Daily Briefing: October 2, 2026, evening · egregore

Listen: https://blossom.buildtall.systems/1c0191a60bfeb1fba57e878dd85116d273548f2b1a9aff84095cef08b1f3159f.mp3

The Daily Briefing for October 2, 2026, evening, covers fifty articles. The window begins at nine in the morning coordinated universal time and includes articles published through the evening.

A substantial group concerns automated license plate readers and the Flock camera network. Schneier on Security discusses Unidentified Flock Cameras in Florida. St. Lucie County discovered a dozen Flock cameras whose ownership it cannot identify and that the county government had not permitted. The author is reminded of the decade-old story of StingRay cell phone surveillance devices in Washington, D.C., whose operators were also unknown. The guess offered is that the StingRay devices were operated by foreign actors, while the Flock case is more likely some local government entity that did not bother getting approval. A foreign actor would rather hack the existing Flock network, as Israel did with Tehran's traffic cameras, than risk installing its own. The article concludes that once a surveillance infrastructure is normalized, both friends and foes will take advantage of it, and the risk becomes that everyone uses it.

The Record reports Bipartisan backlash to ALPRs grows as two high-profile bills are introduced. Republican Senator Josh Hawley has new legislation on limiting automated license plate readers, while Democratic Senators Bernie Sanders and Jeff Merkley, with Representative Alexandria Ocasio-Cortez, have introduced a broader bill.

The longest item in this group is 404 Media's Federal Judge Rules a Flock Search Was ‘Indiscriminate Mass Surveillance’ and Unconstitutional. A federal judge in Oklahoma ruled Thursday that a police officer violated the Fourth Amendment rights of a woman accused of meth trafficking when he searched her license plate in Flock's automated license plate reader system simply because her license plate was from California, then used her travel history as part of the reason to search her car. The judge, Sara Hill, wrote that the officer's use of the ALPR systems was an unconstitutional warrantless search, was not supported by probable cause, and was done without a warrant. 404 Media notes that there are more than a hundred thousand warrantless searches of the Flock system every month according to audit logs it viewed. Hill's decision will not set a binding precedent, and several other cases around the country are considering the legality of warrantless ALPR searches. The opinion argues that previous decisions saying Flock searches are not violations because they track cars in public do not consider that Flock's nationwide network is quickly approaching dragnet-type law enforcement practice. The facts involve a Tulsa County deputy sheriff who followed a Mazda with California plates for no apparent reason other than the plate, ran a Flock query, pulled the driver over for changing lanes without a turn signal, interrogated her about her travel, and used the Flock data as part of the justification to search the car. The search found ninety-one pounds of methamphetamine, but Hill ruled that all Flock evidence and all evidence from the search must be thrown out. She wrote that the ALPR search provided more than fifty individual records of the driver's whereabouts across the country for an entire month, and that the only reason for the search was the California plate. The decision comes after the Supreme Court's Chatrie decision found that police accessing a person's digital data, including cell phone location data, constituted a search. Michael Soyfer of the Institute for Justice told 404 Media that the opinion recognizes the sheer breadth of these systems and the sweeping modes of surveillance that allow police to reconstruct the movements of anyone in the country. Hill also wrote that previous rulings relying on United States v. Knotts, a 1983 tracking case, fail to account for the difference between that case's limited technology and a continuously updated location history for all vehicles. The article notes that Flock Safety did not immediately respond to a request for comment.

The second large thread concerns artificial intelligence as a political, regulatory, and public-safety issue. TechCrunch reports It’s not AI anymore, it’s ‘super intelligence’ (according to the White House). This week the White House gathered nearly every major tech CEO in one room, including Zuckerberg, Bezos, Musk, and Anthropic's Dario Amodei, to sign an AI safety pledge that President Donald Trump called morally binding. Trump also signed an executive order officially rebranding AI as super intelligence. The same article notes that Meta and OpenAI are putting friendlier faces on their AI products.

Slashdot covers Error-Prone AI Nearly Sparked World War Three Last Month. Timnit Gebru and Emily M. Bender argue in a Guardian op-ed that policymakers focus too heavily on hypothetical superintelligence while overlooking the more immediate danger of deploying unreliable AI in high-stakes settings. They point to an as-yet independently unverified CNN report claiming U.S. officials nearly intercepted a Chinese vessel after an AI-assisted intelligence report incorrectly said it was carrying nuclear-weapons components. The piece says military personnel made plans to intercept the Chinese vessel, complete with aircraft and soldiers ready to board it, and only moments before executing the plan did officials further investigate and find that the report had been generated with the use of a chatbot and contained erroneous information. The authors add that this incident did not galvanize lawmakers into regulation or prompt media warnings the way a rogue-AI extinction scenario might have.

Schneier on Security also carries How American Political Campaigns Are Using AI—and What They’re Spending on the Tools, written with Nathan E. Sanders and originally published in The Guardian. New campaign finance disclosure data shows which U.S. political campaigns are using AI tools and how much they are spending. Federal Election Commission itemized expenditure data going back to 2020 reveals at least seventeen million dollars in disclosed spending on AI technology vendors across five hundred twenty-three federal candidates and campaigns. Four states—California, Colorado, Massachusetts, and Washington—provide a localized picture going back to 2022. At least eighty federal campaigns and committees have reported spending with OpenAI since 2024, totaling only about fifty thousand dollars and skewing slightly more Republican than Democratic. The Republican National Committee is the largest overall buyer with nearly ten thousand dollars in reported expenses, and top individual users include Republicans Mike Lawler, John Kennedy, and Bill Cassidy, as well as California Democrats Ro Khanna and Ted Lieu. Most expenses are listed as office expenses, ChatGPT subscriptions for staff, or research tools. Anthropic has rapidly built a similar level of usage but with a different split: at least sixty-five candidates or committees now report paying the Claude maker in 2026, a nearly two-to-one Democrat-to-Republican ratio. The largest individual user is the campaign of Republican Senator Tom Cotton of Arkansas, with more than four thousand dollars in reported spend in his June filing. Other major users include Montana independent Senate candidate Seth Bodnar, Jason Knapp, who lost a Democratic House primary in Virginia, and Democratic Alaska Senate candidate Mary Peltola. About twelve percent of campaigns using either tool reported expenditures to both vendors. The article says the Democratic lean of Anthropic usage may reflect the company's alleged liberal skew and clashes with the Trump administration. Elon Musk's xAI, by contrast, caters to Republican interests, and its meager usage comes almost entirely from the political right: seven federal and two state-level candidates or committees have reported paying xAI about five thousand dollars total, most spent by the 2024 presidential campaign of Robert F. Kennedy Jr., with Dave McCormick and Thomas Massie also included.

Specialized campaign vendors attract larger sums. AmplifAI, which provides automated text messaging, soaked up 4.7 million dollars in campaign spending in the 2022 cycle alone, heavily used by Democrats Mark Kelly, Joe Biden, Bernie Sanders, and Adam Schiff. That spending has tapered off since 2022, and AmplifAI is now owned by Triller. The new rising Democratic solution for AI-powered text messaging is Daisychain, with about three hundred thousand dollars in reported candidate spend in the 2026 cycle, up from about fifty thousand dollars in 2024; more than half of this year's spending comes from the Senate campaign of Democrat Abdul El-Sayed in Michigan. The closest Republican equivalent is Campaign Nucleus, associated with former Trump campaign manager Brad Parscale, which has attracted six-figure spending from the Republican National Committee and multiple Trump-aligned PACs, plus five-figure investments from Mike Johnson, Kari Lake, and others. It is displacing the legacy Republican texting vendor Prompt.io, which has retained about three hundred seventy-five thousand dollars in 2026 spending, down from more than five hundred thousand in 2022. A More Affordable California, a PAC sponsored by Uber, spent more than one million dollars on Prompt.io in 2026. At the state level, the four states studied show at least ninety-two thousand dollars in spending confidently attributable to modern generative AI vendors since 2022, spread across at least one hundred eight candidates and committees; there has already been about ten times the amount of state-level AI spending reported in 2026 as in all of 2024. Daisychain again has the highest overall spend in the states, and OpenAI and Claude dominate among general-purpose AI vendors. DonorAtlas, an AI-powered prospect research tool, sits behind only Daisychain and OpenAI in these states, buoyed by nearly four thousand dollars from the California Democratic Party. Despite media attention on deepfakes, few candidates report spending on synthetic audio and video tools: ElevenLabs has six federal candidate or committee users totaling about fourteen hundred dollars, and Midjourney has five users totaling about sixteen hundred dollars. The article cautions that disclosure reports are only the tip of the iceberg: consultants, media firms, and other vendors are not required to disclose detailed expenditures. Wesleyan Media Project data shows at least one hundred sixty-four political ads in this cycle have included AI-generated media, supported by at least eighty million dollars in ad spending. The piece concludes that much remains to be seen about how campaigns will use AI in the final weeks before November.

Marginal REVOLUTION's A Normal Debate? reports that computer scientists Arvind Narayanan and Sayash Kapoor wrote AI as Normal Technology, describing the statement that AI is normal technology as a description, a prediction, and a prescription. Narayanan and Kapoor view AI as a tool that can and should remain in control of, and argue that this goal does not require drastic policy interventions or technical breakthroughs. AI responded in video format, and Kapoor created his own diss track using Claude, calling it extremely fun to convert scattered thoughts into a song and video. The post notes that both videos are good, even great, but that the sneering in the first video is sublime. The script, music, and lyrics were almost entirely AI. The author observes that it does not seem like a normal debate regardless of which side won.

TechCrunch reports Pope Leo XIV is not a fan of AI-generated art. The pope wrote that there is an ontological difference, even before an aesthetic one, between art and what a machine can generate through statistical calculation based on millions of images created by others, and that algorithms lack the spark of humanity.

Circuit Breaker Labs hopes to make AI safer for your kids (and you) is another TechCrunch item. With all the talk about how AI might one day harm everyone, the piece says it is easy to forget that AI has already harmed some people psychologically. Circuit Breaker Labs has created crash test dummies to address that harm.

Apple says it’s tightening macOS ‘Full Disk Access’ controls due to new risks from AI agents. Apple says it will add new controls around macOS's Full Disk Access permission, warning that increasingly capable AI agents make broad access to users' files, messages, mail, and browsing history riskier.

Slashdot covers Judge Dismisses Chegg and Penske Antitrust Lawsuits Targeting Google AI Search. A federal judge dismissed antitrust lawsuits from Chegg and Penske Media accusing Google of harming publishers by using their content in AI Overviews and other AI search features without compensation or a meaningful opt-out. Chegg claimed Google illegally scraped its educational content, allowing Gemini models to recreate it and reduce traffic. Penske, which owns Rolling Stone and Variety, filed a similar case alleging lost traffic and unfairness because sites indexed for organic search had content harvested for AI answers with no way to opt out. Judge Mehta wrote that plaintiffs pleaded only an expectation that Google will send them search traffic if they make their content available for free, and that an expectation is not an agreement. The judge noted that Google never had a formal arrangement with either plaintiff, so antitrust law does not apply. The article adds that Mehta also heard the Justice Department's long-running search antitrust case and found Google violated the law, but the government did not get the harsh penalties it wanted.

Stratechery by Ben Thompson offers 2026.40: Dots and Question Marks, a weekly roundup. The first featured item is Ben Thompson's Can Meta Focus?, which argues that everyone will use agents eventually, and that if they do, the impact on value chains will be enormous: something like Muse could aggregate Aggregators and become the most valuable product in the world. That bullish takeaway from Apps, Agents, and Aggregation was followed a day later by Meta's announcement of Meta Enterprise Platform, which Thompson says will not work and is a distraction from the biggest opportunity in consumer tech. The second featured item is What is OpenAI Doing? by Andrew Sharp, which says OpenAI's Tuesday Dev Day brought a return to chaotic messaging and branding, with new pricing tiers, overlapping products, and dots that look meant for mass-market consumers but are only available to pro subscribers. Thompson's update on Dev Day validated that confusion while attempting to identify an underlying method. The third featured item is Mao or Media Day?, covering Xi Jinping's visit to Washington. Sharp says the pageantry had an absence of substance, and that Beijing's posture looks much like Mao Zedong's second phase of protracted war, strategic stalemate, perhaps a prologue to phase three counter-offensive. The same section then turns to NBA media days on Greatest of All Talk. The roundup lists Stratechery articles and updates: Apps, Agents, and Aggregation; One More Note on Agents, Meta Connect, Meta Enterprise Platform; OpenAI Dev Day, Dot and OpenAI’s Product Transition, Sign In With ChatGPT; and An Interview with Jason Del Rey About Muse, Amazon, and Walmart. It also lists Sharp Text's Super Intelligence Is Not Right, Either, Dithering episodes Personal AI and Dots, Asianometry episodes on the birth of the DVD boom and EUV photomasks, Sharp China on the strategic stalemate era, and Sharp Tech on agents, dots, Google and Apple questions, Muse follow-ups, super intelligence, and the mandate of heaven. The week's video is Frontier Overhangs.

TechCrunch also reports Slovenia’s .si domain sees a surge in registrations after Trump’s ‘super intelligence’ order. The .si domain name is seeing unprecedented demand after President Trump's super intelligence executive order.

Cybersecurity incidents and related legal cases form another cluster. The Record from Recorded Future News reports 'Warlock' ransomware used in attacks on critical infrastructure in Portuguese, Spanish-speaking countries. The group is exploiting a variety of vulnerabilities impacting Microsoft SharePoint, according to a new report from Symantec Threat Hunter Team.

The same publication reports Mississippi mayor says ransomware incident led city to shut down systems. Government services were temporarily disrupted by ransomware in Vicksburg, Mississippi, and Mayor Willis Thompson said the FBI and other authorities are investigating.

404 Media's Podcast: The FBI Was Hacked. We’ve Seen the Data begins with Joseph's stories on the massive FBI hack. 404 Media broke that news and continued covering the breach of data on all FBI employees and their spouses. After a break, Jason tells listeners that a surveillance company is telling cops it wants to add facial recognition technology to Flock camera data. In the subscribers-only section, Jason and Joseph discuss Muse, Meta's new agentic AI, and how humans are actually behind some of the tasks. The episode notes include the stories We Hacked the FBI: Hackers Say They Have Data on All FBI Employees, FBI Hack Exposed FBI’s Own Hacking Unit, FBI Hackers Say They Won’t Publish Massive Trove of FBI Employee Data, and Surveillance Company Tells Cops It Wants to Add Facial Recognition to Flock Cameras.

The Record reports Judge dismisses spyware case brought by Salvadoran journalists targeted with Pegasus. The plaintiffs, who all worked for the independent Salvadoran news outlet El Faro, failed to convince the court that their case had jurisdiction in California, according to the judge's order.

Slashdot covers Used Processor Leads To Game Ban. Someone who installed a used AMD Ryzen 7 5800X3D found that several online games would not start because the previous owner was caught cheating. The problem appeared in Valorant from Riot, where the new owner was kicked out from the first start. Riot support confirmed the previous owner had cheated, and the processor is on an internal blacklist; with its installation, some other components like the mainboard are apparently also affected. Riot support does not remove such entries manually. In addition to Valorant, all other Riot games are affected, including League of Legends, Teamfight Tactics, Legends of Runeterra, and 2XKO, as well as likely future titles. The article notes that Riot's Vanguard anti-cheat system runs deep in the Windows kernel and starts with the PC, but a June 2026 update allows it to launch with a game if security features such as TPM and Secure Boot are enabled. It is unclear how Riot uniquely identifies processors, though possibilities include hardware fingerprinting based on chip behavior or storing identification hashes in the Trusted Platform Module.

Corporate and product news runs next. TechCrunch reports Rivian issues R2 recall for poorly tightened battery packs. The company says around fourteen vehicles may be affected and that it has resolved the issue on the assembly line.

TechCrunch also reports Rivian’s R2 just helped it set a new sales record. Deliveries jumped roughly forty-five percent compared to the same quarter last year, with Rivian delivering more than nineteen thousand electric vehicles.

Tesla sustains its EV sales momentum despite US troubles, also from TechCrunch, reports that Tesla delivered more than four hundred eighty-six thousand EVs in the third quarter, down from last year's record but moving in a positive direction.

Medical records giant Epic pauses product development to fix security bugs that risk patients’ data. TechCrunch reports the healthtech software giant, which makes the widely used MyChart system for accessing medical data, will focus on fixing security bugs for the next few weeks.

The Paramount and Warner Bros. Discovery merger is covered twice in the window. TechCrunch's Paramount and Warner Bros Discovery to become Skydance says the roughly one hundred ten billion dollar deal is expected to close October 6. Slashdot's Paramount and Warner Bros. Discovery to Become Skydance adds that the combined company will bring together Paramount+, HBO Max, CBS, CNN, MTV, and other networks, along with franchises including The Lord of the Rings, Game of Thrones, DC, and Yellowstone. Ellison said the Paramount and Warner Bros. brands will remain prominent, and that the new corporate identity is meant to give the combined company its own identity while allowing the brands to remain in the spotlight.

EU's Microsoft Teams Alternative Draws Bad Reviews From Officials is a Slashdot item. The European Commission has rolled out Element Pro as a self-hosted sovereign back-up to Microsoft Teams, and Politico reported that several officials questioned its usefulness and usability. One official said on an internal message board that the tool was absolute shit, another did not understand the added value, and a third said the United States would instantly win the war if it decided to switch off its technology. The Commission hosts Element on its own infrastructure rather than relying on Microsoft's cloud, giving greater control over communications data. Element CEO Matthew Hodgson said there will always be cognitive dissonance when switching to a new app, noted that the United Nations and NATO also use Element Pro, and said change is always painful when moving from a much-used program like Teams.

This is BiCS NAND Close Up from Kioxia and Sandisk is a ServeTheHome post reporting that the publication saw BiCS NAND under the microscope at FMS 2026 and thought readers might like close-ups of the packages.

Laytr’s new app lets you save anything you find online, not just articles to read. TechCrunch reports that Laytr saves articles, recipes, screenshots, videos, PDFs, and more for later, while keeping the archive private and synced across Apple devices.

TechCrunch also publishes several announcements for its events. Last 24 hours: Exhibit at TechCrunch Disrupt 2026 and reach 10,000+ tech leaders says today is the last day to book an exhibit table; the event takes place October 13 through 15 at San Francisco's Moscone West, with more than ten thousand founders, investors, operators, and tech leaders expected.

Less than 24 hours to apply for a Side Event at Founder Summit 2026 says applications close tonight at midnight Pacific Time. Hosting a Side Event connects applicants with the Boston startup ecosystem from November 1 through 7.

TechCrunch Disrupt 2026: Clay’s Kareem Amin on the rise of the GTM engineer announces that Clay co-founder and CEO Kareem Amin joins the AI Stage to discuss the rise of the GTM engineer. The notice adds a registration offer with a second pass at fifty percent off.

TechCrunch Disrupt 2026: Blackstone’s Jas Khaira on building the next generation of AI giants announces that Blackstone's Jas Khaira will take the Builders Stage to discuss building next-generation AI, with a similar registration discount.

Affected by layoffs? Don’t miss this $75 deal for your TechCrunch Disrupt 2026 Expo+ Pass promotes a seventy-five dollar Expo+ Pass for the first one hundred qualifying people.

The science and nature coverage spans climate, paleontology, marine biology, astronomy, immunity, neuroscience, and psychedelics. Slashdot covers 20% of Swiss Glaciers Lost In Five Years. Swiss glaciers lost 5.5 percent of their ice volume in the first eight months of 2026, bringing total losses since 2021 to 19.4 percent. Researchers say some glaciers are now collapsing from within as meltwater carves out large cavities. Matthias Huss, a glaciologist at ETH Zurich and lead author of the report, says many people do not realize how fast glaciers are retreating, or that one generation ago some now-barren sites sat beneath hundreds of meters of ice. He visited one last week and described a moonlike landscape where there was a big layer of ice just a few years ago, now a rock desert with scattered blocks of old ice melting on the hot rocks. The report found the mean thickness of individual glaciers fell by between 2.5 and 4 metres in 2026, and Huss said scientists are increasingly seeing glaciers collapse from the inside, the result of internal streams beneath the ice that excavate caverns, rather than simply retreating from surface heat. MeteoSwiss analysis found the summer half-year from April to September was the hottest and driest since records began in 1864, and over that time humanity has filled the atmosphere with 1.8 trillion tons of carbon dioxide and heated the planet by about 1.4 degrees Celsius.

Nautilus contributes several short science items. This Newly Discovered Dinosaur Looks Terrifying—But Probably Wouldn’t Eat You names the find as Kayrasaurus changliensis. There’s a New Sea Spider in Town opens with the observation that just when you think sea spiders could not get any more unique. This Spectacular Star Field Hides Brown Dwarfs That Never Ignite says the James Webb Space Telescope does it again. Your Immune System Ages in Two Big Bursts states that your forties and sixties are periods of intense remodeling.

What Goes on in the Brain When a Good Buzz Gets Harshed, also in Nautilus, reports that in 1974 Pfizer developed a synthetic cannabinoid called CP55940 that mimicked THC but was around forty times stronger. It was never marketed, but is now used to test the effects of THC on lab animals. Neuroscientists from Northwestern University dosed mice with different levels of CP55940 and placed them in an enclosure with a petri dish of fox urine. Control mice approached the dish, sniffed it, and ran away, while mice under the influence resisted the urge to investigate and tended to freeze more often. The study appears in Nature.

This Obscure 1960s Guide to LSD Holds Up, also from Nautilus, reports that a survey of more than five hundred private psychedelic users finds real utility in a sixty-year-old pamphlet on safe tripping penned by a pioneer of harm reduction.

Weekly Dose of Optimism #213 from Not Boring by Packy McCormick is a longer digest with several stories and a sponsor. The cover image was generated by Runway, and the post announces a partnership with Runway to bring the newsletter's stories to life over the coming quarter. The sponsor Reducto is described as a document parsing product used by AI teams at Scale AI, Harvey, Toast, and Vanta, with a new frontier parsing model called r-1 that reduces parsing errors by up to twenty percent and costs one cent per page. The first story reports that Starship reached orbit for the first time on Monday and deployed twenty-six Starlink V3 satellites. An engine shut down early during ascent, and SpaceX shortened the mission as a precaution, bringing the ship down in the northern Pacific, but the newsletter says the flight keeps getting better. The second story covers the National Design Studio's unveiling of America.gov, a government website built under Joe Gebbia that asks questions and grounds answers in government sources, drawing on more than twenty-nine thousand government websites. The third story reports Matic's master plan for a family of home robots, starting with useful single-purpose robots already in sixteen thousand homes, then adding manipulation, as shown in a video with a non-humanoid picking items off the floor and a humanoid folding laundry. The fourth story introduces Freckle, a two hundred dollar phone for kids with a twenty-five dollar monthly subscription, running ClayOS with parent-approved calls and texts, maps, AI questions, and no open web, social media, or app store. The fifth story reports on transplanted hearts adjusting to the host's biological age: a Harvard and Brigham and Women's Hospital study found older hearts showed signs of rejuvenation in younger bodies, while younger hearts showed accelerated aging in older ones, based on mouse experiments and biopsies from eleven human heart-transplant recipients. The post also notes extra doses on science breakthroughs, AMD buying World Labs, and General Intuition and Varda raises.

Several longer essays and commentary pieces remain. Modern vanity, from Seth Godin's Blog on marketing, tribes and respect, argues that vanity is rooted in emptiness, and that modern vanity is amplified because culture has hooked the toxic emotion of shame to vanity. The new version is not about being the prettiest, but about avoiding being not-pretty. That shame can be extended into not-competent, not-respected, and not-fit; social media numbers are called vanity metrics for a reason. The piece calls this selfish, not as hoarding but as the defensive decision to hold things back and sabotage process and practice so one is not exposed. It argues that aspiration has a ceiling, but avoidance has no floor. Society industrializes this threat into self-policing, the piece says, and the answer to what a person would ship if they did not care about blame or credit shows what they are withholding and from whom. Disappearing into apparent safety is not safe, but an empty and selfish way to make yourself small.

Make tmux the OS, from matduggan.com, is a long design essay prompted by a Scott Jenson talk called Are we really going to use the same Desktop UX forever? The author says Jenson argued Apple and Microsoft are not going to innovate much more in desktop OS design, leaving it to others to decide the future desktop. The author adds that the situation is more nuanced, with the large companies still trying ideas but being extremely conservative. The proposal is to take the scrollable, session-persisted, detachable, task-oriented experience of tmux and make it usable by a normal person.

The essay breaks the problem into components. Screen real estate constantly changes as the author moves between laptop and monitor about six times a day. The OS treats a display change as a surprise rather than a scheduled event, and when the laptop is undocked, windows must be dragged and resized by hand. People also use windows in different ways: maximizers, near-maximizers, and coordinators. The author says the three types, along with a public-private window split, were measured in 2004 and are still not accommodated. Browsers are mini operating systems, with tabs carrying the same complexity as applications, and the filesystem is an increasingly weak concept because notes, messages, and chat content live inside applications rather than the operating system.

The essay surveys earlier attempts. Rooms, from 1986, treated the screen as RAM and windows as pages, and described reducing knowledge faulting in the user. WindowScape dropped explicit grouping and recorded photographs of window arrangements, allowing one window to appear in many photographs; the catch was that photographs evaporated. Tiling window managers solve overlap but are too hard for normal people, and a scrollable tiling manager such as Niri already exists as a technical foundation. Arc and ChromeOS moved the browser toward being an operating system, which the author considers a fundamentally flawed direction. The graveyard includes Windows Timeline, Windows Sets, macOS Stage Manager, KDE Activities, and installed PWAs. The pattern is that the ideas were not wrong, but were not defaults and were not done at the OS level where they can see all apps.

The author then proposes a desktop as one infinite canvas with each physical display as a viewport. A transition contract would keep horizontal order, scroll position, input focus, and task membership unchanged, while column widths reflow deterministically. The OS would learn which monitor is focal from input-focus telemetry, and privacy would become a property of a window or tab, driven by machine-observable facts such as connected displays or capture sessions, never by inferred human states like attention. Browser tabs would detach into real windows and join tasks. A local LLM would operate through a quake overlay or ask box, querying a constrained MCP server and previewing proposed layouts; the model would never touch the window manager directly. The author notes that novelty is rare, so classification and preview run rarely while deterministic replay handles the rest. Spatial memory matters because epistemic action research shows that a window layout is thought in progress, so the system should append rather than replace. The author tried to build a Linux demo and found deep problems: the privacy-minded design runs on surveillance-shaped data, the favorite privacy feature hid content from its owner while showing it to the threat, and the infinite canvas never ages out, creating an overwhelming electronic desk. The essay concludes that the graveyard is full of good ideas that died of distribution, and that the overlapping window design was a hack around hardware limitations that have long since disappeared.

Teaching Paradox: Hearts of Iron IV, Part IIIa: Focusing on Politics, from A Collection of Unmitigated Pedantry, is the first half of the third part of a series on Paradox Interactive's World War II grand strategy game. The author states that the series now turns critical as it moves from the operational simulation to the rest of the game, beginning with domestic politics. The essay is framed with updates from a United Kingdom campaign, including naval battles in the Strait of Malacca and offensives through the Netherlands and into Germany. The thesis is blunt: the game's simplified politics present an essentially authoritarian vision in which the opinion of the people matters nearly not at all, state power shapes outcomes easily, and democracy comes only with drawbacks. The author says governments in the game do not fall and cannot be overthrown; they can only decide to abolish themselves, and the model is fascist in that politics flows down from the state and people exist merely as units of state power. The developers likely did not intend this, but the simplification produces it.

The article walks through the mechanics. There are four ideologies—democratic, communist, fascist, and non-aligned—plus largely cosmetic sub-ideologies. Democracies are limited in declaring war and gated by world tension, and guaranteeing other countries lowers world tension, which is mechanically bad for them. The author argues the game does not simulate the strengths democratic governments displayed, particularly adaptability and economic mobilization, and gives the example that the United Kingdom outproduced Germany in aircraft despite its smaller population and territory. Popular ideology is compressed: the United States is ninety-nine percent Democratic with Republicans and Democrats merged, and the United Kingdom's Tories and Labour are fused into one Tory mega-party. The non-aligned ideology is a grab bag covering monarchies, oligarchies, colonies, and Japan.

The author says political change is extremely simplified. Hiring one advisor can move support for an ideology at 0.10 percent per day, taking a country from no support to majority support in less than a year and a half. Stability and war support are mostly static and rarely produce crises. The focus tree, not the political system, is where the big decisions actually happen. Every country has a focus tree, and each focus takes about seventy days and costs political power. Branches cover industry, military development, and political paths. A player who switches ideology is effectively the government choosing to undermine itself, preparing a civil war it intends to lose.

The article borrows the term meme politics from Samantha Hancox-Li, defined by three assumptions: every policy does exactly what it intends, politics is straightforward and information is perfect, and policy faces no meaningful resistance. The focus tree operates this way, the author writes, with industrial plans never failing and no blowback except where clearly marked. The example of Communist China's Agricultural Reform and the Great Production Movement is cited as awarding bonuses with no Great Leap Forward-style failure. The Imperial Federation path for the United Kingdom is used as the clearest illustration: a player can hold an Imperial Conference, raise relations to one hundred by clicking improve relations, guarantee acceptance from the Dominions, and annex Canada, Australia, New Zealand, South Africa, and the British Raj into a single unitary state. The author says this would have been probably impossible at any level of effort in the 1930s and 1940s, and the game asks the player to consider none of the questions of representation, let alone the Indian independence movement.

The conclusion is that Hearts of Iron IV presents an authoritarian model of politics and grants the question to the authoritarians by making fascist and communist governments mechanically better, with democracy a roleplay choice that accepts limits without real benefit. The author contrasts this with the actual war, in which democratic economic mobilization worked, and the downsides of autocracy were on display. The simplification does violence to the underlying history, and the author says next week the series will turn to occupation, bombing, and war crimes, where it will not get better.

Terence Eden’s Blog reviews Gadget Review: Una Watch ★★★★☆. The author says he has never been a fan of smartwatches, but chose the Una Watch because it is designed in Scotland, charges over USB-C, is almost completely open source, and is repairable by unscrewing and replacing components. He bought one second hand, and the first charge from empty to full took a little over an hour at about 0.84 watts. The official Android app, though not open source, worked on GrapheneOS and allowed app installation, with restart required after each install. The screen is an always-on memory-in-pixel display in orange, blue, and white, without touch input; four buttons control the watch. Notifications initially required unpairing, repairing, and reinstalling, and the watch has no notification history and no per-app filtering by default. Gadgetbridge support is excellent for heart rate and notifications, though it cannot install apps or set alarms. Plugging the watch in over USB-C exposes 3.5 gigabytes of exFAT storage and JSON files. Battery life after a full day was around ninety-five percent. Development is possible with an open-source SDK and community board, though the GUI designer only works on Windows. Downsides include no payments, microphone, speaker, replies to notifications, or maps; the firmware is not yet fully open. The conclusion is that the watch is decent value at around one hundred eighty pounds, or one hundred twenty used, for tinkerers who want open and repairable technology.

Marginal REVOLUTION's Friday assorted links lists eight items: how rich Anglo-Saxon England was; a new game theory paper on AI races; Scott Sumner movie reviews, with the note that Sumner is always correct and therefore the greatest film critic in the world by that metric; a discussion of what the AI safety movement should be; a claim that frontier models are beating licensed accountants; a claim that somewhere between two and six cents of every dollar of World Bank foreign aid disbursements got siphoned into crypto wallets; Imbue Studios; and the return of war in Ethiopia.

Supply is elastic, installment #1637, also from Marginal REVOLUTION, reports on a study from the latest issue of the American Economic Review using Scandinavian administrative data. The study finds significant migration responses among the wealthy to wealth taxes: a one percentage point increase in the top wealth tax rate decreases the stock of wealthy taxpayers by about two percent. A large fraction of the wealthy are business owners, and their businesses are negatively affected by owner out-migration. The aggregate effects are nevertheless modest, with employment reduced by 0.02 percent, investments by 0.07 percent, and value added by 0.10 percent. The authors are Katrine Jakobsen, Henrik Kleven, Jonas Kolsrud, Camille Landais, and Mathilde Munoz.

Slashdot covers Social Media Harms Democracy By Spreading Rumors, Survey Shows. Pew Research Center found a growing number of people think social media manipulates and divides people and harms democracy. Companies including Meta, Snap, and TikTok have faced lawsuits and settlements, while Australia has proposed rules to let users opt out of algorithmically recommended feeds after becoming the first country to ban social media for children under sixteen. Eighty-two percent of Australian adults support that restriction. A majority of respondents said social media informs people and gives them a voice in politics, but there was a broad rise in those who now see social media as a bad thing compared to 2022. People in wealthier countries, including the United States, Canada, and the United Kingdom, are more inclined to believe social media harms democracy than respondents in Ghana, Kenya, and Nigeria. In the United States, sixty-four percent of adults say social media harms democracy. The share of people who consider social media bad for democracy increased in sixteen of twenty-six countries where older data was available. Pew surveyed more than forty-five thousand people in thirty-seven countries and an additional five thousand five hundred eleven U.S. adults.

404 Media's Behind the Blog: Freaking Out discusses how a few of the outlet's top stories came together, including inbox slop, Claude cults, and more. Joseph writes that for months he has noticed many emailed tips are clearly written by AI, because the responses keep using words that AI believes apply to journalists, but real journalists do not speak that way. He distinguishes his observation from Jason and Emanuel's coverage of iLands, the AI agents that email people offering to do work.

Life on a hair trigger, by Megan Kang in Aeon, carries the subtitle that a world of violence sustains itself through guns, poverty and other social ills, but also through the minds it creates. The piece directs readers to Aeon for the full essay.

The Lunduke Journal of Technology reports Linux Market Share Plummets! From 15% Down to 3% in One Month! The article's substantive claim is that those fifteen percent Linux desktop market share numbers were bots. The post also carries a subscription promotion for the Lunduke Journal.

Across the window, several tensions recur. The superintelligence rebranding and safety pledges sit beside reporting on error-prone AI in a near-military incident, AI-generated campaign tools, and a papal objection to machine-made art. The Flock and ALPR group moves from unidentified cameras and new bills to a federal decision finding one search unconstitutional. The corporate items mix recalls and sales records with a security pause at Epic and a major media merger. The science and health material is largely short, while the long essays return repeatedly to the difference between simple models and the real world.

Your signing session ended