Hello {{first_name|Motivated and Miffed Community}},

The AI industry has been very busy spending money it doesn't have, cutting people to fund the spending, and simultaneously launching tools that do the jobs of the people it just cut. This week that loop became impossible to ignore. Three stories landed in the same news cycle, and they all pointed at the same thing: the infrastructure race is now being paid for with headcount — and the agents being built with that infrastructure are getting uncomfortably competent.

Here's what happened.

⚡💸 Building AI is expensive. Someone's paying for it — and it's not the shareholders.

🤖🏛️ Anthropic drew a line with the Pentagon. The DoD didn't like it.

🦀🤖 OpenClaw went viral, got acquired by OpenAI, and spawned an ecosystem in weeks.

🧓🏽🧩 The father of algorithm analysis read Claude's work and wrote "Shock! Shock!" at the top of his paper.

Attio is the AI CRM with conversational AI built directly into your workspace. Every morning, Ask Attio handles my prep:

All in seconds. No searching, no switching tabs, no manual updates.

Ready to scale faster?

Start now →

Oracle is reportedly planning to cut between 20,000 and 30,000 jobs — potentially 12–18% of its global workforce — to free up $8–10 billion in cash flow for AI data center expansion, according to Bloomberg. The same week, Block (Jack Dorsey's fintech company behind Square and Cash App) announced it was cutting roughly 4,000 roles — nearly 40% of its staff. Dorsey said explicitly that AI tools had made those positions redundant.

These are two different stories with the same shape. Oracle's cuts are largely driven by a cash crunch from building AI infrastructure — the company is raising $50 billion in debt and equity and analysts say it won't see positive cash flow until around 2030. Block's cuts are a company saying the quiet part out loud: we don't need as many people because the tools got better. Whether the layoffs are caused by AI or just funded by the AI bet, the message hitting workers is identical. According to research firm RationalFX (via IBTimes), at least 9,200 of the 59,000+ tech jobs cut in 2026 so far are directly attributed to AI adoption — roughly one in five. The rest are at least partially subsidized by AI ambitions.

Why it matters: The cost of building AI and the cost of having AI are now being paid out of the same budget — and that budget is human jobs.

This one has been developing for a few weeks but it landed hard this week. Anthropic and the Department of Defense hit a hard stalemate over how the U.S. military could use Anthropic's AI. Anthropic refused to allow its models to be used for mass surveillance of Americans or to power autonomous weapons operating without human oversight. The DoD cancelled the contract and officially designated Anthropic a "supply chain risk" — a national security label that could block the company from certain government contracts. Anthropic sued.

OpenAI, meanwhile, agreed to a version of the DoD's terms. That agreement triggered what some are calling the "#QuitGPT" movement — reportedly drawing over 2.5 million supporters, with ChatGPT uninstalls surging by 295% overnight according to aggregated data. Claude hit number one on the U.S. App Store for the first time. Sam Altman later amended the contract language, but critics remained skeptical. Microsoft and employees from Google and OpenAI have since filed in support of Anthropic's legal position, arguing (per The Verge) that the "supply chain risk" designation sets a dangerous precedent for the entire tech industry.

Why it matters: AI companies are being forced to decide, publicly, what their tools are for — and users are paying attention to who draws which lines.

According to TechCrunch, February was "the month of OpenClaw" — a vibe-coded AI assistant wrapper built by developer Peter Steinberger that let users talk to models like Claude, Gemini, GPT, and Grok through iMessage, Discord, Slack, and WhatsApp. It went viral almost immediately, spawned spinoff companies (including Moltbook, a Reddit-clone for AI agents later acquired by Meta), ran into privacy issues, and then got acquired by OpenAI — with Steinberger joining the company. The full arc happened in under two months.

Meanwhile, OpenAI launched GPT-5.4 on March 5th — a model explicitly designed for professional multi-step work tasks. In internal benchmarks (the GDPval test of real-world job performance), it achieved an 83% success rate versus 70.9% for its predecessor. It also shipped with native computer-use capabilities — meaning it can navigate software UIs by interpreting screenshots and issuing mouse and keyboard commands on its own. OpenAI also released a ChatGPT add-in for Excel, which I'm sure will make spreadsheet people feel very calm and secure.

Why it matters: Agents that can actually use your computer are no longer a demo — they're in your Excel install now.

Donald Knuth — Stanford computer science professor emeritus, author of The Art of Computer Programming, and widely considered the father of algorithm analysis — published a paper in early March titled "Claude's Cycles." It opens with the exclamation "Shock! Shock!" Knuth had been stuck on a complex graph theory problem (constructing Hamiltonian cycles in a 3D directed graph) while preparing a new volume of his life's work. Claude solved it. Knuth called it a "dramatic advance in automatic deduction and creative problem solving."

Why it matters: When the person who invented the formal study of algorithms is writing papers about being surprised by what AI can do — that's a different kind of signal than a benchmark score.

This week rhymed: the infrastructure costs jobs, the jobs fund the agents, and the agents get good enough to surprise Knuth. That's not a coincidence. It's the loop everyone's been circling around.

Stay MOTIVATED,

Gio