AI, the Punch You Don't See Coming

Most people get knocked out by punches they don’t see coming. Not because the punch is too fast. Because they’re not looking.

A few weeks ago, I sat with a friend late into the night. He’s a web developer. Smart. Skilled. Employed. He’d been quiet for most of dinner, and when everyone else left, he finally said what was on his mind.

He sees it coming. The wave. The shift. Whatever you want to call it. He’s been reading the headlines, watching the demos, noticing which jobs are quietly disappearing from the postings. And he doesn’t know what to do.

We talked for hours. Not about tools or prompts or productivity hacks. About fear. About identity. About what it means when the thing you spent years mastering becomes a commodity overnight.

I didn’t have clean answers for him. I still don’t. But I’ve been thinking about that conversation ever since.

I understood his fear more than I wanted to admit. Because I’d felt it once too, the moment you realize the ground is shifting under you faster than you can adjust.

I Was Skeptical Too

Ten years ago, I was building chatbots and machine learning systems. I knew the limitations intimately. Not enough training data. Not enough capability. The outputs were toys. Impressive demos that collapsed in production. I watched hype cycles come and go, and I learned to tune them out.

When ChatGPT launched, I tried it like everyone else. Email polishing. Unit tests. Useful, but I didn’t see what the fuss was about. The chat interface felt clunky. The outputs were generic. I went back to my regular workflows.

About six months ago, a friend was building an app to summarize books. OpenAI couldn’t take the whole book at once, so he needed to run his own models. I helped him test what was available on Hugging Face: BART, GPT-2, GPT-J, LLaMA.

The results were nowhere close to usable. Summaries missed the point entirely. Key themes got lost. And even if quality wasn’t an issue, the cost to run these models at scale was unrealistic for a startup.

He put the project on pause.

Last month, he picked it back up. Now he can batch the calls, and OpenAI summarizes an entire book for him in minutes. What was impossible six months ago is now a feature he ships.

That’s not incremental improvement. That’s a different capability entirely. Nothing prepared me for the jump in quality. It wasn’t “a little better.” It was a different species.

Won’t This Just Blow Over?

I hear this a lot. It’s a bubble. The valuations are insane. It’ll collapse and things will go back to normal.

Look at the investment patterns. Nvidia, OpenAI, Anthropic, Microsoft, they’re all investing in each other. Valuations inflating on circular capital. It looks like a bubble. It might be a bubble.

But here’s the thing about bubbles: the technology is often real even when the valuations aren’t. The dot-com bubble popped. Amazon, Google, and the internet didn’t disappear.

The infrastructure might be shaky. The capability is not. The change in how work gets done is happening regardless of what happens to stock prices. Waiting for the bubble to pop before paying attention is waiting for a signal that won’t save you.

What I Hear From Friends

No time to learn another tool. Tried it once, wasn’t impressed. My job is too nuanced. It’s overhyped, I’ll wait for things to settle.

I get it. I said the same things.

But the gap between “email polishing” and “real work output” closed in months, not years. Decisions made based on the AI of six months ago are decisions made about something that no longer exists.

The pace is what people underestimate most. This isn’t like learning software that stays stable for a decade. The capability curve is steep and accelerating. Every few months, tasks that seemed impossible become trivial.

Most people think they’re evaluating a tool. They’re actually evaluating a timeline.

They’re not scared of AI. They’re scared of losing the script of their lives.

My web developer friend understands code. He can see the trajectory. That’s why he’s scared. Most people don’t have that window. They’re making career decisions based on a snapshot that’s already outdated.

The Numbers

Seventy-seven percent of American workers would face financial difficulty if their paycheck were delayed by one week. Over a quarter of households spend more than 90% of their income on necessities.

There’s no buffer. No runway.

In October, U.S. employers announced 153,074 job cuts, up 175% from October last year. Through the first ten months of this year, over 1.1 million cuts have been announced. That’s the highest October total in over twenty years.

Between Q1 2023 and Q1 2025, white-collar job postings fell 35.8%. Forty percent of white-collar job seekers in 2024 failed to secure a single interview.

The jobs aren’t moving to other companies. They’re leaving the economy entirely.

When I told my friend these numbers, he went quiet. Then he asked: “Why isn’t anyone talking about this?”

I still don’t have a good answer.

One Thing You Can Do Monday

I don’t have a comprehensive solution. Policy makers should be paying attention. They’re not. Industry leaders should be preparing the workforce. They’re not.

So here’s something concrete.

Pick one task you do regularly. Something that takes 30-60 minutes. A report you write. An analysis you run. A document you review. This week, try doing it with AI. ChatGPT, Claude, Gemini, pick one.

You might not get it right the first time. The output will be generic, or wrong, or useless. That’s what happens to everyone.

But pay attention to what’s missing. What did the AI not understand? What context would it need to do this well? That gap is where the real learning is.

You don’t need to get good at AI tomorrow. You just need to stop guessing what it can do.

My friend started doing this. Not with the goal of automating his job. With the goal of understanding what the technology can and can’t do. He’s not frozen anymore. He’s learning.

The pace isn’t slowing. It’s layering.

The blanket is being woven. Every thread is another capability, another tool, another company figuring out how to do more with fewer people. Eventually it covers everything. The question is whether you’re holding part of it or lying under it.

The punch is coming. You might not be able to stop it. But you can see it. You can move. You can adapt.

Or you can stand still and wonder what happened.

The round has already started.