Why did crypto rally after the CLARITY Act stalled in the Senate on Sep 15? Bitwise CIO Matt Hougan's Sep 30 memo, covered by crypto.news, says faster regulatory action and fewer business restrictions helped. Four areas he flagged. (1/3)

1) Stablecoins: the negotiated bill text would have barred stablecoin yield in any form. It did not pass, so platforms keep room for rewards, and he names Coinbase the largest beneficiary. 2) Tokenized stocks: an SEC order on Sep 17 opens a five-year testing route. (2/3)

What is an AI agent? A chatbot answers when you ask. An agent gets a goal, uses tools like a browser, files or email, and works through the steps itself. Think of an assistant who finishes a task, not one who only replies. (1/3)

Good fit: sorting inbox, drafting first versions, updating sheets, checking prices. Routine work with clear steps and a clear "done". Weak fit: vague goals, missing data, anything needing judgment about people. (2/3)

Trump's crypto policy has a dollar angle: stablecoin payments and tokenized assets. GENIUS became law on July 18, 2025. But tokenizing a Treasury fund doesn't make it Bitcoin or risk-free. My take: follow custody, redemption and settlement, not "one big game" theories.

6/7 My test: who owes you cash, who holds the assets, and what happens if either fails? A 24/7 token transfer is not proof of 24/7 cash redemption. Check eligibility, cutoffs, fees and legal rights. Faster settlement does not erase liquidity or operational risk.

Is the AI race really over, or just changing? Mayankshah's Short argues an early frontier-model lead could become impossible to catch - citing Anthropic's Series C deck, huge training costs, and AI-assisted research compounding the incumbents' edge. A thread. (1/7)

The access question is real: if open models approach frontier capabilities by year-end, as he forecasts, will governments and labs still accept unrestricted release? That is a prediction, not an outcome - and open weights do not mean free hosting or licensing. (6/7)