45,000 AI Layoffs. Zero Displacement

Jensen Huang just unveiled chips “the world has never seen before” to 39,000 people in a hockey arena. Tesla announced a $25 billion chip factory launching this week. The Pentagon blacklisted one of America’s top AI companies for refusing to build autonomous weapons. And here’s the twist: Anthropic’s own research says AI hasn’t actually displaced any workers yet.

Here’s what you need to know—and do—this week.

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. 

Jensen Huang’s GTC keynote revealed the Vera Rubin architecture—NVIDIA’s successor to Blackwell—delivering a 5x inference throughput boost with HBM4 memory and liquid-cooled, rack-scale NVL144 systems. Samples ship to cloud providers by late 2026, with full production in early 2027. Huang called it “the new unit of compute for the modern enterprise.”

Source: NVIDIA Blog

The Trump administration designated Anthropic a supply chain risk—a label normally reserved for foreign adversaries—after CEO Dario Amodei refused to let Claude power autonomous weapons or mass surveillance. Anthropic filed two federal lawsuits. Dozens of rival employees at OpenAI and Google filed amicus briefs in support. Claude downloads surged past ChatGPT on the App Store the next day.

Source: CNBC / NPR

The Jira maker slashed 10% of its workforce and replaced its CTO with two “next-gen AI talent” leaders. This follows Block cutting 4,000 jobs in February. Global tech layoffs in 2026 have surpassed 45,000. Atlassian’s stock is down over 60% in a year as investors fear AI agents will make traditional SaaS subscriptions obsolete.

Source: TechCrunch / Bloomberg

Elon Musk confirmed Tesla’s chip fabrication facility launches this week—targeting 2nm process technology and 100–200 billion custom chips annually for Full Self-Driving and Optimus robotics. If it works, Tesla joins a handful of entities on Earth producing frontier AI silicon in-house. Jensen Huang publicly questioned whether they have the decades of yield expertise required.

Source: Reuters / FinTech Weekly

While companies cite AI to justify layoffs, Anthropic’s new research paper tells a different story. Their “observed exposure” framework—combining theoretical AI capability with real-world usage data—found no systematic increase in unemployment for highly AI-exposed workers since late 2022. The catch: younger workers in exposed occupations show early signs of slower hiring. The displacement may be coming, but it hasn’t arrived yet.

Source: Anthropic Research

Beneath every headline this week is one story: the AI war has moved from software to silicon.

NVIDIA isn’t selling chips anymore—it’s building “AI factories” measured in gigawatts. Tesla is spending $25B to fabricate its own processors rather than wait for anyone’s roadmap. And the Pentagon just proved that governments will weaponize supply chain access to decide which AI companies survive.

Here’s what most coverage misses: the industry is splitting into those who own physical infrastructure and those who rent it. That’s your new risk metric—not which model is smartest, but where your AI stack’s silicon comes from.

Prediction: Within 18 months, at least three more major AI companies will announce in-house chip programs. The companies that own the atoms—not just the algorithms—will set the terms for the next decade.

🔗 Go deeper: If value is concentrating in the physical layer, where is it draining from? This week’s deep dive goes inside the SaaSpocalypse—the $2 trillion wipeout in enterprise software stocks, who’s most exposed, who’s pivoting, and the moves you should make before Q2 earnings.

Read the full deep dive →

Turns any complex decision into a structured risk assessment with a built-in asymmetry detector.

Why it works: The “reversal cost” lens is the key move. It shifts the model’s analysis from “what’s the best outcome” to “what’s the safest bet with the most upside”—something humans rarely do naturally when listing pros and cons.

The transferable principle: Any prompt that introduces an asymmetric evaluation axis (like reversal cost vs. upside) produces sharper analysis than a symmetric pros-and-cons list. Apply this pattern to hiring decisions, vendor selection, or product roadmap calls.

A command-line tool that lets AI agents read your Gmail, search Drive, manage Calendar, and edit Docs—through a single interface.

Why you need it: Google quietly dropped this open-source CLI on March 5. It hit 17,000 GitHub stars in five days. One install command gives AI agents structured access to your entire Workspace—no custom API wiring. It even includes a prompt injection defense layer (Model Armor) that scans Workspace data before it reaches your AI.

One-liner: “The shortcut that turns your Gmail, Drive, and Calendar into an AI agent’s toolkit—in one npm install.”

Rating: ⭐⭐⭐⭐ (4/5) — Incredibly useful, but not officially supported by Google yet. Use in production at your own risk.

Key features:

Best use case: Building AI agents that need to triage email, search files, or manage calendars without stitching together five separate OAuth flows.

Link: github.com/googleworkspace/cli

After any AI analysis or recommendation, add one line: “Now give me the strongest possible argument against your own conclusion.”

Why it works: LLMs have a confirmation bias problem—they tend to reinforce whatever direction you steer them. The Steel Man Close exploits the model’s ability to argue any position by explicitly asking it to attack its own output. This surfaces risks, assumptions, and blind spots that the initial analysis glossed over. You get a built-in adversarial review without needing a second opinion.

Limitations: Works best on analytical and strategic tasks. For creative work (writing, brainstorming), it can kill momentum. Don’t use it mid-draft—use it after you have a recommendation to stress-test.

Pro move: Chain it with the Decision Landscape prompt above. Run the full landscape first, then Steel Man the top recommendation. You’ll get analysis that rivals a consulting engagement—for the cost of two prompts.

You just learned:

Now implement one.

Most readers will skim this and move on to their next tab. The ones who install one CLI or restructure one decision prompt this week will be the ones their team turns to when the infrastructure questions start hitting their inbox.

Reply with which move you’re making first. I read every response.

— R. Lauritsen

Share iPrompt: [insert share link]

P.S. If this issue made you think, forward it to one colleague who’s still sleeping on the silicon shift. They’ll thank you later.

P.P.S. Haven’t read the deep dive yet? It’s the other half of this week’s argument—$2 trillion in SaaS market cap gone, which companies are pivoting fastest, and the exact moves to make before Q2 earnings. Link is in Our Angle above.