I posted this on Tuesday night:

4,200 likes. 227 retweets. My DMs exploded.

Not because I said something controversial. Because the market just put a price tag on what I've been writing about for months.

January 30th: Anthropic launches a plugin marketplace for Claude Cowork.

It wasn’t even an official product announcement, just a side note on a technical documentation page.

But here’s what it does: Contract review. NDA triage. Compliance workflows. Legal briefings.

The kind of work lawyers bill $400/hour for.

The kind of work Thomson Reuters built a $60 billion business around.

Four days later, Tuesday, February 3rd:

The software sector dropped 13% in six sessions. Steepest decline since March 2020.

$300 billion in market cap. Gone.

About a year ago, DeepSeek spooked the market by building AI models for under $6 million. Nvidia dropped 17%.

That selloff hit the sellers of AI infrastructure.

This selloff hit the buyers of software. The companies that charge subscriptions for things AI can now do itself.

DeepSeek said: "Maybe AI is cheaper to build."

Anthropic said: "Here's an AI that does your lawyer's job."

One threatens supply. The other threatens demand.

I've been skeptical of "AI agent" hype.

Most of it is marketing slop that breaks in production. Demos that look impressive on Twitter but fall apart when you try to use them for real work.

This week I finally set up Claude Cowork properly. Connected my Google Drive. Gave it access to my content folder — newsletters, drafts, strategy docs, everything I've written for the past year.

Something shifted.

For months, my biggest frustration with AI was context. Start a conversation, explain my business, explain my voice, explain what I've already written — then the context window fills and we start over.

Like that Tom Cruise movie Edge of Tomorrow. Respawn. Repeat. Dread.

I even put instructions in my project memory: "Call me out if I'm going back to building tools instead of focusing on distribution."

But then I'd be working on internal tooling, and Claude would flag it, and I'd have to explain — again — and again — and again, that this is internal, here's the reasoning.

After the fifth correction, I felt defeated. We're having the same conversation for the fifth time.

This week was different.

Cowork read my previous newsletters. Understood my content pillars. Knew what I'd already covered. It saw the pattern in my writing. It understood my audience.

We organized my entire content system in an afternoon. Built an inventory of every newsletter I've published since October. Identified gaps in my pillar rotation. Set up tracking files. Created templates.

The kind of work that would've taken me a full day with an assistant. Done before lunch.

I'm writing this newsletter with Cowork right now. It pulled the market data. Organized the Perplexity research. Created the charts you see in this email.

I'm writing the words — the voice is mine, the opinions are mine — but the context layer? The research layer? That used to require a person. Or hours of my own time.

This isn't a chatbot anymore. It's a collaborator that actually knows my business.

What I experienced this week isn't unique to me. It's the preview of what's coming for everyone — companies, startups, employees, even governments.

The shift isn't "AI can answer questions." We've had that for two years.

The shift is "AI can work alongside you with context."

That's a fundamentally different product. And it changes habits everywhere.

The old model: Hire specialists. Build departments. Scale headcount with revenue.

The new model: One person with AI does what five people did before. The team of 5 startup competes with the 100-person plus company.

I run my content operation with four people (quite frankly just two of us works on content). We're outproducing teams with twenty. Not because we're smarter. Because AI handles the work that used to require junior staff.

If you're a startup founder still hiring like it's 2019 — researchers, analysts, coordinators, junior anything — you're burning money. The winners in 2026 are the ones who figure out how to stay lean while AI handles scale.

This is where it gets uncomfortable.

Big companies have layers. Middle management. Coordinators. People whose job is to move information between other people.

AI doesn't need coordinators. AI doesn't need someone to summarize the meeting and send it to the other team. AI doesn't need a project manager to track status updates.

The companies that adapt will flatten. Fewer layers. Smaller teams. More output per person.

The companies that don't will get outmaneuvered by startups that never had the bloat to begin with.

The 46,000 layoffs on last Tuesday — Amazon, UPS, Pinterest — weren't random. They were the first wave of companies realizing they can do more with less. I can’t believe it was just last week from my last newsletter.

The uncomfortable truth: If your job is primarily moving information around, you're exposed. If your job is making decisions that require judgment, relationships, and accountability, you're safer.

The people who thrive will be the ones who use AI to multiply their output. Not "I play with ChatGPT sometimes." Actually integrate it into every workflow.

The gap between "AI-native" workers and everyone else is about to become the gap between employed and unemployed.

Governments are the slowest to adapt. That's usually fine. This time it's dangerous.

The countries that figure out AI-native public services — licensing, permits, tax filing, benefits administration — will have a massive competitive advantage. The ones that don't will watch their best talent leave for places that move faster.

Estonia did this with digital government a decade ago. Now everyone wants to copy them. The same thing is about to happen with AI-native governance.

The governments still running on paper forms and six-week processing times? They're about to look as outdated as a fax machine.

I see government programs months ago that are helping businesses adopt AI with grants became almost entirely useless today. The speed of everything is that fast.

Not everyone's at equal risk.

High risk: Companies that sell access to information or automate knowledge work.

Thomson Reuters charges $15,000/year for Westlaw. If AI can read case law and draft briefs, why pay?

Intuit charges for TurboTax. If AI can categorize expenses and file returns, why pay? (I for one has not been using TurboTax for the last 2 years)

ServiceNow charges per seat for workflow automation. If AI can triage tickets and route requests, why pay?

Lower risk: Companies with proprietary real-time data that doesn't exist elsewhere.

But even they're not safe. If AI can synthesize information from multiple free sources, the premium for any single source shrinks.

Here's what nobody's talking about.

Ares, KKR, Blue Owl — all dropped 9%.

Why? Because private equity loaded up on software.

"Recurring revenue. High margins. Sticky customers." The perfect buyout target.

They bought these companies with borrowed money. Those loans now sit in Business Development Companies.

Barclays estimates software is 20% of BDC portfolios — roughly $100 billion in exposure.

Apollo's CEO already said he wants to cut software exposure from 20% to under 10%. He reportedly shorted bonds of several software companies.

UBS warned that if AI triggers "aggressive disruption," default rates in private credit could hit 13%.

This isn't just a stock market story. It's a $100 billion credit bubble.

Stop reading. Open your calendar. Block 2 hours this week.

Go to Claude or ChatGPT. Describe what your product does. Ask: "Build me a basic version of this."

If it can produce 70%+ of your functionality in an afternoon, you have 12-18 months before someone ships a free alternative. Not "maybe." When.

Action: Do this test by Friday. Write down what AI couldn't replicate. That's your real moat — or the signal you don't have one.

If you charge per seat, per transaction, or for access to information — you're exposed.

AI doesn't need seats. AI doesn't care about transaction volume. AI makes information free.

Action: List your top 3 revenue lines. For each one, answer: "Could AI do this for my customer directly?" If yes, you need to change your pricing within 6 months.

AI can execute. AI can't decide what matters. AI can't own the outcome. AI can't build trust over years.

Action: Write down the 3 things your best customers pay you for that require human judgment, relationships, or accountability. Double down there. Everything else is commoditizing.

Not a survey. Actual conversations.

Ask: "If you could get 80% of what we provide from AI for free, what would you still pay us for?"

Their answer is your product roadmap for 2026.

Action: Schedule 5 customer calls in the next 14 days. Not sales calls. Research calls. Listen more than you talk.

This section might be uncomfortable. Read it anyway.

Be honest. What percentage of your job is information gathering, document creation, data analysis, scheduling/coordination, or repetitive decisions with clear rules?

Add those up. If it's over 50%, your role is changing in the next 18 months. Not "might change." Is changing.

Action: Write this number down. Don't lie to yourself. The people who get ahead are the ones who see clearly.

What do you do that requires judgment calls with incomplete information, building relationships over time, taking responsibility for outcomes, making decisions your boss doesn't want to make, or navigating politics?

Action: Write 3 bullet points. These are what you emphasize in every performance review, every 1:1, every conversation about your role. If you can't write 3, that's the problem.

Not "play with ChatGPT." Actually integrate it into your work.

The gap between "people who use AI daily" and "people who use it occasionally" is about to become the gap between "employed" and "looking."

Action: Pick ONE workflow you do weekly. Automate or augment it with AI by end of February. Time yourself before and after. Show your boss the improvement.

Any questions with using it? Feel free to ask me, send me an email. I will reply.

I don't care if it's a newsletter, a consulting practice, a small product, or a service business.

The best insurance against job disruption isn't another job. It's income you control.

Action: Commit to one thing you'll build outside work. Give yourself 90 days to make your first dollar from it. Not "think about it." Start.

Ask directly: "How do you see AI changing our team in the next 12 months?"

If they say "it won't" — start looking. They're either lying or asleep.

If they have a real answer — you know where to position yourself.

Action: Schedule this conversation within 2 weeks. Better to know than to wonder.

$300 billion disappeared because Anthropic shipped some plugins.

Not because the plugins were revolutionary. Because they proved that the moat is gone.

Information wants to be free. Automation wants to be built-in. Software wants to be a commodity.

Jensen Huang called this selloff "the most illogical thing in the world."

He's wrong (or filled with self-interest). The market isn't being irrational. It's being early.

The companies that survive will figure out what AI can't do — and charge for that.

The people who thrive will stop competing with AI and start using it.

I wrote about this months ago. The market just caught up.

Now the question is: will you?

I wrote this newsletter with Claude Cowork helping with research. The words are mine. The opinions are mine. But the workflow is different now.

If you're still using AI the way you used Google - asking questions and getting answers - you're missing the shift. It's not chat. It's agents that understand your context and work alongside you.

We're not all the way there yet. But we're closer than most people realize.

George