My end-of-week morning reads:
• Tokens Too Cheap to Meter. The price of using machine learning intelligence is decreasing by several orders of magnitude a year and shows no signs of slowing. We are likely to see LLMs integrated into every part of computing as infrastructure, not just as a product, in the next year or two. We are likely to see LLMs running locally at current frontier-quality on commodity hardware in the next 3-6 years. Starting very soon, we are likely to see quality and access become the limiting factor to AI 1 use, not sheer number of tokens. (via FT Alphaville) (jyn.dev)
• Owning a Home Is Seen as a Financial Holy Grail. But What If It Isn’t?: Michael Luca and Raymond Fisman, starting from Hoover’s 1931 homily on homeownership, on the unrecognized trade-offs that can make renting the better economic choice. (Wall Street Journal) see also The New Rules for Selling Your Home With Mortgage Rates Near 7%: Agents give much different advice than they did five years ago. (Wall Street Journal)
• How China Thinks About AI: Derek Thompson works through seven big questions — and many answers — about the thing everybody’s talking about. 7 big questions—and many answers—about the thing that everybody’s talking about (Derek Thompson)
• A Rocket Supply Crunch Is Making It Harder to Hitch a Ride to Space: A scramble across the industry as SpaceX winds down its flagship vehicle and rivals struggle to ramp up. (Wall Street Journal)
• America’s Hot, New Retirement Destination: Paris: Business Insider on the Yabrows, who turned a layoff and a three-city Europe trip into a one-way move — “I deleted that spreadsheet.” A wave of Americans are making permanent moves to France — and some are paying way too much (Business Insider)
• An Army of Paid Creators Is Carpeting the Web With Political Content: Stuart A. Thompson on the blizzard of cheap videos shifting public conversation — no celebrity followings required. Unlike celebrity influencers with large followings, these creators produce a blizzard of videos for a fraction of the cost of traditional online campaigns and, in the process, shift public conversation. (New York Times)
• These Tiny Startups Are Getting Even Smaller With Help From AI: By holding off on hiring, and sometimes eliminating existing positions, entrepreneurs are growing their products but not their head counts. (Wall Street Journal)
• Americans are drinking less since the pandemic began, but one group bucks the trend: Studies previously found alcohol consumption and deaths surged during the pandemic as Americans grappled with stress and isolation. Drinking among those ages 50-64 is still higher than before the pandemic; the group’s overall alcohol consumption is up by almost 4 percent and heavy drinking increased by 36 percent between 2018 and 2024. (Washington Post)
• Trump Has Finally Lost the Plot. Revolutions can fail but still scar the landscape. And Mr. Trump, with just under two-and-a-half years left in his second term, might emotionally and psychologically blow apart. But things are different than they were… Peter Wehner on a presidency whose frantic pace makes days seem like weeks — and the search for rhyme or reason. (New York Times)
• Tennis for Savages: Victory Journal traces handball back to 2000 BC Egypt, by way of Homer — “the ball incessant flies.” Handball was in Jerome Liebling’s blood. The legendary photographer spent much of his life documenting its practitioners, from Brooklyn to Miami Beach. (Victory Journal)
Video of the day: No Wonder Everyone is Obsessed with Reacher
Be sure to check out our Masters in Business interview this weekend with Adam Frank, Head of Wealth Planning and Advice at JPM, responsible for $1.3 trillion in client assets. Previously, he was head of Wealth Management for JP Morgan Securities. JPM’s combined total assets, including global, institutional, private bank, wealth, and retail, are over $7.7 trillion.
The AI Build-Out Is Becoming the Biggest Economic Bet in U.S. History
Source: WSJ
Sign up for our reads-only mailing list here.
Facts Only
* The price of machine learning intelligence is decreasing by several orders of magnitude annually without slowing.
* LLMs are likely to be integrated into computing infrastructure in the next one to two years.
* LLMs may run locally on commodity hardware within three to six years.
* Quality and access are expected to become the limiting factors for AI use, rather than the number of tokens.
* Homeownership is presented as a financial ideal, with unrecognized trade-offs making renting a potentially better economic choice.
* Agents selling homes face different advice due to mortgage rates near 7%.
* Content creators produce videos for a fraction of the cost of traditional online campaigns, shifting public conversation.
* Entrepreneurs are growing products without increasing headcount due to hiring freezes or eliminations.
* Alcohol consumption among those aged 50-64 has increased by almost 4 percent since the pandemic began, with heavy drinking increasing by 36 percent between 2018 and 2024.
* A scramble in the industry is occurring due to a rocket supply crunch affecting space travel.
* Americans are making permanent moves to Paris.
Executive Summary
Full Take
The juxtaposition of rapid technological deflation (AI cost reduction) against persistent, tangible human constraints (housing economics, physical supply chains) reveals a tension between exponential progress and real-world friction. The narrative implies that the next significant economic battles will be defined not by raw computational power or token volume, but by access, quality control, and structural shifts in societal priorities regarding where to allocate resources—whether toward physical assets like homeownership or human capital deployment for content creation.
The pattern suggests a focus on emergent constraints: as abstract, high-leverage technologies (like LLMs) become ubiquitous, the bottlenecks shift from input scarcity to quality control, distribution mechanisms, and macroeconomic realities. The reference to China’s approach to AI alongside the trajectory of personal finance and geopolitical shifts indicates a broader thematic concern about global power dynamics and how technological acceleration is being negotiated within existing socio-economic frameworks. The tension between the desire for frictionless progress (AI integration) and the reality of systemic inertia (housing markets, supply chain limits) suggests that future resilience will depend on navigating these conflicting realities simultaneously rather than solving them in isolation.
Bridge Questions: If quality becomes the limiting factor for AI access, what institutional structures are necessary to enforce universal quality standards? How does the shift toward distributed content creation re-evaluate traditional metrics of cultural and political influence? What are the long-term economic consequences when high-demand real estate aspirations conflict with broader economic opportunities?
Sentinel — Human
This text appears to be a collection of curated headlines and short excerpts assembled into a newsletter format, showing strong human thematic selection rather than seamless machine generation.