else.

They use these tools every day. They've built muscle memory for prompting. They've wired up agentic workflows and seen what happens when you let a good model run for hours unattended. They notice when a new model ships because it changes their week. Every Friday they're a little more leveraged than they were on Monday.

Then there's everyone else. Regular people. Most have tried ChatGPT once or twice, asked it to write an email, thought "huh, that's neat" and went back to writing emails the way they always have.

The gap is widening every week. It's not about intelligence, education, or access. The tools are mostly free. The gap is about exposure and practice. Someone who has used Claude Code daily for six months is operating in a different reality than someone who has opened ChatGPT three times. The leverage difference is closer to a factor of ten than a factor of two.

And because the frontier moves every week, the gap compounds. The people in the bubble get better at using the latest tools the day they ship. Everyone else hasn't even tried last year's tools. A year from now this will be a chasm, not a gap.

> "The future is already here. It's just not evenly distributed."
>
> **William Gibson**

That line has been quoted to death. It has never been more accurate than it is right now.

## The org chart is the bottleneck

Here's the part that gets missed in most of these conversations. The technology isn't really the problem anymore. Most companies are wearing an organisational shape designed for a completely different way of working.

A typical company is built around departments. Design, engineering, marketing, finance, legal, ops. Work moves between them in handoffs. A request becomes a brief, the brief becomes a ticket, the ticket sits in a queue, the queue gets reviewed in a stand-up, the stand-up feeds a sprint. Four weeks gone by the time the loop closes.

Drop AI into that machine and mostly nothing happens. The AI makes each step a bit faster, but the structure doesn't shrink. The handoffs are still there. The five reviewers are still there. Procurement still takes a quarter. Security review takes a quarter on top of that. You end up with people writing prompts to draft documents that go through the same fifteen-step approval pipeline as before.

[I wrote a longer post about this](/thoughts/the-ai-organisation-why-most-companies-are-doing-it-wrong). The short version is that AI doesn't reward bigger teams. It rewards smaller, faster ones, where one person can take a problem from definition to shipped solution. That setup is incompatible with how most companies are structured today. None of "I'll get my marketing manager to talk to your design lead and we'll set up a kick-off next Tuesday" survives contact with a small team that just builds the thing.

The companies pulling ahead aren't the ones with the biggest AI budgets. They're the ones willing to throw out the old shape. Cut the layers. Shrink the teams. Let small groups own real problems end to end. Most companies won't do this voluntarily. They'll keep tweaking the edges until a competitor with a tenth of the headcount eats their lunch.

## Why is the discussion missing?

This is the part I genuinely don't understand. If a meteor were on a five-year trajectory toward Earth, we'd be talking about nothing else. The arrival of machines that can do most of our cognitive and physical work is roughly that scale of event. The public conversation barely registers it.

A few theories.

It's invisible. AI doesn't look like anything from the outside. It's a chat box. There are no rockets, no smoke. Hard to feel the magnitude of something that fits inside a browser tab.

It's threatening. If you sit with what's coming, the implications for your job, your savings, your kids' careers are uncomfortable. Most people protect themselves from things they can't yet act on by simply not looking.

The people in charge would rather not have this conversation. Politicians don't want to be the ones to tell voters that half the jobs are about to change shape. Executives don't want to announce a fifty percent headcount reduction. Easier to let the next administration, the next CEO, the next quarter handle it.

Whatever the mix of reasons, the result is the same. The biggest story of our time is being told mostly in private.

## What to actually do

When the public conversation finally catches up, it's going to feel sudden. A bit like COVID, except instead of staying home for a few months, the rules of work and money and value will have quietly rewritten themselves while everyone was looking the other way. [I wrote about the slow burn and the sudden shock here](/thoughts/two-paths-to-the-future). Which one it ends up being depends mostly on when the rest of the world wakes up.

If you've read this far, you're already ahead of most. The next step is to stop reading about it and start using it. Pick one thing this week you've been doing the long way and try to do it with AI instead. A boring report. A slide deck. A research dive. A campaign brief. Anything.

You don't need to predict where this goes. You need to develop a feel for what the tools can already do, because the tools you'll have in twelve months will make today's tools look like toys. [Practising beats studying](/thoughts/just-do-things-why-practising-ai-beats-studying-it). It's the only real preparation.

I think the wake-up is coming. I just hope it arrives early enough for most people to have a soft landing instead of a hard one.