Daily TEA: Anthropic+OpenAI Eat 89%
Saudi $12.5B tokenization, AI revenue concentration, side doors over apps, arXiv bans AI slop, Malta goes free ChatGPT, RTK token killer
Hello, dear TEA-mates! Here is what you need to know today.
1. 🪙 Saudi Arabia Pushes $12.5B Tokenization Drive, Eyes Late-2026 Stablecoin Property Settlements
Saudi Arabia is advancing a $12.5 billion tokenization initiative led by Faisal Mannai, head of droppRWA and a key architect of the Kingdom’s electronic payment infrastructure. The project targets real estate, energy, and industrial assets, and recently executed its first real estate deed transfer via blockchain, collapsing transaction times from days to milliseconds. Stablecoin-based property settlements are projected to go live by late 2026 in coordination with the Capital Market Authority and Saudi Central Bank, letting developers receive cross-border capital in minutes instead of days. The broader plan integrates stablecoins and tokenized assets into a sovereign network by 2030. Tokenized US Treasuries hit a record $15.5 billion in May 2026, and the global stablecoin market now exceeds $300 billion. (Read More)
🫖 TEA For Thought: “Sovereign-grade tokenization at this scale is no longer theoretical. When a state economy moves on-chain, the rest of the region follows.”
2. 💰 AI Startup Revenue Hits $80B, But Anthropic and OpenAI Take 89% of It
34 AI startups now generate nearly $80 billion in collective annual revenue, up 112% in just six months according to data from The Information. Anthropic and OpenAI control 89% of that total, with Anthropic recently surpassing OpenAI on the strength of AI coding tools. Perplexity, ElevenLabs, and Cognition each clear $500 million, but everyone else is small. The economics under the hood are brutal: Anthropic shares revenue with Amazon and Google, OpenAI remits 20% to Microsoft through 2030, and the two combined burn over $30 billion annually on training. Sequoia and other investors frame it bluntly: most of the value sits with the model makers, not the app companies built on top. (Read More)
🫖 TEA For Thought: “Well, tons of companies are just Claude wrappers or another harness layer.”
3. 🚪 Velvet Noise: In an AI-Saturated Job Market, Side Doors Beat Applications
Writer Maja published a Velvet Noise essay on May 15, 2026 arguing that traditional front-door job applications no longer carry signal in a market flooded by AI-polished submissions. Her thesis frames hiring as a search problem and says specific, public, and direct outreach now wins. The tactics she lays out: research individuals and companies thoroughly before contact, create small undeniable proof units (memos, essays, teardowns, technical guides), build in public to become discoverable, and send earnest, specific cold messages rather than templated ones. Her core reframe: “A job is a bundle of problems someone wants solved badly enough to pay another person to solve them.” The deeper point is that AI has compressed the value of generic applications to near zero, making specificity the only remaining moat for candidates. (Read More)
🫖 TEA For Thought: “The application stack is dead. Build something visible, then knock on the side door.”
4. 📄 arXiv Imposes One-Year Ban for AI-Slop Papers
Thomas Dietterich, chair of arXiv’s computer science section, announced on X on Thursday evening, May 15, 2026 that arXiv will impose a one-year ban on authors who submit papers containing “incontrovertible evidence that the authors did not check the results of LLM generation.” Telltale signs include hallucinated references, plagiarized or biased content, and meta-comments left in the manuscript like “here is a 200 word summary; would you like me to make any changes?” or “fill it in with the real numbers from your experiments.” After the ban expires, future submissions must first be accepted at a reputable peer-reviewed venue before arXiv will accept them again. The policy does not prohibit AI use outright. It places full responsibility on the authors regardless of how the content was produced. (Read More)
🫖 TEA For Thought: “Finally. A real bar. Use the tools, just check the output before you ship.”
5. 🇲🇹 OpenAI Gives Every Malta Resident One Year of ChatGPT Plus Free
OpenAI and the Government of Malta announced on May 16, 2026 a first-of-its-kind national partnership giving every Maltese resident one free year of ChatGPT Plus (normally €23 per month). Residents must first complete a free online course called “AI for All,” developed by the University of Malta, before applying via Malta’s online identity system. The offer also extends to Maltese citizens living abroad. Microsoft is co-partnered: participants can alternatively pick Microsoft 365 Personal Copilot (valued $10 to $39 per month). The deal is the latest under OpenAI’s “OpenAI for Countries” initiative, which targets governments moving from interest to strategic national AI adoption. Malta is the first country worldwide to receive this scope of access. (Read More)
🫖 TEA For Thought: “National-scale AI distribution with a literacy course attached. Smart wedge. Expect more countries to follow.”
🛠️ Skill of the Day
rtk-ai/rtk — Rust Token Killer. A single-binary CLI proxy that intercepts dev commands (git, cargo, docker, 100+ supported) and applies smart filtering, grouping, truncation, and deduplication before output reaches your AI assistant. A 30-minute Claude Code session shows roughly 80% token reduction (118K down to ~24K), with individual commands like git add/commit/push hitting 92%. Works with Claude Code, Cursor, Copilot, Gemini CLI, and 9 others. 50.4k stars.
TEAHEE Moment
Stay sharp, stay informed. See you tomorrow.
If you enjoyed this TEA, follow along on social for more:
Twitter/X