Can the desert cure insomnia?; why you should quit your job and fix typewriters; and why China is set to pull ahead in AI. Plus Jacob Ward on AI safety (and croissants).
From Lux Capital
Josh’s annual Lux LP Letter is out, with new reflections on our origin story, some of our biggest mistakes, and where things are headed from here. I found this part especially useful:
You can find his annotated version of the letter on X.
In other Lux news, Trial Library just raised a $10 million Series A to better match patients with clinical trials. Portfolio company Varda, which is building space factories to launch and return space-made products for terrestrial applications, was profiled in TechCrunch:
When Will Bruey talks about the future, the timelines are shorter than most might imagine. The Varda Space Industries CEO predicts that within 10 years, someone could stand at a landing site and watch multiple specialized spacecraft per night zooming toward Earth like shooting stars, each carrying pharmaceuticals manufactured in space. Within 15 to 20 years, he says, it will be cheaper to send a working-class employee to orbit for a month than to keep them on Earth.
Finally, check out Wired’s reporting on Fortell’s AI-powered hearing aid (formerly known as Chromatic). The portfolio company’s device is inspiring long waiting lists of people who want to try it out — and tears of joy when they do.
From around the web
1. Daylight savings
How’s your sleep? Rowan Jacobsen’s is horrible. In a new article for Harper’s, he chronicles his attempt to cure his chronic insomnia by spending a month in the Arizona desert living without artificial light after dark. The key insight: it’s not just avoiding light at night that matters, but maximizing the amplitude between bright days and dark nights — the intense daylight exposure our ancestors evolved with but modern indoor life has erased. Good luck with that this dark December.
It feels strangely radical to have the sunset as one’s entire agenda. It passes through a dozen stages, from fiery orange to astral lavender, a menagerie of photons that left the solar corona eight minutes ago and banked their way into me. All the while my light meter is slowly ticking down, from a few hundred lux to just dozens. Finally, the color is gone, except for some peaceful blue bands along the skyline. It seems clear that we’re built for this slow fade, rather than lights-out at eleven.
2. Shift happens
In more charmingly off-beat “self help,” our scientist-in-residence Sam Arbesman recommends Kurt Streeter’s “How to Fix a Typewriter and Your Life” in The New York Times. The story follows a burned-out facilities manager named Paul Lundy, who in 2015 read a newspaper article about 92-year-old typewriter repairman Bob Montgomery. Paul became his apprentice, eventually bought his business, and found a new purpose late in life.
He showed Lundy the right way. No anger. No frustration. Just quiet insistence that good enough was not good enough.
Sometimes Mr. Montgomery would partly disassemble a machine and leave it on Lundy’s bench — a test, a puzzle, a method of teaching as old as apprenticeship itself.
3. Grid expectations
If the United States wants to remain competitive with China in AI, it needs to do something new, say Jared Dunnmon, Avanika Narayan, Jon Saad-Falcon, and Chris Ré. While China is limited by export controls on advanced chips, the United States remains deeply constrained by power grid bottlenecks. The solution: maximizing “Gross Domestic Intelligence” — the total computational intelligence a nation can deploy — rather than just building more data centers. This will mean activating the millions of AI-capable laptops and phones already in American homes through hybrid local-cloud inference systems, which could boost usable AI capacity by 30–40% without building new infrastructure.
Maximizing deployed intelligence requires two simultaneous objectives: making computation more intelligent per unit of power (better models, better chips), and maximizing the power available to run that computation. Miss either dimension and you lose: efficient models without power sit idle; abundant power without efficient computation drives up costs with diminishing returns. In this framing, the absolute amount of intelligence a nation can deploy is the product of intelligence per watt and the amount of power usable for computation — what we call Gross Domestic Intelligence (GDI).
4. Green tech, red tape
Also harming U.S. competitiveness: over-regulation. Laurence liked Peter Reinhardt’s discussion of his travails trying to start two climate-tech hardware companies: Charm Industrial (carbon removal) and Revoy (electric trucking). Peter argues that more than half the cost of building each company went to regulatory burdens. He details how a 5.5-year wait just to determine which permitting pathway to use cost Charm $125M directly and $300M in public health costs from continued pollution.
Regulatory delays and bottlenecks have added millions of pounds of pollutants like PM2.5, NOₓ and CO₂ to our air from the continuation of business as usual, instead of the deployment of clean technologies from my two hardtech efforts alone. While CO₂ is a long-term climate issue, PM2.5 and NOₓ are immediate major drivers of asthma and excess morbidity. Both operations have high bipartisan appeal—and we’ve never been denied a permit—because we’re fundamentally cleaning up things that matter to everyone: dirty air, wildfires, orphaned oil wells. Revoy is also helping deflate the cost of long-haul freight. But none of that has made getting freedom to operate easy. For creative new technologies the default answer is “no” because there isn’t a clear path to permitting at all, and figuring out that path itself takes years — time that startups can’t afford to wait.
5. The thin red line
Laurence also suggests checking out the debate over the poverty line between and Jeremy Horpedahl. Michael argues America’s poverty line is catastrophically outdated, and that updating the formula with today’s spending ratios yields a real crisis threshold of around $140,000 for a family of four (rather than closer to $31,000). This means, he says, that the median household earning $80,000 is actually living in poverty.
When you run the net-income numbers, a family earning $100,000 is effectively in a worse monthly financial position than a family earning $40,000.
At $40,000, you are drowning, but the state gives you a life vest. At $100,000, you are drowning, but the state says you are a “high earner” and ties an anchor to your ankle called “Market Price.”
Jeremy retorts that Michael made three major errors: he underestimates how much income American families have; treats average spending as minimum spending, overstating costs by 30-40%; and ignores that the Census Bureau already has a better measure than the poverty line — the Supplemental Poverty Measure. The debate between the two is ongoing.
