Why we keep ending up in the same place, the case for AI data centers in space, Lego (smart and dumb), and more. Plus, my interview with Charles Yang on American reindustrialization.
A special announcement from Lux Capital
We’re very fortunate to announce our new fund, Lux Ventures IX. With over $1.5 billion, it’s the largest fund in our history. While the number is new, the thesis is not: we’re going to continue to invest where we have since our founding at the cutting-edges of frontier science and technology. “We happen to be in the companies delivering this at a critical time,” Josh Wolfe noted to Bloomberg. “Many feel very confident that this will be a company-making next few years for some” startups.
As we wrote in our announcement:
Today entire sectors of aerospace, biotech, defense, industrials, transportation and beyond are being reinvented by a new generation of brilliantly ambitious, often irreverent, scientists and engineers. And while markets have changed and capital reshapes around fewer companies and larger checks, our focus on the craft remains constant.
Thank you to our pathbreaking founders, our enthusiastic limited partners and the entire Lux ecosystem who make this possible.
From around the web
1. What’s new?
Starting something new in 2026? Be warned. It might end up looking pretty similar to everything that came before. In a new post, our scientist-in-residence Samuel Arbesman explains why: canalization. It’s an idea borrowed from biology that explains why lots of different genetic combinations end up producing similar traits despite variation, environmental forces and randomness. Sam argues the principle applies outside of biology, too. For example, new research organizations tend to converge toward familiar institutional forms despite attempts to create novel structures.
With organizations that end up looking university-like, it comes down to risk, or more precisely, risk aversion. Imagine you build a weird institution devoted to some odd interdisciplinary topic. Then, you hire people to work in it. These are often academic researchers with varying levels of exposure to academia’s incentives and requirements. Each hire then begins to think the following: “What if this organization crashes and burns? Or what if I’m simply not a good fit for it? I need to make sure I can get another job afterward.” And for the most part, “job afterward” means another academic job.
2. Think tank theater
For me, Sam’s piece raises the question of what all the new think tank formats sprouting up are really for — a question I also got into this week on the podcast. Coming at this issue from a slightly different perspective is an article Laurence found over on Brookings. Author Jeremy Shapiro’s surprisingly amusing narrative about the secret rituals of meetings between government officials and think tank experts shows how the blob uses “thinkers” to validate their preexisting choices rather than to learn.
[The thinker] clears his schedule of any conflicting brown bags on separatism in South Ossetia and, after a suitable interval to keep the government guessing as to his availability, replies that he might be able to squeeze it in to his schedule. Citizenship data and social security numbers are provided for security purposes, times are confirmed and ground rules are established in a multitude of emails with a seemingly never-ending array of staffers, all of whose titles include the word “special.” The thinker says nothing directly to his colleagues, but searches desperately for opportunities to obliquely allude to the meeting: “I’d love to come to your roundtable on uncovered interest rate parity, but I unfortunately have a meeting with the secretary of defense.”
3. Burn, baby, build
Speaking of validation. This week, Lux’s Shaq Vayda recommends a great new exploration of the economics of AI data centers in space from Andrew McCalip (of portfolio company Varda). The idea, although far less “obviously stupid” than Andrew says he first thought, is really only feasible under aggressive assumptions and if one company (ahem, SpaceX) decides to throw money at the problem basically for funsies.
I’ll go one step further and say the quiet part out loud: we should be actively goading more billionaires into spending on irrational, high-variance projects that might actually advance civilization. I feel genuine secondhand embarrassment watching people torch their fortunes on yachts and status cosplay. No one cares about your Loro Piana. If you’ve built an empire, the best possible use of it is to burn its capital like a torch and light up a corner of the future. Fund the ugly middle. Pay for the iteration loops. Build the cathedrals. This is how we advance civilization.
4. Turing taste
On to more earth-bound concerns. Laurence also liked Vauhini Vara’s latest in The New Yorker. She asks what happens if people end up liking AI-generated fiction more than the real deal. Based on my usage of these tools, it sounds far-fetched. But maybe not. Vauhini reports on a computer scientist who fine-tuned AI models on the complete works of 30 prominent authors and found that creative writing students preferred the AI-generated prose to human imitations in nearly two-thirds of cases.
Reading the authors’ original passages alongside the A.I. imitations, I was startled to find that I liked some of the imitations just as much. The A.I. version of Han’s scene, about the newborn’s death, struck me as trite in places. But, to me, the line about the mother’s chant was more surprising and exact than the original. I also spotted some good bits in an imitation of Junot Díaz. In “This Is How You Lose Her,” Díaz writes, “The one thing she warned you about, that she swore she would never forgive, was cheating. I’ll put a machete in you, she promised.” To my ear, the A.I. rendition was more rhythmic and economical: “She told you from the beginning that if you ever cheated on her she would chop your little pito off.”
5. Toy story
If you’re like me, this first week back after the holidays was a bear, so let’s ease on out with something fun. In March, Lego is launching “Smart Bricks” — bricks with tiny computers that communicate with NFC-equipped tiles and minifigures to add interactive sound, light, and motion effects to builds. Many on the Lux team have already added them to our Christmas lists for next year.
Feeling lo-fi? Perhaps PastaTube is a better fit.
This collection brillantly demonstrates why specialized newsletters beat algorithm-driven feeds for discovering genuinely interesting ideas. The canalization piece nails something I've observed repeatedly—new research orgs inevitably drift toward academic structures because individual researchers need career insurance, not becaus the model itself is optimal. I cofounded a tech incubator years ago and dunno how many times we saw founders unconsciously recreate traditional corporate hierarchies despite explicitly trying to build something different. The space datacenter economics are wild but the broader point about funding "irrational, high-variance projects" is exacty right for advancing civilization.