Hello, dear TEA-mates! Here is what you need to know today.
1. 💵 Traders Ran to Stablecoins Before the Fed Hike
Crypto traders moved capital into stablecoins ahead of Wednesday's Federal Reserve decision, according to CoinDesk research cited in this report. Markets priced a 92 percent probability of a 25 basis point hike that would lift the target range to 3.75 percent to 4.00 percent, the first increase since 2023 after five consecutive meetings held at 3.50 percent to 3.75 percent. The policy update landed one day after the Senate blocked the CLARITY Act. Data from Talos showed a 28 percent net buying preference for stablecoins going into the meeting, while buying conviction for Bitcoin fell to 3 percent and Ethereum to 9 percent. Bitcoin traded between $76,000 and $80,000 with volatility at a one-month low, Ethereum near $2,404 and XRP near $1.29. August inflation held at 3.4 percent annually, with core inflation easing to 2.4 percent. (Read More)
🫖 TEA For Thought: "And with the CLARITY Act being vetoed, the yield generated by holding stablecoins is another call out to those disappointed in the USD. With stablecoins continually being adopted in the AI agentic world, this might just be the beginning."
2. 🏢 Microsoft Says Access to AI Is Not the Advantage
Kathleen Hogan, Microsoft's Chief Strategy and Transformation Officer, published the company's own AI transformation report card on September 17, drawn from hundreds of internal efforts and packaged as the Frontier Playbook. The blunt admission up front: licensing a tool to over 200,000 people did not change how work got done, and early sales usage plateaued with no measurable impact. Five lessons follow, including starting from the business outcome, redesigning whole workflows rather than single tasks, and putting employees at the center. The numbers behind them: sales close rates up 20 percent and revenue per account manager up 9.4 percent; more than 111 agents deployed across cloud supply chain workflows, cutting planning cycle time from roughly 10 business days to under 2.5; a nine-person team shipping the first Copilot Cowork release in 35 days; and 58 percent of AI users saying the tools let them do work they could not do before, rising to 80 percent among advanced users. (Read More)
🫖 TEA For Thought: "I just have to say I don't quite agree with 'redesign the entire workflow, not just individual tasks' and also 'put employees at the center.' This feels still more like integrating AI, with AI just being the tool, instead of rebuilding the company inside out to make it AI native. And those that are not AI native will eventually still be replaced."
3. 🚦 Huawei Says China Cannot Feel the Risk Yet
Eric Xu, Huawei's rotating chairman, told reporters at the company's Connect conference in Shanghai on September 17 that Chinese AI developers may not yet be advanced enough to perceive the safety risks leading US labs are reporting, and should speed up rather than slow down. "Maybe the type of risk from AI that they can feel, they can perceive, is something that people in China or AI model providers in China cannot," Xu said, adding that Chinese providers "may need to speed up their pace to the level that they could also feel the risks." The remarks land as OpenAI commits to regularly disclosing unexpected or unauthorized model behavior and Anthropic CEO Dario Amodei calls for slowing development. Beijing is drafting a mandatory national standard for AI agent safety. Huawei forecasts agents will carry more than 90 percent of global AI processing traffic by 2035, with up to 900 billion active agents. (Read More)
🫖 TEA For Thought: "Basically the CCP running dog Huawei is saying that Chinese AI is not advanced enough to be slowed down yet, so just speed up."
4. 🧬 37,000 AI Agents Went Looking for a Cancer Drug
A system called the Virtual Biotech, described in Science and led by Stanford computer scientist James Zou, assembles as many as 37,000 AI agents into the org chart of a drug company, with a chief scientific officer agent directing divisions for target identification, clinical trial design and more. Versions of Anthropic's Claude powered the agents. Tasked with analyzing published results from more than 55,000 clinical trials, the system assigned 37,075 agents to one late-stage trial each and found that drugs targeting proteins active in specific cell types were nearly 50 percent likelier to reach market. Directed to assess the protein CD276 as a lung cancer target, it confirmed the candidate from existing data and proposed an antibody carrying an anticancer drug, which external reviewers called promising. The caveat is large: none of the predictions have been validated in experiments, let alone in patients. (Read More)
🫖 TEA For Thought: "Lung cancer is probably one of the deadliest ones, and this is definitely very promising news!"
5. 💷 UK Workers Spent £958 Million of Their Own Money on AI
A new Deloitte survey found that 63 percent of British workers have used generative AI for work, and 17 percent are personally paying for at least one tool they use on the job, spending £958 million between them. Nearly half use free tools at work, 34 percent use external tools their employer pays for, and 31 percent say they are using generative AI without their employer's knowledge. Secrecy is part of the pattern: 23 percent believe there is a stigma attached, 64 percent worry managers will conclude AI can do their jobs, and two thirds say their organization lacks convincing leadership on how the tools should be used. The most common uses are searching for information (43 percent), drafting emails (43 percent) and creating summaries (31 percent), and nearly half have had no formal training. The UK's National Cyber Security Centre warned last week that unapproved AI use risks data breaches and lost intellectual property. (Read More)
🫖 TEA For Thought: "This bring-your-own-AI to work is the new shadow tech that folks adopt. I guess when enterprise is slow at rolling out AI tools, folks just find their way around and adopt on their own!"
🛠️ Skill of the Day
The Blank-Page Rebuild: stops you from bolting AI onto a broken process and forces you to design the process as if it never existed.
You are an operations strategist. I am going to describe a process my team
runs today. Do not help me speed it up. Assume the process no longer exists
and we are building it from a blank page, with AI agents available from day one.
The process: [DESCRIBE IT, WHO TOUCHES IT, AND HOW LONG EACH PART TAKES]
Give me four things.
1. THE REBUILD. What this process looks like if designed today, start to
finish. Say what a person does, what an agent does, and where a person
must approve before anything continues.
2. THE DELTA. The three biggest differences from how we work now, and what
each one is worth in time, cost, or quality.
3. THE FIRST MOVE. The smallest change we could make next week that moves
toward the rebuild instead of patching the old process.
4. THE HONEST OBJECTIONS. Who loses something in this design, what breaks
first, and what evidence would prove the rebuild wrong.
Ask me for anything you need before you answer. Never use an em dash.Paste into ChatGPT, Claude, or your tool of choice. Replace the bracketed part with a process you actually run every week.
TEAHEE Moment
Stay sharp, stay informed. See you Monday.
If you enjoyed this TEA, follow along on social for more: