AI is coming for your job.

And your company’s moat.

Or at very least, is the perfect excuse for headcount cuts and short selling.

It’s getting real now, as Block let 4 out of 10 staff go.

Block had become “bloated” (growing headcount from 3,800 in 2019 to 10,000 in 2025), but this isn't simply AI washing.

Jack took to X to say he takes responsibility for building two company structures instead of one. That is the inefficiency. But says, with AI now they’re targeting $2m gross profit per head, 4x higher than 2019.

The stock ripped 24% after-hours. Not because Block is struggling. Gross profit hit $10.36 billion in 2025, up 17%. Q4 gross profit surged 24% to $2.87 billion. They raised 2026 EPS guidance to $3.66 — crushing the $3.22 estimate.

Block fired 4,000 people from a position of strength. And the market loved it. This is a company who’s stock wouldn’t budge despite delivering good results. Now every CEO saw this and learned the market likes the model.

What started 3 weeks ago as a SaaSpocalypse is now meaningfully spreading throughout sectors.

Block built an open-source AI agent called Goose (powered by Anthropic’s Model Context Protocol) and deployed it across the entire company. One engineer says 90% of his code is now written by Goose. Non-technical teams are using it to write SQL queries, close support tickets, and manage inventory without waiting for engineers. 

Block’s CTO told Lenny’s Newsletter it saves employees 8 to 10 hours per week. When you multiply that across thousands of people, you start to understand how a company can look at its org chart and realize half the seats are redundant.

Where are the gains coming from? Coding.

Karpathy wrote a prompt telling an agent to set up a video analysis dashboard: install tools, download models, write code, test it, debug it, set up system services. The agent worked for 30 minutes. Hit problems. Researched solutions. Fixed them. Came back with a report.

Perhaps the poster child for how this is impacting the amount of stuff a company can ship is Anthropic themselves.

If you’re not watching their engineers on X, I’d suggest it. They’re shipping incredible new features daily. Possibly my favorite is a little tool that helps you migrate all of your memory and context from ChatGPT or Gemini to Claude.

Every company now has a higher gross profit per head benchmark. Because extreme productivity is possible. Anthropic is showing the way.

Again this week, we’ve seen them bring scheduled tasks:

So you can get a daily summary of Slack messages, or a personal EA checking for spam and deleting those emails for you.

Then they launched memory for Claude Code, so it remembers what it learns across sessions (not just random facts that appear inappropriately in another chat, like, ahem, some other chatbots).

And that’s missing about 80% of the big features they shipped.

Anthropic is shipping so fast because they can.

Even when people try to kill it.

The recent push by the Department of War to get Anthropic to drop its safety standards appears to have resulted in them becoming more popular than ever. It is now number 1 in the US app store.

And it appears the US military used it during their operations in Iran over the weekend. And what does this have to do with them pushing back on the Department of War?

It attracts and retains the best talent in the industry.

Which in turn, helps them ship fastest.

Tech companies are shipping so fast because they can too.

And if this is what the tech companies are doing today, you can expect the rest of the economy to be heading there soon. And what the tech nerds are doing today will slowly ripple out to the rest of the economy over time.

And it’s not just Anthropic.

Izzy Kaminska called this a “citrini doom loop.”

Citrini Research published “The 2028 Global Intelligence Crisis” — a fictional report from June 2028 describing the S&P down 38% and unemployment at 10.2%. It racked up 28 million views on X. 

And in its wake the software ETF (IGV) hit a 52-week low, erasing all gains since ChatGPT launched. A Substack post. Moved billions. From card networks, software stocks, and anything implicated by AI.

Citrini never called it a doom loop. But that's what it is.

The loop:

Then Block went and did exactly this. On earnings day. Stock ripped 24%.

But is the doom loop real economics, or a narrative that creates its own reality? Or both? 

It’s largely agreed that Anthropic has won enterprise, but the new OpenAI funding round could change that.

Not to be outdone. OpenAI casually completed a $110bn mega-round at a $730bn pre-money valuation.

This dwarfs the largest IPO in history (Saudi Aramco at $29bn), and comes with strings attached. OpenAI agreed to run on AWS Bedrock infrastructure. Bedrock becomes the exclusive place to use OpenAI’s enterprise governance stack. Amazon gets to sell picks and shovels to both sides of the AI war.

$110 billion is a black hole for capital. That money comes from somewhere. Investors are dropping positions in traditional sectors to chase AI, accelerating the very sell-off the doom loop describes. 

As OpenAI becomes much more capable of displacing knowledge work, and going deeper into enterprise, the trend accelerates.

The round itself is part of the mechanism.

Citadel Securities macro strategist Frank Flight pointed to Indeed data showing software engineer job postings up 11% YoY in early 2026. 

The St. Louis Fed found daily use of generative AI at work is “unexpectedly stable” with “little evidence of imminent displacement risk.” New business formation in the US is expanding. Data center construction is driving localized hiring booms.

The historical argument is strong. 

Productivity booms have never caused long-term unemployment spikes. Keynes predicted mass technological unemployment in the 1930s. It didn’t happen. There used to be a job called a “computer,” large rooms full of people who did calculations for NASA and the US military. These people were ultimately “replaced” by actual computers. 

But NASA still hires plenty of nerds. The frontier moved.

But zoom out from Silicon Valley.

Entry-level roles in retail, hospitality, and logistics are shrinking. The graduates entering the job market right now are competing against tools that can do junior-level knowledge work for $20 a month.

Anthropic’s CEO Dario Amodei said AI will wipe out 50% of entry-level lawyers, consultants, and finance jobs in 1 to 5 years.

Do you buy that?

I actually do.

A lawyer recently shared “The Claude-Native Law Firm” — Zack Shapiro’s viral thread about running a two-person boutique competing against firms with thousands of lawyers, powered almost entirely by Claude.

Look at this line: 

That’s an iMessage. From a real lawyer. Casually confirming that AI has replaced junior staff in their practice. Harvey AI went from $50M to $190M ARR in a single year.

When Anthropic launched its legal plugin for Claude Cowork, it wiped $285 billion from legal tech stocks in a single day. Thomson Reuters fell 16%. LegalZoom down nearly 20%.

The AI doom loop. In real time. In one profession.

Maybe it’s a third thing.

Markets are reflexive. 

Block’s 24% stock pop is a clear signal: Wall Street wants you to fire people and replace them with AI. If you don’t, your competitors will, and the market will punish you for hesitating.

Gross profit per employee will become the defining measure of how well a company is using AI.

Individual companies may have less staff, but their productivity will have to increase. 

The doom loop might not need to be “true” in the classical economics sense to be real. It just needs enough people to believe it, act on it, and create the conditions it describes.

Which, come to think of it, is how most of economics works anyway.

So what does any of this mean for you?

Get good now.

My belief is that we’re all software engineers now.

And software engineers can do design, designers can do code, product people can do some mix of all of it.

In this world the ability to do the more complex stuff, production grade code, and run teams of agents becomes a rare skill and highly sought after. This means each head can do more, but needs to know how to do more.

Personally, I’ve become WAY more productive. 

I was always fast — I could write fast, make decks fast. 

But now I’m a Dad with multiple companies and a compelling job at Tempo. AI has become existentially important to my ability to do all of this. 

(Don’t get me wrong, I shout at claude code daily, and I still feel bottlenecked, and outrun by others who are far better. But the point is the tools work).

What’s happening in tech now will come for other sectors. It will just take a while.

Ask yourself if you’ve become wildly more productive since December?

The AI agent coder is your new customer, and this customer has preferences.

How do you make AI pick your product? This tool leaderboard tells you what AI already prefers.

I’d be obsessed with this chart and reverse-engineering why the AI is using these tools. 

If I had to speculate about why these tools are winning, it’s the fact that the tools had existing human engineer support, clean, readable docs, and have taken the steps to make their docs and tools more accessible to AI Agents (including LLMs.txt, MCP servers and agent-friendly tool calling).

In a handful of months a chunk of your company could look like this:

Nvidia stock is falling. 

The market can see a ceiling even if the builders can't. What happens when all of the money in the world isn't enough to keep the compute buildout going? The doom loop assumes AI gets better and cheaper. But the capital required is astronomical. If the investment thesis cracks, the layoffs don't reverse. They accelerate.

I’ve never seen a Substack post move markets like this. I’ve never seen a stock rip 24% on news that 4,000 people lost their jobs. And I’ve never seen so many anecdotes about displacing junior jobs in companies that don’t match the official jobs numbers.

Something is different this time. Maybe not in the way Citrini describes. Maybe not on their timeline. But the signal is there, and it’s getting louder.

The doom loop might have a brake. I just can’t see it yet.

ST.