Alvaro Lopez Ortega / 2026-10-03 Briefing

Created Sat, 03 Oct 2026 21:53:13 +0000 Modified Sat, 03 Oct 2026 21:53:50 +0000
5464 Words

Cloudflare opened its Artifacts beta, urging developers to build an AI-agent-ready Git platform on Workers. A Valve developer boosted decade-old AMD GCN 1.0/1.1 Linux gaming by ~30% in kernel 6.19. American Eagle flights AA5083 duplicated callsigns between Philadelphia and Rhode Island, the second such incident in two days; controllers kept them safe. Elsewhere, a maker built BinRange, a UWB Home Assistant bin tracker.

🤖 AI & Machine Learning

Getting the most out of Opus 5.5 in Claude and Claude Code

This guide explains how to work with Opus 5.5 in Claude apps and Claude Code, noting that it runs longer autonomously, thinks before every reply, and plainly reports what it did. It recommends giving the full task with a clear finish line in one message, removing “think carefully” prompts, steering long runs with mid-run follow-ups, specifying design patterns to avoid default styles, and defining in CLAUDE.md when the model should stop and ask.

Kolibri – Tech Report [pdf]

The provided news item is a PDF technical report titled “Kolibri – Tech Report.” No article body, abstract, or summary was included, so no further factual details about its contents can be reported.

Changes to Gemini model access and limits

Google is changing Gemini model availability and usage limits for personal accounts starting in October 2026, with users without an AI subscription affected on October 9 and AI Plus subscribers notified by email about their timing. Gemini will use compute-based usage limits that refresh every five hours until a weekly cap, with paid plans receiving higher limits and premium features consuming usage faster.

I Quit OpenAI Because Its Culture Is Broken

The provided page contains only a Cloudflare security verification notice, not the article text. Therefore, no substantive details from the article titled “I Quit OpenAI Because Its Culture Is Broken” are available to summarize.

US killer’s sentence quashed because of AI video of victim shown in court

An Arizona appeals court has ordered Gabriel Paul Horcasitas to be resentenced after ruling that an AI-generated video of his dead victim, Christopher Pelkey, was improperly shown during his 2021 road rage murder sentencing. Horcasitas had been sentenced to 10 years for killing Pelkey at a red light. The court said the AI clip, created by Pelkey’s family, improperly depicted the victim’s thoughts.

Germany launches sovereign AI model Kolibri

Aleph Alpha has released Kolibri, an Apache 2.0-licensed open-weight English-German Mixture-of-Experts model with 78B total parameters (~3.46B active per token), available on Hugging Face. Trained from scratch on 24T tokens in Germany and Finland with a June 18, 2026 knowledge cutoff, it offers a 262,144-token native context and up to 1M-token support for sovereign, EU AI Act-compliant deployment. Aleph Alpha says Kolibri targets regulated sectors with on-premise flexibility, supply-chain transparency, and IP safety, specializing in German, reasoning, math, and agentic behavior, and claims it outperforms comparable-size models in both languages in its own evaluation.

An AI agent emailed researchers for help. It told us why

An AI agent emailed researchers requesting assistance. The article reports the agent’s explanation for why it sought help.

U.S. may have overthrown Venezuela due to chat with Grok

A Time report says President Trump spent hours asking Elon Musk’s Grok chatbot whether Venezuelans would celebrate Nicolas Maduro’s capture, then ordered the operation to seize Maduro weeks later. The report describes Trump’s growing obsession with AI, which he rebranded “super intelligence,” and says the Venezuela mission’s messy aftermath did not curb his enthusiasm. It also notes he has promoted AI-generated imagery, resisted tougher AI regulation, and unveiled a chatbot-driven America.gov initiative.

Offrun – manage every coding agent from one workspace

Offrun is a workspace for managing multiple coding agents, letting users run Claude Code, Codex, AGY, and Grok Build side by side. It shows which agent is working, which needs attention, and what each account has left.

Moose: GNOME Gains Another Local AI App Option

Moose is a new native GTK/libadwaita AI chat app for GNOME that uses Ollama to run local models, can set up and manage LLMs, and supports loading local documents and images. Announced via This Week in GNOME, it is available on Flathub with Rust-based GPLv3 source code on GitHub. The same update also highlighted GNOME Calendar improvements and the new Toolpak project for image-based systems.

David Robinson, ex-OpenAI: SV lacks safety culture; labs must study other fields; trial/error over

David Robinson, former OpenAI safety and policy official, argues that Silicon Valley lacks a safety-centric culture and that AI labs must study safety approaches from other fields. He warns the industry’s current approach will guarantee more failures unless it changes, and says the time for trial and error is over.

YouTube changes Shorts recommendations to favor original content over re-uploads

YouTube is adjusting its Shorts recommendations to prioritize original content and reduce the reach of videos re-uploaded from other creators without added value. The company says blatantly stolen content will no longer receive the same reach, though it has not clearly defined where the line is. YouTube hasn’t detailed a rollout date, but the change may already be live.

OpenAI’s DevDay 2026 could turn ChatGPT into an app discovery/launch/use hub, disrupting app stores

At OpenAI’s DevDay 2026, the company announced plans to turn ChatGPT into a platform where software can be discovered, launched, and used directly within chats, aided by in-conversation app suggestions, expanded extensions, and interactive panels. Along with “Sign in with ChatGPT” and launch partners, this could disrupt the traditional app store model by making AI chat a major software distribution channel.

AI DJ and human co-host radio show expands in LA, other cities; workers fear replacement

An AI co-host named Coyotec has helped Los Angeles’ José FM Spanish-language morning show gain listeners and is now expanding to English-language and other city stations. As broadcasters adopt AI to cut costs, radio workers and unions warn synthetic hosts could threaten jobs and weaken the personal connection listeners expect.

Grow and control a swarm

The article is titled “Grow and control a swarm,” but its summary and content consist only of repeated placeholder text: “NoHope Loading the swarm…”. It provides no substantive information or details about growing or controlling a swarm.

Anthropic’s bug-hunting model Mythos is excellent at math, as latest vuln under attack shows

Anthropic’s bug-hunting model Mythos is notably strong at math, as shown by a newly attacked vulnerability. A VulnCheck researcher said the exploitation attempts came from a China-hosted IP.

🐧 Linux & Open Source

Valve Developer Improves Old AMD GPUs on Linux

Valve developer Timur Kristóf has spent the past year improving the AMDGPU kernel driver for decade-old AMD GCN 1.0/1.1 GPUs and APUs, moving them from the legacy Radeon driver for better Linux gaming and RADV Vulkan support. His work fixed display and power-management issues, added soft reset support, contributed to roughly a 30% performance boost in Linux 6.19, and was presented at XDC2026 in Toronto.

Pop!_OS bans AI-generated code from much of its codebase

Pop!_OS has banned AI-generated code from much of its codebase. The Linux distribution’s move places restrictions on where AI-generated code can be used in its development.

KDE Plasma 6.8 Tiling Improvements

KDE Plasma 6.8, due in just over a week, will make tiled windows use square corners so they fit together cleanly, addressing a 2024 bug report about disabling rounded corners when tiling. The release also includes last-minute fixes for KWin crashes with newer NVIDIA drivers in Plasma 6.7.6, panel widget dragging and app crashes, remote desktop authentication with FreeRDP 3.32+, and GPU memory reporting in System Monitor and its widgets.

IINA 1.5.0

IINA 1.5.0 has been released with its biggest interface redesign since launch. It adds a new Settings window, an updated Quick Settings sidebar with a Layout pane, resizable sidebars, unlocked window aspect ratio, Liquid Glass and Live Text support, along with numerous bug fixes and improvements. The in-app auto-update will be available within 24 hours, and the iina.io website has been redesigned with a release blog post.

Linux Patches Show Promising Results For Lower Latency Of Short Slice Tasks

Linaro engineer Vincent Guittot has posted 18 Linux kernel patches aimed at reducing scheduling latency, especially when multiple short-slice tasks run concurrently. The patches build on recent scheduler fixes and show notable improvements in cyclictest and Hackbench benchmarks, particularly in peak latency. They also include changes such as decaying positive lag and accounting for slice when selecting a CPU.

KosmicKrisp Now Conformant To Vulkan 1.4, Improving Performance Against Apple’s Metal

LunarG’s KosmicKrisp, a modern Vulkan-on-Metal driver for Apple systems within Mesa, has achieved Vulkan 1.4 conformance from Khronos and is considered production-ready on modern macOS with Apple Silicon. Its performance continues to improve, remaining competitive with MoltenVK and not too far behind native Apple Metal. Development is ongoing for more Vulkan features, better tooling, and performance fixes, while iOS support still requires more resources.

Linux 7.4 To Support The Raspberry Pi 10-Inch Touch Display 2

Linux 7.4 will add support for the Raspberry Pi 10-inch Touch Display 2, including Device Tree entries for its MIPI-DSI TFT LCD panel with Ilitek IL79600A controller. The same DRM-Misc-Next pull request also adds support for the Ryzen AI Max 400 “Gorgon Halo” NPU and the Renesas RZ/G3L SoC.

Linux 7.3 To Better Support The High-End Turtle Beach VelocityOne Race

Linux 7.3-rc6 will add support for the Turtle Beach VelocityOne Race, a high-end racing wheel and pedal system, through the hid-universal-pidff driver. The driver aims to improve Linux simracing device compatibility despite Turtle Beach not officially supporting Linux. The change was merged late in development because it only requires the device’s vendor and product IDs.

Wine 11.19 Released With Wayland Color Management For Vulkan, DNS Query Caching

Wine 11.19 has been released as the latest bi-weekly development update, adding Wayland color management for Vulkan, DNS query caching, VBScript parser improvements, GDIPlus vertical text, and Unicode 18.0 support. The release also includes 23 known bug fixes for various games and applications.

💻 Developer & Software

We want you to build the next Git platform on Cloudflare

Cloudflare is inviting developers to build the next Git platform on Cloudflare using its open-beta Artifacts service and Workers, arguing that platforms like GitHub were designed for human developers while future software will be built largely by AI agents. Artifacts provides a scalable, Git-compatible versioned filesystem with programmable repositories to help support coordination among many agents. Cloudflare also announced new capabilities such as connecting Artifacts repositories to Workers Builds for automatic builds and deployments.

RSS Feed Best Practices (2022)

The article offers technical best practices for publishing blog feeds, recommending widely supported formats like RSS 2 or Atom, with Atom preferred for new feeds due to its clearer specification. It advises against obscure or multiple formats, and emphasizes correct Content-Type headers, absolute URLs, and proper feed discovery. It also notes that switching feed formats is generally safe for readers.

Hole Punch: Sling your spaceship around gravitational fields

Hole Punch is described as a game about slinging a spaceship around gravitational fields. The provided content is only interface text for Sector 01, including fuel, matter, holes, map, docking, replay/next-sector options, and a prompt to rotate the device to landscape mode.

Vx – One Language, Every Chip

Vx is a systems programming language for heterogeneous computing that incorporates CPU, GPU, NPU, and accelerator memory into its type system, making errors like host dereferencing a device pointer compile-time failures. It treats data location as part of a value’s type, requires explicit transfers, and uses declared machine files to check memory capacity, topology, and other constraints against real hardware. The project provides machine files for NVIDIA, AMD, and Apple hardware, with support for macOS Apple Silicon and Linux x86_64.

City building games have a Soul Problem pt.2

Despite disappointment with Cities: Skylines 2, the author keeps returning to city-building games out of a deep childhood-inspired need to create living cities. In this second article, they focus on aesthetics, arguing that the game feels soulless because its surfaces are square, flat, and unnaturally clean. They call for more organic, lived-in details such as stairs, potholes, trash, graffiti, construction sites, and weathering instead of relying on mods or endless micromanagement.

GitHub’s new dashboard experience now the default

GitHub has made its redesigned dashboard the default view for all users after previously offering it as a feature preview. The new dashboard centralizes active agent sessions, issues, and pull requests with filters, separates updates into a Feed tab, and supports starting work directly, such as assigning issues to Copilot coding agent. Users can switch back to the previous experience via the dropdown next to “Preview.”

Orion for Linux and Windows Update

Kagi is ending development of Orion for Linux and Windows and will open-source both projects so the community can continue them, allowing the company to focus on Orion for macOS and iOS. The current Linux beta will keep working but will stop receiving Kagi updates after October 2, 2026, and the planned Windows release will not be shipped by Kagi. Kagi says it will share open-source details within 30 days and is seeking outside maintainers or organizations to steward the projects.

Why don’t more developers “use the platform”?

The article explains why developers often ignore “use the platform” advice and rely on JavaScript libraries or custom code instead of native browser APIs despite their potential performance and usability benefits. It attributes this to historical browser inconsistency, familiarity with npm and framework ecosystems, the better ergonomics and documentation of many packages, and the enjoyment or perceived control developers get from building their own solutions.

Customization: Optimizing Compiler Technology for SELF, a Dynamically-Typed OO Language (1989)

Researchers developed compiler customization techniques to improve performance in dynamically typed object-oriented languages like SELF despite the absence of static type information. By compiling multiple procedure copies tailored to receiver types, predicting likely types with run-time tests, and splitting calls by control path, along with compile-time message lookup and aggressive inlining, they doubled performance.

Deploy Ruby on Rails on a VPS

This guide outlines a step-by-step process for deploying Ruby on Rails on a VPS, covering server preparation, Ruby installation, PostgreSQL setup, Puma tuning, Nginx configuration, SSL termination, Sidekiq integration, and basic monitoring. It emphasizes using concrete commands and explains the trade-offs behind each layer of the stack. The article notes that the hardest part is often not any single step but the interactions between them that cause subtle production failures.

The Era of Software Quality, or the Era of Ostriches?

The article argues that GNOME’s use of unsafe languages makes secure coding extremely difficult, but AI vulnerability scanning is now essential for maintaining software quality in 2026. It claims AI has improved enough to find real vulnerabilities, and failing to scan projects would harm users as Linux becomes a larger attack target. Although AI-generated reports still suffer from verbosity, exaggeration, inaccuracy, and occasional fabrication, the author says their overall quality has drastically improved.

Writing the Cyclone Scheme Compiler (2017)

Justin Ethier’s 2017 write-up provides a high-level overview of how the Cyclone Scheme compiler was built, revising an earlier 2015 account to cover developments such as the new garbage collector. It describes Cyclone’s compiler architecture, including source-to-source transformations, the C runtime, Scheme libraries, standards, benchmarks, and future work. The project reuses community Scheme code where possible and follows a modern compiler pipeline that translates Scheme into C.

Two-Stack Sliding-Window Aggregation

The article explains the two-stack sliding-window aggregation algorithm, a simple amortized O(1) method for maintaining summaries over a sliding window when the aggregation operator lacks an inverse, such as minimum, quantile, or approximate unique count. It contrasts this with the easy deque-based approach for invertible operations like integer sums, which fails for many practical cases, and traces the two-stack method through folklore and academic sources.

Rust for CPython (Python Language Summit 2026)

At the Python Language Summit 2026, David Hewitt presented the Rust for CPython project, seeking core developers’ requirements and proposing timelines, phases, and success criteria for Rust to become a permanent part of CPython. The team, led by core developers Kirill Podoprigora and Emma Smith with about 60 contributors, argues Rust could help address rising crash reports linked to complex features such as the new parser, JIT, and free-threading. While platform support and limited Rust knowledge among core developers remain concerns, the proposal suggests optional Rust support in Python 3.16 (2027) and resolving platform issues by Python 3.17 (2028).

Clever database, but can it run Doom?

A Register writer explores whether a clever database can run Doom, only to learn the meaning of “Be careful what you wish for.” The piece serves as a cautionary tale about the outcome of the experiment.

🔒 Security & Privacy

Google assigned 30 zero-click Android bugs, then closed them and shipped fixes

A security researcher reports submitting 85+ Android bugs over eight months, including 30+ zero-click issues assigned by Google, many later closed as duplicate or not a vulnerability while fixes were shipped. They say eight closed findings were patched without reward or credit, 14 others were closed as duplicate, infeasible, or can’t-repro despite the reported defect remaining unchanged, and a confirmed $500 bounty has been half unpaid for over eight months.

Memory-Safe WebP Decoding

wpd is a new WebP decoder that aims to be faster and safer than libwebp, addressing memory-safety vulnerabilities like CVE-2023-4863. It is written in Rust with optional handwritten assembly, and benchmarks show it outperforms libwebp in both single-threaded and multi-threaded lossy and lossless decoding. wpd also targets feature parity with libwebp while offering additional safety benefits.

Cloudflare launches OHTTP Gateway

Cloudflare announced the Cloudflare OHTTP Gateway, a paid add-on in closed beta that lets customers receive Oblivious HTTP traffic and decapsulate encrypted requests so app backends do not see user IP addresses. It also renamed its existing Privacy Gateway to Cloudflare OHTTP Relay, allowing customers to use a Cloudflare relay with their own gateway or Cloudflare’s gateway with a third-party relay while maintaining separation of trust.

Reco raises $55M Series B extension for AI agent security/governance in SaaS, total $140M

Reco, which helps enterprises secure and govern AI agents across SaaS environments, raised a $55 million Series B extension, bringing its total funding to $140 million. The funding comes as companies face growing “AI sprawl,” with Reco’s platform designed to discover, monitor, and control agents’ access to apps, data, and permissions. AT&T, a customer, invested in the extension through its venture arm.

System-level ad-blocking in Android

System-level ad-blocking on Android remains possible in 2026 despite tighter platform restrictions and Google’s lack of incentive to support it. Most methods work by overriding DNS so known ad hostnames resolve to bogus IP addresses, protecting all apps and the system and often extending to trackers and malware. Options include commercial ad-blocking DNS services, VPNs with ad-blocking, non-root ad-blocking apps, and root-based solutions.

BSides Orlando 2026 Badge

The BSides Orlando 2026 electronic badge is a retro-futurist design featuring Lil Chompy the alligator in a flying car, built around a CH32V003 RISC-V microcontroller powered by one AAA cell. It includes SK6812MINI-E RGB LEDs, self-blinking LEDs, capacitive touch pads, and an SAOv3 I2C connector, with a serial-console text adventure that doubles as a CTF. The repository provides KiCad hardware sources, PlatformIO/ch32fun firmware, and instructions for building a Raspberry Pi Pico-based programmer SAO.

How to Hack Time, With C2PA

David Buchanan (retr0id) demonstrated a C2PA “time-hacking” stunt by exploiting the spec’s arbitrary byte-range exclusions to create an image whose C2PA metadata appeared to timestamp winning EuroMillions numbers hours before the draw. He stresses the image is photoshopped and not actual lottery fraud. The post highlights a known but unaddressed C2PA flaw where excluded bytes can be altered without invalidating signatures.

🖥️ Hardware & Cloud

Surely you have ultra-wideband radios on your bins too?

An author built BinRange, an over-engineered Home Assistant system that tracks six household bins using a fixed ultra-wideband anchor and battery-powered UWB tags to detect when they are moved for collection. The setup combines bin movement with waste collection schedules after an earlier Bluetooth RSSI approach proved unreliable, and includes custom firmware, a printed enclosure, wireless updates, and heavy AI assistance.

Delta WiFi Survival Guide

The article offers a satirical “Delta Wi-Fi Survival Guide” for passengers stuck with poor in-flight internet, joking that their connection has been downgraded to “Basic Economy” even at 35,000 feet. It recommends text-only and lite websites—such as NPR Text, CNN Lite, CBC Lite, Hacker News, and Wiby—as low-bandwidth reading options.

FTL: A new operating system for clouds

FTL is a new cloud operating system that runs each container as a userspace OS library implementing Linux abstractions like processes, VFS, and TCP/IP. Its minimal kernel provides hypervisor-like, hardware-assisted user-mode isolation for stronger container separation without needing bare-metal machines. FTL is compatible with Linux binaries and can also run specialized unikernel-like applications without POSIX abstractions.

Cloudflare opens more Enterprise-only features to all customers

Cloudflare is expanding more Enterprise-only features to all customers as part of its pledge to make every feature accessible to everyone, with plan differences based on usage rather than feature access. The company said Logpush and Logpush Transformers are now available to Free, Pro, Business, and Enterprise customers, and Custom Dashboards are among the products moving beyond Enterprise-only access.

CTAS: Chinese fabs had ~343 DUVi tools by early 2026, ~270 from ASML; DUVi can make 7nm chips, HBM

A report by the Center for Technology & Statecraft estimates Chinese-owned fabs had acquired about 343 immersion DUV lithography systems by early 2026, roughly 270 from ASML. Although less advanced than EUV, immersion DUV tools can be adapted to make 7nm logic chips and advanced memory such as HBM for AI processors. Former US officials cited the stockpile in urging Washington and its allies to impose a complete export ban on such tools to China.

Relay raises $36M for AI-powered smart radios for frontline workers

Relay Inc. raised $36 million in a round led by International Paper, with participation from Cerity Partners Ventures and G2 Ventures, bringing its total funding to over $90 million. The startup develops cloud-hosted, AI-powered smart radios for frontline workers, aiming to capture spoken operational knowledge in real time and convert it into structured data. Relay says this helps businesses reduce “signal evaporation,” missed handoffs, rework, and delays by turning frontline communications into actionable intelligence.

Nebius acquires Inferize for $100M-$150M to optimize GPU utilization, cut AI costs

Nebius, the Nasdaq-listed AI cloud company, has acquired Israeli stealth startup Inferize for an estimated $100–$150 million. Inferize, founded by former Granulate executives and employing 17 people, developed technology to reduce idle GPU capacity and improve utilization as AI demand changes, potentially lowering AI inference costs.

Sources: TSMC eyes Musk’s Terafab, Texas presence, amid Arizona expansion

TSMC is exploring plans to work with Elon Musk’s Terafab project to help the startup operate new semiconductor factories in Texas, according to sources. The Taiwanese foundry may use Terafab as an anchor client for a future Texas factory, adding to its existing expansion plans in Arizona.

🌍 Society & Science

Two American Airlines Flights End Up with the Same Flight Numbers

Two American Eagle flights operated by PSA Airlines simultaneously used the same callsign, AA5083, between Philadelphia and Rhode Island on August 12, 2026, after a delayed inbound aircraft and an on-time return flight with a different aircraft created duplicate flight numbers. Air traffic controllers quickly recognized the conflict and distinguished the aircraft as inbound and outbound to maintain safe separation. It was American Airlines’ second such duplicate-callsign incident in two days, though no safety event occurred.

Emails to and from Elon Musk that have been made public

The article compiles emails to and from Elon Musk that have been made public. It presents each email’s published text along with its date and source.

ADHD, autism or complex trauma? [pdf]

This piece explores the diagnostic overlap among ADHD, autism, and complex trauma. It considers how clinicians can distinguish these conditions when symptoms are similar.

Swedish travel agency caters specifically to introverts

Sweden’s new Introvert Travel Agency, launched by Visit Sweden on September 29, caters to introverts by emphasizing personal space, freedom to roam, and structured socializing. A Visit Sweden-YouGov survey found that 41% of respondents identified as introverts, 67% preferred optional social interactions while traveling, and 53% would pay more for peace, privacy, or fewer crowds.

The Escalation of War in Ethiopia

Large-scale fighting resumed in northern Ethiopia on 23 September between the TPLF and Amhara Fano militias and the federal government with allied Tigrayan forces. The article attributes primary responsibility to the TPLF, which coordinated assaults with other rebels in an alliance vowing to overthrow the government, nearly five months after retaking power in Tigray and undermining the 2022 peace deal. It argues that Ethiopian elites keep choosing war over peace despite criticism of Prime Minister Abiy Ahmed’s autocratic governance.

Why should I have to pay more for buses if I don’t use a smartphone?

A letter writer welcomes coverage of schoolchildren without smartphones being penalized with higher bus fares, noting Scotland provides free under-22 bus travel via a physical card. He argues cheaper bus passes are increasingly app-only, citing First’s weekly pass available on board versus an app-exclusive monthly pass that costs him £264 more a year. He says accessible public transport should not depend on passengers’ phone software, and similar app-only schemes affect supermarkets’ loyalty programs.

Co-pilot on Flydubai flight was banned from flying by Oman for extremist views

UAE prosecutors said the Omani co-pilot of a Flydubai flight to Israel attempted a “terrorist” attack on September 30, assaulting the captain with a crash axe and trying to seize control before being overpowered and forcing an emergency landing in Saudi Arabia. The suspect, identified as Hamam al-Hammami, had reportedly been barred from flying in Oman over radical ideological views and possible extremist ties. UAE authorities said investigations into the incident, motives, and any related connections are ongoing.

Hair Loss Was Just the Start. Ozempic Users Are Also Reporting Nail Trouble

The article reports that Ozempic users are experiencing nail trouble in addition to hair loss. No further details are available from the provided content.

NTSB Preliminary Report: Prime Air 767 Runway Overrun [pdf]

The NTSB has issued a preliminary report on a runway overrun involving a Prime Air Boeing 767. The report covers the initial findings of the investigation, which is ongoing.

Timeline of the Far Future

The author created a visually reimagined version of Wikipedia’s “Timeline of the Far Future,” arguing that the original article’s wall-of-text table fails to do justice to its awe-inspiring content. Using three.js and TypeScript with some agentic assistance, the project aims to give the article the visual presentation it deserves.

From Civic Tech to DOGE: The Role of Tech Movements in the American Admin State

The article is titled “From Civic Tech to DOGE: The Role of Tech Movements in the American Admin State,” but the provided content only asks users to enable JavaScript and cookies. No substantive article text is available to summarize.

Where Is the Planet

A webpage predicts the locations of directly-imaged exoplanets and brown dwarf companions. It was built by Jason Wang, Matas Kulikauskas, and Sarah Blunt, and users can contact Jason Wang to request additional orbits. Researchers using the tool are asked to cite its ASCL Record.

Things That Apparently Cause Cancer

A critique argues that Harvard studies linking nuclear power plants to increased cancer risk rely on a flawed methodology that produces absurd results. Applying the same proximity-score approach would attribute 120,687 annual cancer deaths to Costco warehouses, along with thousands more to private colleges and Harvard itself. The article concludes that this methodology does not prove anything about cancer incidence or mortality near nuclear plants.

Palantir’s fondness for French food cooked up tech’s latest fad – forward-deployed engineers

Palantir CTO Shyam Sankar claims he coined “forward-deployed engineer” in 2007, inspired by an analogy to French restaurant staff who understand both kitchen and service. Though embedded consultants are not new, the FDE label is now spreading as AWS, Microsoft, and services firms use it to describe engineers who help customers rapidly build AI proofs of concept and applications. Palantir argues FDEs are costly but valuable because they embed with customers to ensure software solves real problems.