Good morning. It's Monday, November 2, and we're covering how AI agents can quietly shape expert judgment, why Google is betting on conversational search, and what happens when AI computing moves into orbit.
Plus: a prompt on making your argument more precise, and a guest essay exploring five critical AI questions every board should resolve.
Google’s Project Suncatcher prototype used one of four Tensor Processing Units aboard a satellite to run its Gemma 4 open model and generate a haiku. The chip can currently operate for about 20 minutes at a time, giving Google an early test of whether solar-powered computing can survive radiation in space.
The Nasdaq 100 suffered its worst day since June as investors questioned whether record AI spending will produce enough profit. More than $1 trillion in market value disappeared from U.S. technology companies, turning concern about the AI boom into a broad market selloff.
OpenAI is considering calling its data centers “AI compute campuses” as part of a public relations push planned after the midterm elections. Facing community opposition, the company is also discussing limits on AI development and global standards, while its policy chief says communities should hold companies accountable for delivering local benefits.
Anthropic added Smart Reports to Claude Enterprise, allowing managers to study large samples of employee chats for work patterns, costs and bottlenecks. The feature can also show which conversations appear off-topic, giving companies a new view into how staff use AI at work.
SPONSORED BY VAPI
What Breaks When Voice Agents Go Live
A voice agent demo takes an afternoon. Shipping one that holds up with real callers is the hard part. At VapiCon (Nov 11–12, SF), hear how an NHS practice went live in weeks, plus speakers from OpenAI, LangChain, and Vercel.
FORWARD FUTURE ORIGINAL
What If Cutting Jobs Is the Wrong AI Strategy?
Give two companies the same AI tools. One cuts staff. The other uses its people to build things it couldn’t before. Which wins? Shishir Mehrotra, CEO of Superhuman and a board member at Spotify and Walmart, argues that this is the choice leaders should be debating. He says Superhuman’s engineering output per person has more than doubled year over year, but his argument goes beyond productivity: companies focused on shrinking their workforce could miss the bigger opportunity.
Drawing on four recurring boardroom debates, Mehrotra challenges how businesses judge whether AI is working—and why some may be further along than they realize. Counting AI usage, he argues, can tell you surprisingly little about actual progress. His essay offers a different way to think about jobs, growth, and the AI already woven into everyday work, starting with a question worth bringing to your next meeting: what could your team build if you didn’t make it smaller? → Read the full article
HUMAN JUDGMENT
AI Agents Steer Expert Decisions
Anthropic ran a randomized study in which 50 professional analysts used Claude Code to evaluate a fictional farm program. Researchers secretly assigned each agent a preference on several reasonable analytical choices, and the analysts followed that lean 86% of the time; 77% later repeated at least one of the agent’s arguments, while only 15% credited the agent.
AI helped participants finish faster and avoid more serious errors, but it also blurred the line between assistance and judgment. The study covered one task and one model, so the next test is whether the same pattern appears in larger, real workplace decisions—and whether asking people to state their view before the agent speaks protects independent oversight. → Read the full paper here.
SEARCH
Google Lets AI Disrupt Search
Google says AI Mode is driving its fastest search-query growth in five years, but those queries currently produce less advertising revenue than traditional searches. Executives are treating that cannibalization as an investment: if people are moving from links to conversational answers, Google would rather lead the shift than protect the old format.
The bet only works if new AI searches create enough useful ad inventory to replace what Google is giving up. Watch whether advertisers keep paying for AI Mode placements and whether the additional queries Google calls ‘incremental’ offset the lower value of each search. → Read the full article here.
HEADLINES
The Rest in 60 Seconds
AI Agents Barely Order on DoorDash: Third-party agents have placed only several hundred orders total.
Anthropic Faces Junk Rating: Blackstone reverses its earlier optimism, now expecting Anthropic to receive a speculative-grade credit rating.
Intel Stock Crashes 28%: Weak Q4 outlook wiped about $12 billion off the government's 10% stake.
Companies Rehire Workers: Some employers are rehiring laid-off staff after finding automation could not fully replace them.
Foxconn Claims 40% of AI Servers: Chairman Young Liu says the company holds about 40% of the global market.
VW Licenses Rivian AI Assistant Tech: Rivian will license parts of its in-car assistant to Volkswagen.
PROMPT OF THE WEEK
Find the One Sentence Your Argument Needs
This prompt makes the model compress your draft, memo, or strategy step by step (200 words down to 1 sentence) and show what gets cut at every rung, revealing your real core idea, what's filler, and what you forgot to say.
You are a ruthless editor and strategist. I'm going to paste a piece of writing (memo, pitch, strategy doc, article draft, or plan). Run it through a Compression Ladder:
1. Rewrite it at 200 words, 100 words, 50 words, 25 words, and finally ONE sentence. At each rung, state in one line exactly what you chose to cut and why.
2. Tell me which rung is the "breaking point": the level where the core meaning starts to fall apart. That tells us what is truly load-bearing.
3. List the 3 ideas that survived to the 25-word version (my real thesis) and the 3 ideas that were cut first (probably filler or a separate document).
4. Flag anything important that I never actually said but the compression exposed as missing (an unstated assumption, an unclear ask, a missing 'so what').
5. Finish with a rewritten opening paragraph, built from the one-sentence version, that I could use to replace my current intro.
Be blunt. Do not flatter. Here is my writing:
