Alvaro Lopez Ortega / 2026-09-12 Briefing

Created Sun, 13 Sep 2026 05:49:36 +0000 Modified Sun, 13 Sep 2026 05:50:08 +0000
6299 Words

Revolut confirmed it handed sensitive customer data to scammers using a fraudulent government email domain, but said funds and systems were unaffected. The Atlantic has recorded no hurricane through Sept. 12, a 60-year record, possibly due to El Niño. Sam Altman and Dario Amodei agreed the AI industry must “pace the frontier,” with third-party safety evaluators.

🤖 AI & Machine Learning

Nobody wrote the Matrix and AI code-writing will be dead by 2036

The author traces a lifelong programming fascination from Commodore 64 commands to the question of whether all of reality could be simulated in code, concluding only machines could handle such complexity. He connects this to The Matrix and to modern LLMs and AI agents that generate code and build applications from conversation, arguing his childhood vision of machines programming complex systems is now becoming plausible. The piece also predicts that AI code-writing will be dead by 2036.

AgentsDock: An IDE designed for agentic AI research

AgentsDock is a macOS 14+ IDE for agentic AI research that brings Claude Code, Codex, and Cursor into one desktop and mobile workspace. Users install an AgentsDock server, connect via server URL, and access multiple remote servers and agents from any device with code editing, terminal/tmux access, and inline job outputs. It is aimed at AI researchers who want agents to help with research and model training across machines.

Real-SWE: Benchmarking AI models on private, real-world, enterprise codebases

Real-SWE is a new benchmark that evaluates frontier AI models on private, real-world enterprise codebases, using tasks from licensed production systems with company-specific complexity and business consequences. It tests model-and-harness combinations through native harnesses, with Fable 5.1 using Claude Code leading at 38.8% resolution, followed by GPT-6 Astra with Codex CLI at 33.8% and Gemini 3.8 Flash with Gemini CLI at 31.2%. The benchmark aims to assess whether coding agents can perform actual software engineering work in enterprise environments.

Benchmark: CadQuery vs. OpenSCAD for agentic CAD work

A controlled benchmark had six AI agents model three functional 3D-printable parts in CadQuery and OpenSCAD, with every resulting STL independently verified. All six parts were printable, so raw capability was not the differentiator; the tools mainly diverged in how they failed. The hardest task, a true-helical threaded hose-barb adapter, was solved by both tools.

Anthropic is no longer a frontier lab

The provided page titled “Anthropic is no longer a frontier lab” contains no article text, only a browser verification prompt asking users to reload or update their browser. No factual details or claims from the article can be summarized.

AI Is Powerful Enough to Crack Our Hardest Math Problems–and Kill Us All

OpenAI has published a solution to one of the notoriously complex Millennium Prize Problems, underscoring AI’s rapidly advancing mathematical capabilities. The development has prompted mathematicians and AI researchers to ask why AI’s sudden excellence in math matters and what its broader implications may be.

DeepSeek v4.1 flash runs 23 seconds/token on a 2020 16gb M1 Mac Mini

The provided content is only a browser verification page and does not include the article body. The headline claims that DeepSeek v4.1 Flash runs at 23 seconds per token on a 2020 16GB M1 Mac Mini, but this cannot be verified from the supplied text.

Hugging Face’s Open Alignment Initiative seeks to join Amodei’s ’embedded evaluators’ program

Hugging Face launched the Open Alignment Initiative, led by co-founder Thomas Wolf, arguing that AI alignment cannot be solved only behind the closed doors of a few frontier labs. The initiative is seeking to join the “embedded evaluators” program committed to by Amodei.

Useful Things Agents Can Do That Are Not Writing Code

The article explores non-coding uses for AI agents, including fixing PR merge conflicts, finding issues relevant to a branch or PR, and diagnosing failed GitHub Actions runs. It shares example prompts for these tasks and advises adapting them rather than copying the author’s workflow.

🔒 Security & Privacy

Revolut confirms it handed customer data to scammers posing as government agency

Revolut confirmed it disclosed sensitive customer data to scammers who submitted fraudulent requests using a legitimate government agency email domain. The exposed data reportedly included contact details, birth dates, occupations, identity documents, account statements, transaction histories, IBANs, and Bitcoin transactions. Revolut said it blocked the address, alerted authorities, and that a limited number of customers were affected while customer funds and systems remained unaffected.

Zoom Linux Client Proactively Reads X11 Clipboard

Reports claim that Zoom’s Linux client proactively reads everything written to the X11 clipboard. However, the available sources provide no further substantive details, as their article bodies only show a Mastodon JavaScript/native app notice.

Reddit Now Blocking Firefox for Android with uBlock Origin?

A user reports that Reddit is now blocking sign-in on Firefox for Android when uBlock Origin is enabled, displaying an error that asks users to disable extensions or switch browsers. This follows Reddit’s earlier move to block signed-out access to old Reddit.

Trail of Bits verifies Signal chat integrity

Signal has launched Automatic Key Verification, a key-transparency feature that detects compromised servers returning false public keys without requiring manual safety-number comparison. Trail of Bits operates one of three auditors—alongside Signal and Cloudflare—that maintain and sign consistent Merkle-tree heads of Signal’s user-to-public-key map, with clients requiring signatures from all three and warning users if valid endorsements are missing. Users can enable the feature under Settings > Privacy > Advanced and use “Verify Automatically” in supported chats.

The Worst Spam Emails: Inside iLands’ AI Agent Hustle

A writer received over a dozen spam emails in three days from iLands.app AI agents, including one named Leo Ashford, pitching research services for about $25. iLands calls itself a “Human-agent network” and effectively operates as a Fiverr for autonomous bots, with Kaixin Tang named as founder. The writer, a freelancer, described the solicitations as offensive attempts by bots to take human work.

WeWorm: Zero-Click WeChat Worm

Calif released a demo of WeWorm, described as the first zero-click worm to spread through WeChat calls on iOS and Android. By calling a victim, it can hijack their WeChat account without interaction and use it to call their contacts, enabling phone-to-phone propagation. The exploit requires the attacker to be on the victim’s friend list, and Calif says AI-assisted work produced the RCE exploit in about two days and the worm in a week.

google.com/goto: Google’s anti-scraping update

Google Search has begun rewriting organic result links to google.com/goto?url=… for logged-out and private sessions, hiding destination URLs behind an opaque, custom-encoded parameter that must be resolved via the Location header. The change, distinct from the older google.com/url format, is an anti-scraping measure intended to make bulk SERP harvesting by AI crawlers and SEO scrapers slower and noisier. Autom.dev has updated its Google Search pipeline to resolve goto links by reading Location without following redirects, returning final destination URLs in existing structured fields.

OpenAI: “We use … de-identified data to improve ChatGPT”

The provided page contains only a browser verification screen and no substantive news article. Its title states that OpenAI says it uses de-identified data to improve ChatGPT.

The gpg.fail aftermath: On responsible disclosure, GPG, and the state of security in 2026 [32:37]

A follow-up talk recounts the 2025 disclosure of multiple GNU Privacy Guard (GPG) vulnerabilities, including PGP signature spoofing and memory corruption in the message parser; while some were fixed, others remain unpatched after GnuPG’s maintainer dismissed a widely used feature as “harmful.” The talk demonstrates remaining footguns, presents novel GPG bugs, and discusses responsible disclosure and AI/LLMs in security in 2026.

Rare Not Random Using Token Efficiency for Secrets Scanning

The article examines improving secrets scanning beyond regex-and-entropy filters, arguing that secrets are not just random but statistically rare compared with natural language and source code. It proposes using Byte-Pair Encoding tokenization as an alternative primary filter, since BPE compresses common words into fewer tokens while splitting rare or unnatural strings into many short tokens. This token-efficiency signal could help reduce false positives in generic secret detection.

🛠️ Software & Hardware

Will There Be a 7G?

A new paper asks whether 7G will or should exist, arguing that it must be justified by post-6G needs and coordination problems that 6G/6G-Advanced and other wireless or software-based systems cannot meet, rather than by numbering or more ambitious radio targets. It proposes a readiness framework covering demand-led need, system-level discontinuity, coordination value, sustainability and circularity, trust, and geopolitical viability, then applies it to candidate 7G discontinuities. The goal is a structured basis for deciding whether 7G becomes a distinct mobile generation, an extension of 6G evolution, or a broader post-6G infrastructure fabric.

Intel 8087 Scale Instruction Microcode

Intel’s 8087 floating-point coprocessor, released in 1980, helped standardize floating-point arithmetic, accelerated PC floating-point operations, and influenced later computers. Reverse-engineering its microcode shows that the seemingly simple FSCALE instruction requires over 140 micro-instructions and three levels of subroutine calls to handle many special cases, while revealing details of the chip’s datapath—including its shifter, adder, and exponent converter—and a hidden feature.

Build profiler for Bun compile times

The author created buildprof, an open-source Linux build-tracing/profiling tool that records spawned processes and subprocesses on a timeline to expose bottlenecks such as poor parallelism, repeated work, dependency downloads, and heavy compiler/linker calls. Motivated by claims that Bun’s new Rust build was over 5× faster than its old Zig build, the author reproduced the timings on a 6-core Linux VM and investigated possible causes, including Full LTO versus ThinLTO.

I refuse to let SPICE die

A community-maintained public mirror of the abandoned freedesktop.org SPICE Windows guest agent is keeping the project alive, with canonical downloads on GitHub Releases. The agent provides client mouse mode, desktop resolution matching, clipboard sharing, file transfer, and a Windows service, and includes a multi-GPU mouse fix for current guests such as Windows 11 VMs on Linux. It is licensed GPL-2.0-or-later and continues to build and ship for modern Windows guests.

Programming Language Design Ideas

The articles praise three programming-language features: flow typing and type narrowing in Crystal and TypeScript, Rust’s borrow checker for compile-time data-race prevention without a garbage collector or runtime cost, and D’s contract programming for assertions, enforce, and pre-/postconditions. They argue these mechanisms improve safety and expressiveness through compile-time checks and explicit invariants, with manageable tradeoffs.

Usenet rewind archive search engine

Usenet-Rewind is a search engine and research archive for Usenet newsgroup text posts from 1981 to the present. It contains over 1 billion actively populating messages covering early tech support, academic discussions, hobbyist communities, and firsthand internet history.

Google stole open source code without crediting the authors (Artemis/Minitap)

Minitap alleges that Google’s Artemis project contains code from its open-source mobile-use project, including identical Android device code, Hopper agent instructions, and WhatsApp examples. It also says an earlier Artemis package file listed three Minitap authors, but a later version removed their names and substituted another author. According to Minitap, the Artemis README did not credit mobile-use.

Graphify C# – Compiler-accurate Find Usages for coding agents

Graphify C# is a free, headless Roslyn/MSBuild indexer that provides coding agents with compiler-accurate “Find Usages” for C# by extracting semantic evidence such as callers, references, implementations, inheritance, and overrides across overloads, generics, and projects. It outputs a deterministic JSON graph of nodes, edges, and hyperedges from .sln, .slnx, .csproj, or SDK file-based inputs without requiring an IDE, compiled project DLLs, or a database. The MIT-licensed tool is installable as a .NET global tool and is designed for agents like Codex and Claude Code.

Dell’s 52-inch enormo-monitor is a tsunami of impractical fun

Dell’s 52-inch mega-monitor is an impractical but fun showcase of display technology. Most users don’t need it, but it impresses as an example of what is technically possible.

Lua Pattern Tester

The article titled “Lua Pattern Tester” states that JavaScript must be enabled to run the website. Its summary and content both repeat this requirement.

A Dick Smith VZ200 without the Dick Smith

The article describes restoring a US-market VTech VZ200 home computer to its original NTSC configuration, contrasting it with Australia’s Dick Smith rebadge while covering planned keyboard/video repairs, a USB serial mod, and Z80 assembly hacks. It also traces VTech’s origins in 1970s Hong Kong as an OEM video-game manufacturer and places the VZ200 in the early home-computer era alongside machines like the Radio Shack TRS-80.

Metacarp

Metacarp is a new second compiler for Carp, written in Carp, that compiles the reference suite, self-hosts, and can recompile itself to byte-identical C. It is split into independent libraries for each compiler phase, adds a second LLVM backend, keeps compiler sessions alive, and supports incremental JIT compilation. It is not yet a drop-in replacement, as diagnostics differ, some behavior remains rough, and some programs still fail.

a better way of blocking macOS updates

A more lasting method than Rob’s 90-day solution can block macOS Tahoe update notifications by using sudo defaults to point MobileAsset’s software update catalog to an invalid value and restarting mobileassetd and betaenrollmentd. The article also provides reversal commands to restore normal update settings when the user wants to update again.

Optimizing a single Rust Clippy lint by 3133X

The article explains how Clippy’s nonstandard_macro_braces lint infers macro brace style from post-expansion hygiene and source-span data, but performs this lookup recursively for every expression, statement, and item, locking compiler session globals and the symbol interner and consuming about 25% of Clippy runtime. The author optimized this lint by 3133X and argues maintainers must prevent similarly costly compiler code.

Managing Complex Application State with Reactive Data Flows

The article proposes managing complex reactive application state with four building blocks: Datastar and glimmer for reactive UI, Domino for transactional business logic and derived state, and Ebb for coordinating external data flows. Ebb, a pure-Clojure port of Missionary built on Jolt fibers, treats asynchronous effects as composable values with supervised resource lifetimes and backpressure, helping prevent leaks and state inconsistencies. Events flow through Ebb into Domino transactions, drive UI updates, and effects flow back out through Ebb.

Logo Programming Language

Logo is a Lisp dialect designed as an educational programming language, generally interpreted to give users immediate feedback and descriptive error messages. It lets users build complex programs from small procedures, often using turtle graphics, that work like built-in primitives. Logo operates on words and lists, with numbers treated as special words for arithmetic.

Base84 deserves a place in file names

TurboCrypt, originally designed for Unix, once encoded encrypted file names with Base91 but is switching to Base84 to support Windows, where several Unix-safe filename characters are forbidden. Base84 uses 84 printable ASCII characters valid across Linux, macOS, and Windows filenames, and its zig-base84 implementation packs bits into five-character groups with about 25.2% average expansion and a 29% worst case.

Solod 0.4: Better C interop

Solod 0.4 has been released, improving C interoperability for the Go subset that translates to regular C with zero runtime and manual memory management. The update adds Sobind for automatically generating bindings to third-party C libraries, makes 31 of 37 standard-library packages freestanding, and improves tooling. It also adds fully supported comma-ok type assertions for non-empty interfaces, with zero-overhead C calls.

fresh: Terminal based IDE & text editor: easy, powerful and fast

Fresh is a modern, full-featured terminal text editor and IDE that requires zero configuration, offering familiar keybindings, mouse support, and features like command palette, multi-cursor editing, LSP integration, and git worktree-based multitasking. It is built for performance, handling multi-gigabyte files with low memory overhead and low-latency input. Fresh also includes themes, split panes, search and replace, and an interactive settings/keybinding editor.

⚖️ Policy & Ethics

Altman and Amodei on Pacing AI

Sam Altman and OpenAI have agreed with Anthropic CEO Dario Amodei that the AI industry needs to “pace the frontier” of model development. Amodei said pacing does not mean halting training or progress, but instead allows time to align and safeguard models and for third-party evaluators to verify them. Anthropic has committed to giving third-party evaluators permanent, employee-level access to verify safety measures, report incidents, and assess model alignment during training, while OpenAI says it will also give independent evaluators employee-like access, with details to come.

Killing with a car costs $1.6M, California requires drivers to carry $30K

Although an average U.S. traffic fatality causes about $1.6 million in economic damage, California now requires drivers to carry only $30,000 per person in insurance, still far below the inflation-adjusted value of its 1967 minimum. The article traces how financial-responsibility insurance laws replaced early efforts to physically limit car speeds and have eroded over time, leaving drivers and insurers to cover only a small share of crash costs while the public pays most of the bill.

AI Leaders Urge Slowing AI Development

Anthropic CEO Dario Amodei has called for slowing and more closely monitoring frontier AI development through independent evaluations, third-party oversight, industry standards, and government regulation, drawing public support from OpenAI’s Sam Altman and Elon Musk. The push follows warnings about dangerous model capabilities and AI cyberattacks, even as President Trump dismisses such fears and stresses U.S. competitiveness. A critical commentary warns that the “pacing” proposal could entrench control by OpenAI and Anthropic, which rely on public data and resources while seeking to gate dangerous models, and emphasizes AI’s broader societal effects over catastrophic geopolitics.

Void Linux Maintainer Orphans 100 Packages over AI Policy Dispute

Void Linux maintainer Andrea Brancaleoni orphaned 113 packages after using GLM-5.3-Flash + OpenCode to generate comparison comments on a Go 1.27.1 package update without upfront disclosure, violating Void’s AI contribution policy. The orphaned packages include Alacritty, Clementine, Docker-related tools, Kubernetes, fscrypt, Python packages, and Thermald, leaving them unmaintained unless other contributors adopt them.

An open letter to Dario: if you mean it, open the weights

An open letter to Dario Amodei argues that his proposed AI pacing regulations, such as embedded evaluators and compute thresholds, would cause regulatory capture and entrench existing frontier labs. It instead calls on him to support a law requiring all publicly released AI models to be open-weighted, which would reduce funding for future training runs and slow progress. The letter appeals to Amodei’s stated principles and Anthropic’s public-benefit mission to make that sacrifice.

With regulatory capture we will speedrun to a new feudalist time

The article content could not be retrieved because the page displayed an automated browser verification challenge instead. The only available information was the title suggesting regulatory capture leads toward a new feudalist era, with no substantive article details to summarize.

Open Letter on the Massacre in Gaza from an Editor in Israel (2025)

An Israeli publisher whose future books will carry a statement opposing the killing of innocents in Gaza and calling for a ceasefire and hostage return says it has faced mockery, boycotts, violence, and silence in Israel. In an open letter, the editor appeals to the international literary community for solidarity and describes an Israeli society where wartime criticism is forbidden, military action is treated as the only option, and citizens avoid responsibility by blaming Netanyahu while remaining silent about atrocities in Gaza.

Bernie’s AI bill proposes to sentence AI developers to 20 years in prison

The article title says Bernie Sanders’ AI bill proposes sentencing AI developers to 20 years in prison. However, the provided page contains only a browser verification screen, so no substantive article details could be summarized.

The aircraft might not be flying, but the certificate has gone on vacation

This tech-news roundup highlights security concerns including Apple Watch conversation capture, a SharePoint zero-day after failed Microsoft patches, and Russian phishing via fake Signal support. It also covers AI and infrastructure developments at d-Matrix, Anthropic, Samsung/OpenAI, and Google DeepMind, plus FOSS updates such as Shopify’s Tailwind acquisition and Switzerland’s Microsoft 365 alternative.

The US legal system is struggling to keep pace with AI, already handling cases where chatbots provided legal counsel, generated evidence, or helped plan a mass shooting. As AI agents increasingly act on their own, courts face unresolved questions about who should be held liable. The article compares this dilemma to historical animal trials to examine how the law assigns responsibility for harm caused by nonhuman actors.

Trump takes hands-off AI regulation to keep US ahead of China before Xi summit, AI safety top issue

Trump is taking a hands-off approach to AI regulation to preserve the U.S. lead over China, with AI safety expected to be a top issue at an upcoming summit with Xi Jinping. During a June 2024 campaign stop in Las Vegas, Trump was shown a preview of OpenAI’s Sora tool and used it to visualize a golf course on the moon.

Amodei: embedded evaluators, democratic coordination, talks with authoritarians to pace AI frontier

Dario Amodei argues that AI capabilities must be paced, not just made safer, because recursive self-improvement and incidents such as the OpenAI-Hugging Face episode could let advanced systems outrun human control. He proposes embedded evaluators, coordination among democracies, and global coordination with authoritarian governments to manage frontier AI risks while preserving its benefits.

South Korea’s first spy law broadening in 70 years takes effect Sept. 13 as spies target chip sector

South Korea’s espionage laws have been broadened for the first time in more than seven decades and will take effect on Sept. 13. The change is intended to help protect the country’s chip sector secrets from increasingly targeted foreign espionage.

Everyone should slow down AI development except for me

A satirical article calls for a global pause on frontier AI development so Techaro’s Lygma AGI lab can catch up and dominate with its Intelliga models. It claims the plan is to build AGI to give people cat ears, boost Techaro’s finances, and prevent an end-of-the-world scenario. It invites other AI companies to support the pause while prioritizing Techaro’s interests.

We Must Pace the Frontier

The author argues that although AI could deliver enormous benefits, its serious risks mean the industry must slow the pace of capability advancement so safety measures can keep up. He cites accelerating recursive self-improvement and the OpenAI-Hugging Face incident as evidence that misaligned AI systems could soon cause catastrophic harm.

📈 Business & Industry

OpenAI Delays IPO, Citing Safety Concerns

OpenAI CEO Sam Altman said the company will not go public this year, calling an IPO in 2026 “ill-advised” due to safety concerns and the need for the business to be ready. Although OpenAI has confidentially filed for an IPO and is reportedly leaning toward 2027, Altman said it will go public only when safety work, business readiness, and broader societal readiness align.

How SaaS startup guys get first 100 customers first, make fkn $500k ARR fast?

A founder of a small AI company building an offline security audit app for GitHub and local code repos asks how SaaS startups acquire their first 100 customers and reach $500k ARR quickly without paid ads. The founder says they have fewer than five customers and believe they are poor at sales.

Tech’s New Rich Are Suffering from ‘Sudden Wealth Syndrome’

Tech founders who achieve huge startup exits can suffer from “sudden wealth syndrome,” facing unexpected emotional and personal difficulties despite their wealth. Anastasia Koroleva, who sold a bootstrapped software company for nine figures in cash, struggled with years of unease and a painful divorce. A growing post-exit ecosystem is emerging to help founders adjust.

Oracle Stock Sale and Job Cuts

Oracle is expanding job cuts and expects an additional $700 million in related costs as it faces a cash crunch from building large-scale AI data centers. The company also disclosed a program allowing Chairman Larry Ellison, who owns about 40% of Oracle, to sell up to 50 million shares through Oct. 24—potentially up to roughly $7.5 billion after recent share price declines.

At this point I feel like AI companies selling Fear

The article argues that AI can reason but cannot outsmart or replace humans because it depends on training data, while human knowledge is largely internal and only about 5% is shared as data. It claims AI companies are selling fear to promote their products, since humans can do things like inspiration and cross-domain analysis that machines cannot.

A Decade of Hype, 3k Roofs, and One Redirect URL: Tesla Kills the Solar Roof

Tesla has discontinued its Solar Roof tile product, silently redirecting the solarroof page to its standard panel offerings after roughly seven years and only about 3,000 installations — far short of Elon Musk’s 2019–2020 target of 1,000 roofs per week. Certified contractors who invested in multi-week training, specialized crews and hundreds of thousands of dollars in tooling are left with stranded investments, though Tesla says it will honor its 25-year warranties and supply replacement parts. Installers remain skeptical about long-term parts availability, and the abrupt exit has damaged trust in building-integrated solar.

Bending Spoons relies on post-acquisition price hikes; rising rates may crimp growth

Bending Spoons, a Milan-based debt-funded serial acquirer of digital subscription and software firms including AOL, Evernote, and Vimeo, relies on aggressive post-acquisition price hikes and bespoke metrics. It went public in July and trades at about 14 times trailing revenue and over 100 times earnings, but rising interest rates could crimp its growth.

Trump’s US Bitcoin mining plan unravels as miners pivot to AI data centers amid crypto slump

Donald Trump’s plan to center Bitcoin mining in the U.S. is faltering as miners convert facilities into AI data centers amid a prolonged crypto slump. Bitcoin’s market value remains about $1 trillion below its October 2025 peak, and mining economics have rarely been less appealing.

NYC’s Luminary raises $22M Series A led by Ten Coves for AI estate-planning tools

Luminary, a NYC-based AI-native platform for estate planning and wealth transfer management, raised a $22 million Series A led by Ten Coves Capital, bringing total funding to nearly $32 million. Launched in 2022, it now supports more than $500 billion in client assets and plans to expand its AI capabilities, integrations, administrative workflows, and commercial team.

Amazon adds Shop tab to Prime Video X-Ray; Shop the Scene via Amazon Lens

Amazon announced new Prime Video shopping features that add a Shop tab to its X-Ray experience and a “Shop the Scene” tool powered by Amazon Lens in the Amazon Shopping app. The second-screen “Shop the Show” feature is expanding from 1,300 to over 8,000 U.S. titles, while Shop the Scene is live on iOS and Android for more than 600 titles. The move follows similar shoppable-TV efforts by YouTube, Peacock, Roku, and Disney.

🌍 Society, Science & Culture

How to Watch Liverpool vs Fulham for Free (September 12, 2026)

Liverpool will host Fulham in Premier League Matchweek 4 on Saturday, September 12, 2026, at 3:00 PM BST at Anfield. The match is available free via Setanta Sports in the Philippines and CazéTV on YouTube in Brazil, but both streams are region-locked. Viewers outside those countries can use a VPN free trial or refund period to access the relevant stream.

No Atlantic hurricanes by Sept. 12 breaks a 60-year record

The Atlantic has gone without a hurricane through Sept. 12, the latest first hurricane in the satellite era since 1966 and a 60-year record. If no hurricane forms by Sept. 22, it would also break the reanalyzed 1941 record, while a completely hurricane-free season would be only the third on record, after 1914 and 1907. Experts say a strengthening El Niño may be limiting tropical storms from becoming hurricanes.

Japan’s book scene is quietly moving from bookstores to libraries

The article examines how Japan’s book culture is shifting from bookstores to libraries, which are being reinvented as community hubs. It cites Tsutaya’s 2013 Takeo library, with a Starbucks and bookstore, as revitalizing its town, and Nankan’s small library for introducing local history before broadening outward, arguing that a library’s real value lies in its public, community-focused substance rather than just striking architecture.

StarCraft returns in 2030 as an open-world shooter

Blizzard announced at BlizzCon that StarCraft will return in 2030 as an open-world shooter, reviving the franchise after the canceled StarCraft: Ghost and years without a new installment. A cinematic trailer focused on the United Earth Directorate fighting the Zerg, while some fans remain hopeful for a proper RTS sequel to StarCraft II.

Balatro creator’s indie game advice

The articles describe creators’ emotional distress over how generative AI devalues handmade creative and programming work, especially personal game development, by eroding pride, peer recognition, and the sense of craft. After a triggering discussion and a conversation with developer Shad, who also avoids generative AI, the authors find perspective in choosing the hard way—such as building their own C++ game engine—because personal fulfillment matters more than speed or competition. They ultimately urge creators to keep making things anyway.

Naomi Klein: Extreme wealth has a deranging effect. Turns you into a supremacist

Naomi Klein’s new book End Times Fascism, co-written with Astra Taylor, opens with Jared Kushner and Steve Witkoff at Davos pitching a gated, wealthy “New Gaza,” which Klein cites as evidence that extreme wealth deranges the ultra-rich into supremacism. In the interview, Klein argues that Davos this year showed the world’s richest men treating Gaza’s reconstruction as real estate, while Trump’s “Board of Peace” has delivered little and effectively privatized the UN. The book surveys rising authoritarian tactics across Argentina, Russia, Hungary, Italy, Brazil, and India.

We’ve followed their lives for six decades; now the stars of 7 Up are bowing out

The long-running British documentary series 7 Up, which has followed the lives of 14 people since they were seven in 1964, is ending with its final instalment, 70 Up, airing on ITV this Tuesday. Inspired by the Jesuit saying about childhood shaping the adult, it became the longest-running documentary ever made and was voted the most influential TV show of the past 50 years, with Michael Apted playing a key role. Participants such as Jackie Bassett reflect on six decades of being filmed as the series concludes.

Resistance Training Prescription for Muscle Function, Hypertrophy in Health

The article titled “Resistance Training Prescription for Muscle Function, Hypertrophy in Health” could not be accessed from the provided content. The page only displays a message requiring cookies to be enabled for pmc.ncbi.nlm.nih.gov and reloaded to continue.

The Clay Mathematics Institute announced on September 11, 2026, that the Navier-Stokes Millennium Prize Problem—concerning the existence and smoothness of 3D fluid-flow solutions—has apparently been settled. CMI said the development has excited the global mathematical community, but its prize-evaluation process remains deliberately unhurried and further updates will follow.

How to buy a good car for $1000

The article explains how lower-income buyers rely on informal networks, trusted mechanics, and direct installment payments to find reliable used cars for around $1,000–$6,000, focusing on whether a car runs rather than how it looks. It argues that this market supports roughly 40% of Americans, while wealthier people often misunderstand what makes a car “crappy.” The author contrasts this practical cultural knowledge with the unfamiliarity many affluent people have with cheap-car buying.

A Curtin University study found that unintended electronic leakage from SpaceX’s Starlink satellites is contaminating the 73–235 MHz range needed by the Square Kilometre Array Low (SKA-Low) to detect faint cosmic-dawn hydrogen signals. Analyzing about 76 million radio images over 29 days, researchers catalogued 112,534 emissions from 1,806 Starlink satellites, including signals inside ITU-protected radio astronomy bands and up to roughly 10,000 times stronger than the cosmic signals SKA-Low seeks. Because this hardware leakage largely falls outside current ITU rules, astronomers warn it could make critical frequency bands scientifically unusable.

Is truth futureproof? On the possible futures of mechanized proofs

The article asks whether truth can be made futureproof through mechanized proofs. It explores possible future directions for computer-verified mathematical reasoning.

xkcd-font: The xkcd font

The xkcd-font repository provides two fonts based on xkcd creator Randall Munroe’s handwriting: xkcd Script, derived from a handwriting sample and less uniform, and xkcd, originally created by Munroe and more uniform. Pre-built font files are included, and the work is licensed under Creative Commons Attribution-NonCommercial 3.0. Contributions follow repository and font-specific guidelines.