🌀🗞 The FLUX Review, Ep. 241
July 2, 2026
Episode 241 — July 2, 2026 — Available at read.fluxcollective.org/p/241
Contributors to this issue: Neel Mehta, Boris Smus, Erika Rice Scherpelz, Dart Lindsley, MK
Additional insights from: Ade Oshineye, Ben Mathes, Jasen Robillard, Jon Lebensold, Justin Quimby, Lisie Lillianfeld, Robinson Eaton, Spencer Pitman, Stefano Mazzocchi, Wesley Beary, and the rest of the FLUX Collective
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.
“Not everything that is faced can be changed, but nothing can be changed until it is faced.”
— James Baldwin (1962)
🎯🚘 Point, don’t steer
Given a task to build a treehouse, would you rather have dozens of pages of detailed schematics or a simple one pager outlining what someone should feel when they are in the treehouse?
Well, like many things, it depends. Clarity is about the goal (why this treehouse matters, what good looks like for this treehouse). Meanwhile, detail focuses on what to do and how to do it (how do we build this treehouse?). The trap for many is that detail often feels like it increases clarity, but often the opposite happens. When the intent is buried under a hundred specifics, people lose the plot. They build exactly what was described without understanding whether or not it solves the problem.
The military tradition of mission command tells officers the objective and the constraints and leaves the how to the people closest to the fighting. Its opposite, detailed command, scripts the steps from the top. Mission command wins more often not because soldiers are smarter than generals, but because they see the situation on the ground.
Detail is not always the enemy. Sometimes, a single centralized logic needs to be scaled to more than one person’s worth of execution. If an engineering system is having an outage, the incident commander needs to be able to quickly make the calls and have someone execute them. There’s no time to give high level goals and iterate for alignment. A startup three weeks from insolvency is in a similar situation. These situations tend to appear when the stakes are existential and time is scarce.
But when you’re not in those situations, detailed mode becomes a liability. Every decision routes through one person, so the people around them stop exercising judgment. Which means that now the leader has to stay involved, reducing judgment even more. It’s a feedback loop that drops an organization’s throughput to that of the bottleneck. And worse, the leader might not realize that they’re the bottleneck. Instead, they blame it on the lack of autonomy of those around them.
Detail can be valuable, but we need to figure out when and how to present it. Interface design calls this progressive disclosure. First, set the clear goal. Reveal detail on demand, when someone hits a wall. A new hire will need a lot of it; tenured engineers, almost none. The detail arrives as an answer to a real question, not a preemptive script.
As we provide guidance, we can think “does this guidance provide clarity? or does it add detail?” And if it adds detail, “Was this detail requested?” We might find ourselves realizing that clarity is, in fact, the harder job.
(Credit to Matt Manela for inspiring this piece.)
🛣️🚩 Signposts
Clues that point to where our changing world might lead us.
🚏💱 Wealth managers are moving upmarket, leaving non-millionaires to AI
A senior partner at McKinsey observed that “mass-affluent” clients, defined as those with liquid assets between $100,000 and $1 million, can now get “private-banking quality from AI,” meaning that financial advisors who focused on standardized advice for the mass market are now a lot less valuable. Instead, he predicted that wealth managers will shift to serving even richer clients, since those people often have emotional needs (such as which child will get the largest inheritance) that require a human touch. Banks such as Citigroup have actually stepped up hiring for wealth advisors, but the roles are shifting now that internal AI tools can do a lot of basic portfolio analysis. The McKinsey partner predicted that firms would need to start hiring for AI experts, “behavioral data scientists, personalization architects, and human-in-the-loop oversight professionals,” all hybrid roles that didn’t exist in wealth management just a few years ago.
🚏🎰 Influencers promoting a prediction market “won” millions on a fake betting site
A duo of exposés from Politico and the Wall Street Journal found that the prediction market startup Polymarket paid influencers to post videos of themselves winning hundreds of thousands of dollars (or more) betting on things like what a politician would say in a speech. It turned out those bets had never actually been placed, and if you’d actually placed the hundred-plus bets featured in the ads, you’d have lost $166,000. It turned out that Polymarket devs had built a fake lookalike website, “poiymarket.com” (with an ‘i’ in place of the ‘l’), for influencers to post these mock trades. What’s more, Polymarket’s CMO used his personal PayPal account to send over $2.5 million to political influencers, most of whom never disclosed that they were doing a paid partnership.
🚏🛻 A dead-simple electric truck will sell for $25,000
The average new car in the USA sells for $50,000; even used cars sell for upwards of $25,000. But an EV startup called Slate has revealed a bare-bones electric pickup truck that will sell for just $24,950 (not counting fees or taxes). The base model has hand-crank windows, basic gray paint, no infotainment system, and just 205 miles of range, but it’s modular and allows buyers to add things like roof racks, upgraded headlights, or a back section that turns it into an SUV. While the US is indeed lacking in affordable EVs, it remains to be seen if there’s market interest for a low-end pickup truck in a country where the average new pickup sells for over $66,000.
🚏🗳️ An NYC politician was arrested for forgery over AI-generated photos
A politician from Queens who ran for state assembly last year was arrested for forgery after it was revealed that he’d used AI to create fake endorsements, photos, and news articles. He made a social media post claiming the endorsement of the Queens Jewish Alliance, using their logo on a realistic looking (but AI-generated) endorsement sheet; he also swapped his face into a picture of a prominent local politician shaking a man’s hand, making it appear that that politician supported him. In the complaint documents, the Queens District Attorney added that this man had deepfaked photos of his opponent and generated AI videos “appearing to show endorsements from a police precinct and an elementary school.”
📖⏳ Worth your time
Some especially insightful pieces we’ve read, watched, and listened to recently.
Your AI Is Not a Tool (The Convivial Society) — Applies a McLuhanist lens to argue that AI, like most modern digital technologies, isn’t really a tool. Tools exist outside of us and channel certain skills, and they can be picked up and put down at will. But AI is an environment that envelops us and shapes us; you can’t escape it or choose not to engage. This line of thinking refutes the cliché that “technology is just a tool, and what matters is what we do with it.”
Why Flying Within Africa is Harder Than Flying Out of It (Proxima Imperium) — Argues that the lack of competition (and resulting high prices) of intra-African flight routes is due not to corruption but to a collective action problem. Countries signed agreements to allow airlines to freely cross national borders, which would stimulate competition and drive down prices, but each individual country has an incentive to defect and lock down its airspace to prop up its (often struggling) state-owned flag carrier. Thus, good treaties like the Yamoussoukro Decision are enacted de jure, but everyone slow-walks the implementation, leaving the plans as good as useless.
AI Isn’t Management. Try Explaining That to Matthew Prince. (Programmable Mutter) — Political scientist Henry Farrell examines a tech CEO’s “self-congratulatory op-ed” where he talked about growing his business while laying off 20% of its employees; the essay cited the writings of famed management consultant Peter Drucker to support its argument that middle managers (“measurers”) can be eliminated. Farrell argues that Prince misread Drucker; Drucker says that technology can and should be used to automate menial work and free employees up to make higher-level judgments and strategic decisions. This vision would actually lead to more mid-level management, not less, and Drucker would say it’s a good thing because it helps develop humans’ capacity and leads them to self-actualization.
Iran’s Ultimate Banned Book (Amir Ahmadi Arian) — Examines Iranian writer Sadeq Hedayat’s 1936 masterpiece The Blind Owl, which holds the rare distinction of being banned both before and after the Iranian revolution, surviving in a form similar to Soviet samizdat. The hallucinatory, opium-fueled novel follows a pen-case painter who descends into a delirious oblivion, ultimately transforming into the very Quran-reciting old man he despises. Hedayat deeply loved his homeland yet routinely reviled it in letters as “Stinkistan.” Named for a bird that portends catastrophe rather than wisdom in Persian culture, the novel offers an absolute, uncompromising pessimism.
🔍📅 Lens of the week
Introducing new ways to see the world and new tools to add to your mental arsenal.
This week’s lens: prognosis vs. diagnosis.
“You’ll be fine in two weeks,” is a reassuring thing for a doctor to say, but it leaves out what’s actually wrong with you. It is a prognosis — a prediction about where you end up. It says nothing about the diagnosis, which is what’s actually causing that outcome.
Physicians used to be excellent prognosticians and useless diagnosticians — and not that long ago, either. This is at first surprising, since the future is naturally less certain than the present. In 1850 a skilled doctor could often tell you whether you’d survive a fever, without the faintest idea of its source — or worse, with certainty that the fever was caused by humoral imbalance, not bacterial infection.
The two reduce different kinds of uncertainty. Since a prognosis forecasts the outcome, a good one is confident, specific, and correct (often enough). Diagnosis goes after the cause and can identify levers, directing us toward how to intervene.
Sometimes we mistake the prognosis for the diagnosis. “This project will slip three weeks” tells us something valuable, but it doesn’t say why. Maybe a handoff broke. Maybe nobody owned it. You can become a brilliant forecaster of dysfunction and remain completely unable to treat it or, even worse, you can end up treating the symptoms. Adding more status meetings or more staffing to a late project may very well be the blood letting of the organizational world. Bring me the leeches!
Next time something’s wrong, ask: Do I need a prognosis? diagnosis? both? (But, unlike with germs, when you’re working with humans, remember to engage in connection before correction.)
© 2026 The FLUX Collective. All rights reserved. Questions? Contact flux-collective@googlegroups.com.