Daily TEA – Open Algorithms, AI Agents & Stablecoin Fault Lines

X algorithm, stablecoin flaws, AI agents, Venezuela, UAE AI surge

Hello, dear TEA-mates—here’s what you need to know today.

1.🐦X Promises to Open-Source Its Recommendation Algorithm

Elon Musk says X will open-source its new recommendation algorithm within seven days, including all code used to decide which posts and ads are shown to users. The company plans to repeat these releases every four weeks, publishing updated code and developer notes so outsiders can track changes over time. Musk did not provide a detailed rationale, but X has faced user criticism and regulatory scrutiny over how content is surfaced on the platform. Read More (Bloomberg)​

🫖 TEA For Thought: This is huge—curious to see how the algorithm actually works under the hood, and whether it will be meaningfully forkable.

2.💱 Vitalik Buterin Flags Unsolved Flaws in Stablecoin Design

Ethereum co-founder Vitalik Buterin says decentralized stablecoins still face three unresolved structural weaknesses: long-term USD pegs, oracle capture risk, and staking-yield trade-offs. He questions building “resilient” systems around indefinite U.S.-dollar pegs, arguing that inflation and macro shocks over 20-year horizons can erode value and that future designs may need to track broader purchasing power indexes instead. Buterin also warns that oracles—needed to feed external prices on-chain—can be captured by large pools of capital, forcing protocols into economic arms races where they must extract fees or governance power from users to make attacks too costly. A third issue arises when stablecoins use staked ETH as collateral, creating an inefficient competition between validator yields and returns for stablecoin holders; he sketches options like lowering staking yields or creating a low-risk, yield-bearing staking class, but does not endorse a single solution. Read More (CoinMarketCap)​​

🫖 TEA For Thought: Overreliance on USD is the real fault line—petrodollars are becoming “crypto dollars,” while post-gold-standard money printing continues; baskets of currencies or scarce-asset-pegged stablecoins feel like the more robust endgame.

3.🧑‍💻 OpenAI Uses Real Work Artifacts to Benchmark AI Agents

OpenAI, working with data provider Handshake AI, is asking third-party contractors to upload real assignments and deliverables from current or past jobs—including documents, decks, spreadsheets, code, and other files—to benchmark its next wave of AI agents. Contractors are instructed to pair each task request with the original “experienced human deliverable,” giving OpenAI detailed examples of how people actually complete complex, multi-hour or multi-day work. The guidelines tell contributors to scrub confidential, proprietary, or personally identifiable information, with a redaction assistant provided, as OpenAI looks to measure whether agents can not just pass tests but genuinely perform office-style work. Read More (Wired)​

🫖 TEA For Thought: This reinforces that AI agent benchmarks should move beyond test-passing toward assessing whether agents can reliably execute end-to-end real work.

4.🪙 Tether at the Heart of Venezuela’s Oil and Sanctions Saga

A Wall Street Journal report details how Venezuela under Nicolás Maduro turned tether (USDT) into a core settlement rail for its state oil company, helping it sidestep U.S. sanctions while hyperinflation wrecked the bolivar. PdVSA shifted a large share of oil-export payments into tether and other stablecoins via intermediaries, and one local economist estimates that nearly 80% of the country’s oil revenue is now collected in assets like USDT. Tether has since worked with U.S. authorities and froze multiple wallets tied to the oil trade, and analysts say Maduro’s fall is unlikely to dislodge stablecoins from Venezuela’s sanctions-stressed, dollar-starved economy. Read More (MSN / Wall Street Journal)​

🫖 TEA For Thought: If roughly 80% of Venezuela’s oil revenue flows through stablecoins while Tether’s reserves lean heavily on volatile assets like Bitcoin and Ethereum, it feels like only a matter of time before a Terra-Luna–style collapse hits at a sovereign scale.

5. 🌍 Global AI Adoption Widens the Digital Divide

Microsoft’s AI Economy Institute reports that about one in six people worldwide used generative AI tools in the second half of 2025, but adoption is growing much faster in the Global North than in the Global South. The UAE leads all countries with 64% of its working-age population using AI, followed by Singapore and several European states, while the U.S.—despite its frontier models and infrastructure—ranks just 24th with a 28.3% usage rate. South Korea is a notable climber, jumping from 25th to 18th as government policy, Korean-language models, and viral consumer apps pushed adoption above 30% of the population. Read More (Microsoft)​

🫖 TEA For Thought: It’s striking that the UAE and multiple European countries so clearly outpace the U.S. in everyday AI use—the big U.S. AI factories are unmatched, but true mass adoption at home is still lagging far behind.

Prompt Tip of the Day: The Rapid Test Loop

Checks understanding early.

Quiz me with 5 questions on this topic: [topic]. Increase difficulty gradually. Explain mistakes briefly after each answer.

TEAHEE Moment

Stay sharp, stay informed. See you tomorrow.

Follow along on X for more real-time sips of TEA: Twitter/X.