🌀🗞 The FLUX Review, Ep. 198

July 17th, 2025

“FCP-198” // Photo: MK with Midjourney

Episode 198 — July 17th, 2025 — Available at read.fluxcollective.org/p/198

Contributors to this issue: Neel Mehta, MK, Erika Rice Scherpelz, Justin Quimby

Additional insights from: Ade Oshineye, Alex Komoroske, Ben Mathes, Chris Butler, Dart Lindsley, Dimitri Glazkov, Jasen Robillard, Jon Lebensold, Julka Almquist, Kamran Hakiman, Lisie Lillianfeld, Melanie Kahl, Robinson Eaton, Samuel Arbesman, Scott Schaffter, Spencer Pitman, Boris Smus, Wesley Beary

---

We’re a ragtag band of systems thinkers who have been dedicating our early mornings to finding new lenses to help you make sense of the complex world we live in. This newsletter is a collection of patterns we’ve noticed in recent weeks.

“We cannot master everything, taste everything, understand everything, drain every experience to its last dregs. But if we have the courage to let almost everything else go, we will probably be able to retain the one thing necessary for us—whatever it may be. If we are too eager to have everything, we will almost certainly miss even the one thing we need.”

— Thomas Merton

🌱🏗️ Just enough structure

An open-source project gets some attention and becomes kind of a big thing. Within weeks, contributors are fixing bugs, building features, and opening issues from every timezone. For a moment, it feels like magic. Then things start to fray. People propose conflicting directions. Code gets duplicated. The mailing list becomes a coordination graveyard.

This is a natural consequence of scaling. As group size increases, the number of communication channels scales non-linearly—roughly with the square of the number of participants. A team of five people has ten pairwise relationships to manage. A team of 15 has 105. Beyond a certain point, coordination alone can consume the system.

You can mitigate this with hierarchy… if you’re aware of the risks. Hierarchical systems can over-optimize for local concerns. Isolated subsystems can be so well optimized for their own function that they lose alignment with the whole and undermine system-level goals. Subsystems can also become sources of change resistance when gatekeepers use their power to resist changes they oppose.

Yet at a particular scale, hierarchy is unavoidable. Look at it as a system design challenge: you need a structure that allows local autonomy while maintaining systemic coherence. Decisions at one level can’t require perfect information about all others, yet they still need to nest into the broader context. Hierarchy works best when local optimization and global optimization point in the same direction.

Christopher Alexander offers a valuable design principle. In The Timeless Way of Building, he describes how pattern languages—those deep structures that guide how things are made—are inherently hierarchical. There are dependencies, but influence fades with distance. When you’re designing at one level, Alexander advises us to look one step up and one step down. This constrains the problem space enough to keep it tractable without losing the integrity of the whole.

In systems that scale well, units operate with a sense of fractal accountability. They are aware of the whole, responsive to the adjacent, and grounded in their layer. These hierarchies don’t centralize control—they distribute understanding.

🛣️🚩 Signposts

Clues that point to where our changing world might lead us.

🚏🏛️ There are now crypto coins tied to stocks—including those of private companies

Stock-tracking tokens are increasingly popular in the crypto world; these cryptocurrencies aim to mirror the price of a stock, allowing investors to trade stocks 24/7 or utilize other crypto infrastructure for lending, leverage, or anonymous trading. These tokenized equities are becoming available on financial platforms like Robinhood and Kraken, but they’re thinly traded and have suffered from huge price discrepancies: one Amazon-linked token briefly sold for over 100 times Amazon’s actual stock price. Robinhood also has stock tokens that apparently track the performance of SpaceX and OpenAI, both of which are private companies; OpenAI quickly disavowed those tokens.

🚏🏎️ An electric car drove a record 749 miles on a single charge

A car made by EV company Lucid Motors recently covered 749 miles (1,205 km) on a single charge, breaking the old record of 649 miles (1,045 km), which itself was set just last month. The car’s stated range is just 596 miles (960 km), so it probably benefited from regenerative braking during the downhill trip from St. Moritz in the Swiss Alps to the lower-lying Munich. (The course was still tough: the car took a mix of back roads and highways, the latter of which is “usually the nemesis of hypermiling.”)

🚏🤑 A company trying to donate to the inauguration sent $250k to a scammer

A crypto company called MoonPay was trying to donate a large amount to Trump’s presidential inauguration in January when they got an email from ‘steve_witkoff [at] t47lnaugural.com’; note the lowercase “L” masquerading as an uppercase “I.” The email, which purportedly came from the co-chair of the inaugural committee, directed MoonPay’s execs to send $250,000 in the stablecoin Tether to a certain crypto wallet. MoonPay sent the money, but investigators later found that the wallet was registered to a Nigerian scammer, and the email was sent from Nigeria too—so the whole thing was a scam on the donors.

🚏🐧 Linux has reached 5% desktop market share in the US

Linux now represents 5.03% of the US computer market, compared to 63% for Windows, 24% for Mac, and 2.7% for ChromeOS. Indeed, Linux’s growth has accelerated in recent years: “it took eight years to go from 1% to 2%... then just 2.2 years to reach 3%... and a mere 0.7 years to hit 4% (February 2024).” It’s unclear exactly what has driven this growth, but analysts believe that the popularity of the Steam Deck, the upcoming end-of-life of Windows 10, and the rise of Linux-powered crawlers and bots may be factors.

📖⏳ Worth your time

Some especially insightful pieces we’ve read, watched, and listened to recently.

• Why Micropayments Will Never Be a Thing in Journalism (Columbia Journalism Review) — Argues that paying per-article for journalism (as opposed to subscriptions) isn’t a feasible business model. Newspapers and magazines are package deals, with light and heavy, cheap and pricey, and popular and niche stories complementing each other. Separating them out defeats the bundling logic, and since only the hit stories would earn revenue, the per-article cost would have to be a lot higher than most people would be willing to pay.

• Gen Z and the End of Predictable Progress (Kyla Scanlon) — Argues that the classic American wealth-building formula (increased earnings from education, steady career advancement, and home appreciation) has broken down for today’s young people, for whom “the main path to financial security comes through the algorithmic gods rather than institutional advancement.” When a single viral TikTok or lucky crypto trade can make more money than years of professional work, it’s unsurprising that Gen Z’ers are adopting Taleb’s famous “barbell strategy” of either extremely low-risk jobs (e.g., non-college trades) or high-risk ones (e.g., creator economy, AI startups, crypto).

• Who Understands Alignment Anyway (Jessica Hullman) — Observes that all interdisciplinary collaborations tend to be hard, and the one between Machine Learning (ML) and Human-Computer Interaction (HCI) seems likely to be particularly slow and messy. HCI reached its apogee with the advent of desktop and mobile UIs, while ML is currently at its peak; so, what should the HCI expert do? Continue to work on AI within their field or "temporarily reinvent themselves as an ML researcher"?

• How Dinosaur Extinction Gave Us Fruit (PBS Eons) — Explains that, in the chaos of dinosaur-dominated jungles, flowering plants’ seeds were r-selected and thus small and tough. However, when the dinosaurs disappeared and forest ecosystems became more stable, plants could shift to K-selection, resulting in larger seeds with more fleshy coatings—fruit!

🔍🪢 Lens of the week

Introducing new ways to see the world and new tools to add to your mental arsenal.

This week’s lens: loose coupling.

Wikipedia is edited by thousands of people worldwide. They work asynchronously, often anonymously, yet the encyclopedia holds together. How? Loose coupling. Editors operate independently but align through shared protocols such as citation rules, formatting standards, and dispute resolution norms. The structure shapes coherence but doesn’t dictate every move.

Originally from software design, loose coupling refers to components that interact through defined interfaces, without dependence on each other’s implementation details. It’s a way to enable autonomy without fragmentation through alignment tools such as purpose, principles, and protocols.

Loose coupling supports adaptability and resilience. But it’s not a free-for-all; the structure matters. Wikipedia works not because it’s chaotic, but because shared commitments constrain its looseness.

When alignment mechanisms can be expressed as enforceable rules, loose coupling can be an alternative to hierarchy. Hierarchy makes it easier to drive to a specific outcome. Loose coupling enables systems to evolve, learn, and self-correct.

---

© 2025 The FLUX Collective. All rights reserved. Questions? Contact flux-collective@googlegroups.com.