Sponsored by
Today on FindMeAI:
This week the AI economy stopped being a story about potential and became a story about money — who's printing it, who's burning it, and who's losing their job to it.
→ OpenAI filed for IPO — at up to $1T, while bleeding $14B a year. Anthropic just turned its first profit.
→ Gemini Omni shipped — physics-aware video from any input, free on YouTube this week
→ Meta cut 8,000 — and the first government data on AI job loss landed the same week
→ The bills are coming due — Anthropic meters agents June 15, and a poisoned extension breached GitHub
OpenAI filed to go public this week. Anthropic quietly turned its first profit. Only one of them is making money.
On Friday, OpenAI confidentially filed its S-1 with the SEC, targeting a Q4 2026 listing — possibly September — at a valuation between $852B and $1 trillion. Goldman Sachs and Morgan Stanley are leading. The numbers underneath are split-screen:
OpenAI | Anthropic | |
|---|---|---|
Revenue run rate | ~$25B annualized | $10.9B projected Q2 (↑130%) |
2025/26 bottom line | ~$9B loss in '25, ~$14B operating loss projected '26 | First operating profit: +$559M in Q2 |
Compute cost per $1 revenue | Still underwater | 71¢ → 56¢ in one quarter |
OpenAI is generating roughly $2B/month with 50M consumer subscribers and 9M business users — and still burning cash to do it. Anthropic told investors (via a WSJ report) that enterprise demand, led by coding and cybersecurity, is what flipped it into the black. It cautioned profit may not hold all year as compute spend climbs.
The same week, OpenAI co-founder Andrej Karpathy joined Anthropic's pre-training team — using Claude to accelerate the research that builds future models.
Why this matters for builders: the platform you build on is about to have a public stock price and quarterly earnings pressure. Profitability is becoming the new credibility. Watch whose unit economics actually work — that's who sets your pricing in 2027.
Gemini Omni turns a photo into physics-aware video — and it's free on YouTube Shorts this week.
Google shipped Gemini Omni Flash at I/O on May 19. Feed it any mix of text, image, audio, or existing video; get coherent video back, with physics that mostly hold up. The differentiator isn't raw quality — it's that you edit by conversation: "make the background sunset," "have her turn around," and each instruction builds on the last while characters stay consistent.
The honest scorecard:
→ Quality: Seedance 2.0 still tops the Artificial Analysis leaderboard (Elo 1,269 text-to-video), ahead of Veo 3.1 and Sora 2. Omni's clips can look slightly over-smooth on big screens.
→ Length: caps at 10s per clip (Google calls it a deployment choice, not a limit).
→ Access: live now for AI Plus/Pro/Ultra globally, free this week on YouTube Shorts + Create.
→ Catch: no public API yet — you can't build on it programmatically.
So this isn't an automation play yet. It's a creation play. If you make content — product demos, ad variations, explainer clips — you can prototype a full visual concept this weekend without touching a video editor or paying per render.
The move: stop treating AI video as a someday tool. The marginal cost of a first draft just hit zero. Make three this weekend.
Meta cut 8,000 jobs.
On May 20, Meta laid off ~8,000 employees (10% of staff), reassigned 7,000 to AI initiatives, and canceled 6,000 open roles. Zuckerberg's memo: "success isn't a given." That's 30,000+ cut since 2022, against a $125–145B capex plan. Intuit cut 17% the same week.
Then the receipts arrived. The Bureau of Labor Statistics tracked 18 AI-exposed occupations — roughly 10 million jobs — and found they fell 0.2% from May 2024 to May 2025, while overall employment grew 0.8%. Strip out medical secretaries (rising on healthcare demand) and the other 17 categories dropped 1.6% — the second straight year of decline.
Where it's hitting hardest (since May 2022):
→ Credit authorizers, checkers, clerks: −26.2%
→ Broadcast announcers / DJs: −20.8%
→ Sales engineers: −13.2%
→ Customer service reps: −130,180 workers in a single year (−4.8%)
This is the counterweight to every funding-round headline. The capital is concentrating; the entry-level and mid-tier roles in exposed categories are thinning.
The implication: the safe move is no longer "a stable role in an exposed function." It's becoming the person who builds or directs the systems doing the replacing. If your job is on that list, this is your signal to move up the stack — not next year, now.
The bills are coming due: agents get metered, and your VS Code setup just became an attack surface.
Two stories every builder should act on before Monday.
→ Anthropic is putting Claude agents on a meter (starts June 15). Programmatic usage — Agent SDK, GitHub Actions, third-party frameworks — gets split from your chat limits into dedicated monthly credits: Pro $20, Max 5x $100, Max 20x $200. If you run Claude in production pipelines, model your real consumption now or get surprised on the first invoice.
→ A poisoned VS Code extension breached GitHub. On May 18, a compromised Nx Console build (v18.95.0, 2.2M installs) was live for ~18 minutes. In that window, it ran a credential stealer harvesting tokens from GitHub, npm, AWS, Vault, Kubernetes, and 1Password — exfiltrating via HTTPS, the GitHub API, and DNS tunneling, plus a persistent macOS backdoor. Result: ~3,800 internal repositories exfiltrated. Threat actor: TeamPCP.
Two more worth a look this week:
→ OpenAI Codex Goal mode (now GA) — define an outcome + success criteria, let it work toward it across CLI, IDE, and app. Best for: long-running refactors.
→ Antigravity 2.0 (from I/O) — standalone agent desktop app with cron, subagents, and JSON hooks. Best for: scheduled, async agent workflows.
The pattern: agents are moving from free experiment to metered infrastructure — with a real security perimeter. Treat your dev environment like production.
Do These Before Microsoft Build (June 2)
The AI economy got real this week. Here's how to respond before the next wave. Pick two:
☐ Audit your VS Code extensions — Remove anything you don't actively use. Rotate any tokens stored in your dev environment. 20 min. Non-negotiable after the Nx breach.
☐ Make 3 videos with Gemini Omni — Open YouTube Create or the Gemini app. Turn a product photo into a clip. Edit it by chat. 30 min. Free this week.
☐ Model your Claude agent costs — If you run agents in production, estimate your June usage against the new credit tiers before the June 15 switch. 15 min.
☐ Read OpenAI's IPO context — Skim Fortune's breakdown of the $1T filing. Understand the platform you're building on. 10 min.
Don't bookmark. Don't "save for later." Pick two. Start today. →
Stop Paying for 6 Tools. One AI Does It All
Most e-commerce sellers are running their store across 6 to 8 separate tools — and paying hundreds of dollars a month for the privilege. StoreClaw replaces your entire stack with one autonomous AI engine that monitors competitors, optimizes listings, automates marketing, and tracks real profit across Shopify, Amazon, and beyond.
It doesn't wait for you to ask. It runs 24/7 in the background, so you wake up to a full dashboard instead of a list of things you forgot to check.
Connect your store, and StoreClaw gets to work — no prompts, no complex setup, no six-app stack.
Free to start. No credit card required.
Forward this to one builder who still has a sketchy VS Code extension installed and doesn't know it.
Reply with one word: OpenAI, Anthropic, and Google — whose unit economics do you actually trust to set your pricing in 2027?