AI stopped being theoretical this week

Jack Dorsey fired 4,000 people Thursday. His reason: AI makes them unnecessary. Block’s stock surged 24%. Then the Pentagon blacklisted Anthropic for refusing to build autonomous weapons—and OpenAI signed the deal Anthropic wouldn’t. By Monday, Claude crashed under record demand and hit #1 on the App Store.

This wasn’t a normal week. This was the week AI became a power story. Here’s what you need to know—and do.

In Today’s 5-Minute AI Digest. You will get:

1. The MOST important AI News & research
2. AI Prompt of the week
3. AI Tool of the week
4. AI Tip of the week

…all in a FREE Weekly newsletter. 

Jack Dorsey cut 4,000 employees—the largest AI-attributed layoff in S&P 500 history—and predicted most companies will follow within a year. Block’s stock surged 24% overnight. Whether genuine transformation or rebranded cost-cutting, a template now exists: cite AI, slash headcount, get rewarded.

Anthropic refused to let Claude power mass domestic surveillance or autonomous weapons. Trump ordered all federal agencies to cut ties within six months. Hegseth designated Anthropic a “supply chain risk”—a label never before applied to an American company. Hours later, OpenAI signed the Pentagon deal Anthropic wouldn’t, claiming the same red lines.

One Anthropic feature launch. CrowdStrike down 10%. Cloudflare down 8%. The Global X Cybersecurity ETF at its lowest since 2023. Claude Code Security scans codebases for vulnerabilities using reasoning, not rules—and found 500+ bugs in production open-source projects that decades of human review missed. Cybersecurity is now the third sector AI has rattled in a month.

Anthropic’s stance against the Pentagon turned into marketing gold. Claude topped the App Store as users rallied behind the company. Free users surged 60% since January; paid subscribers doubled since October. Then Monday, the servers buckled under “unprecedented demand.” Principles, it turns out, sell.

Everyone’s covering the drama. Here’s what they’re missing.

In three weeks, Anthropic’s Claude Cowork wiped $285B off SaaS stocks. Claude Code Security crashed cybersecurity equities. Then Anthropic defied the President—and gained more users than any marketing campaign could buy. The company is simultaneously destroying the old software business model and becoming the most trusted AI brand in America.

Here’s the 90-day prediction: Anthropic will announce an enterprise product that directly replaces at least one major SaaS category—likely legal research or financial data. The Pentagon standoff will get resolved in court, and Anthropic will be back on classified networks by Q3 with stronger terms. And at least two more S&P 500 companies will announce Block-style “AI restructuring” using Claude as their platform of record.

The bottom line: AI vendor choice is no longer just a technical decision. It’s a political one, a business continuity question, and increasingly a values signal. When the Pentagon can blacklist your AI provider overnight, multi-vendor strategy isn’t a best practice—it’s survival AI Prompt of the Week

[Read the full story]

Stress-test your role against the Block playbook—before your CEO does it for you.

Prompt:

I lead [department] with [X] people. Our recurring tasks:

[list 5–10 tasks]

For each task, give me:

Why it works: Task-level analysis beats role-level panic. You’ll find that the tasks most at risk are the ones you probably don’t enjoy anyway. The tasks that survive—judgment, stakeholder alignment, creative direction—become your competitive moat.

Real result: A marketing director used this and found 4 of 8 core tasks were automatable. She restructured her team around the remaining four before her CEO read the Block news. She kept her headcount. Her peer in sales didn’t.AI Tool of the Week

What it is: A single gateway to every major AI model—Claude, GPT, Gemini, Llama, Mistral—with automatic failover when one goes down.

Why this week: Claude crashed Monday. The Pentagon blacklisted Anthropic Friday. Single-vendor AI dependency broke twice in four days. OpenRouter users switched models in seconds.

Pitch: “It’s like having backup generators for your AI.”

Rating: ⭐⭐⭐⭐⭐

Automatic fallback routing when your primary model goes down

Pay-per-use across all models—no separate subscriptions

Playground mode lets you compare model outputs side-by-side—no coding required

Developers: change one URL in your existing code to get started

Non-technical? Use the Playground at openrouter.ai/playground to test prompts across models in your browser. Zero setup. Find which model works best for your specific tasks—then you’ll know exactly where to go when your main AI is down.

Link: openrouter.ai

The tip: For every AI-dependent workflow, write a one-page fallback doc. Not a full replacement—just enough to keep moving when your provider goes down, gets blacklisted, or changes terms overnight.

What it looks like (steal this template):

WORKFLOW: Weekly client report generation

PRIMARY TOOL: Claude via API

FALLBACK 1: Gemini (account: [login], same prompt works)

FALLBACK 2: Manual template in Google Docs [link]

SWITCH TIME: <5 minutes

DATA NEEDED: CRM export (CSV), last week’s metrics

OWNER: [Name] | LAST TESTED: [Date]

Why it works: Monday’s Claude outage and Friday’s Pentagon ban happened four days apart. The teams that recovered fastest weren’t running the best AI—they’d already answered “what happens when this breaks?” Resilience beats optimization when the landscape shifts weekly.

Limitations: This works for AI as a productivity layer (most of you). If AI is your core infrastructure—like Block’s internal tool Goose—you need architectural redundancy, not a doc.

Pro move: Create the fallback doc for your three most critical AI workflows. Then test one. Actually switch to your backup for a full task cycle. If you can’t complete it, you don’t have a fallback—you have a wishlist. Fix it now while you have time to choose.

AI layoffs are no longer theoretical—Block’s 40% cut and 24% stock surge just wrote the playbook every CEO will study

Your AI vendor is now a geopolitical risk—Anthropic was blacklisted on Friday, Claude crashed on Monday, and cybersecurity stocks got torched in between

Single-vendor AI dependency is a liability—build fallbacks now, while you still have time to choose them

Now implement one.

Most readers will skim this and move on. The ones who run the displacement audit before Friday’s standup, set up OpenRouter’s playground during lunch, or write their first fallback doc before they close this tab—those are the ones who won’t be scrambling when the next Block memo drops. And after this week: it’s coming.

Stay curious—and stay paranoid.

— R. Lauritsen

P.S. Know someone whose company just announced “AI restructuring”? Forward this issue. The displacement audit might save their next quarterly review.

P.P.S. We are building a deep dive on the Anthropic-Pentagon timeline—the full story, the legal angles, and what it means for enterprise AI procurement. Reply “PENTAGON” if you want it.