2026-09-25

Google is about to launch a small data center into orbit — four TPUs that will run for exactly 15 minutes at a time, powered by solar arrays in space. It's a genuinely novel approach to an infrastructure problem: putting compute where the energy is abundant and free, rather than building more terrestrial data centers. The project is audacious enough to be fascinating regardless of whether it scales, and it raises real questions about what "cloud computing" really means when the cloud is literal orbit.

This piece makes a case that AI labs should fund historical research — specifically using LLMs to trace how alchemical knowledge spread across 17th-century Europe through encrypted correspondence. It's not just a thought experiment; the author has done the work, using models to decipher encoded letters and map intellectual networks that conventional history tools struggle with. What makes it compelling is the reversal: instead of asking what AI can learn from history, it asks what AI can reveal about history that we couldn't see before.

Astra's return-to-flight timeline is almost beside the point — the interesting argument here is that the era of truly cheap launch might already be behind us. The dynamics that drove prices down (venture capital, rapid iteration, accepting risk) are colliding with customer expectations around reliability that push costs back up. This is a structural analysis of an industry that promised to revolutionise access to space and is now discovering the economics don't work the way the pitch decks said they would.

The Dutch government is building its own sovereign IT infrastructure on NixOS — a declarative, reproducible Linux distribution — as an alternative to Microsoft's stack. This isn't a small pilot; it's an entire digital workplace platform designed to reduce vendor lock-in at national scale. The project is notable for how it handles configuration management and security at scale, and serves as a real-world test case for whether governments can actually escape the Microsoft ecosystem through open-source infrastructure.

The liver can regenerate even after 70% removal — a trick almost no other human organ can pull off. This piece walks through the mechanics: which cells do the work, what signalling pathways trigger regrowth, and why evolution gave this particular organ such extraordinary redundancy. It's the kind of deep-dive that makes you realize how little we understand about the basic biology of our own bodies, and touches on bigger questions about why regeneration is the exception in mammals rather than the rule.

A thru-hiker reflects on the John Muir Trail after finishing — not just the highlights, but what surprised them, what they'd do differently, and how the experience settled in the weeks after. The "post-trail retrospective" format is a valuable one in outdoor writing: it captures the transition back to normal life, the things that felt important on-trail that faded, and the things that lingered. For anyone who's done a long walk or is considering one, this is the most useful kind of trip report.

A counterintuitive institutional story: organized crime in Italy may be acting as an accidental public-health barrier against the fentanyl epidemic. The argument is that the Mafia's traditional heroin business gives them a vested interest in keeping more dangerous synthetic opioids out of their supply chains, and their territorial control makes it hard for competitors to introduce them. Whether or not the thesis holds, it's a fascinating lens on how illegal markets self-regulate in unexpected ways — and how the absence of fentanyl in a country can be explained by something other than good policy.

A bug tracker that lives entirely inside your Git repository — no server, no database, no API. Issues are stored as Git objects, which means they branch, merge, and sync the same way code does. It's a clever repurposing of existing infrastructure (the Git object model was never designed for this) that solves a real problem for offline-first and decentralized workflows. Even if you never use it, the design thinking behind it — what else could we store in a distributed version control system? — is worth exploring.

Before Maury, ocean navigation was largely guesswork and accumulated experience. He turned it into a data-driven science by systematically collecting ships' logs, charting wind and current patterns, and publishing the first comprehensive maps of ocean routes. It's a story about the power of aggregating observations into a usable system — what we'd now call open data — and how one person's systematic thinking can change an entire industry. A reminder that the biggest leaps often come from organizing existing information, not discovering new facts.

Stumptown has taken over a historic Los Angeles movie theatre and turned it into a coffee flagship — the kind of adaptive reuse that preserves architectural character while giving a building a new purpose. The piece covers how the space was transformed, what design choices were made, and why the location matters. It's a good example of the intersection between coffee culture and urban preservation, showing how a third-wave roaster thinks about physical space as more than just a counter with an espresso machine.