Alvaro Lopez Ortega / 2026-11-02 Briefing

Created Mon, 02 Nov 2026 15:50:16 +0000 Modified Mon, 02 Nov 2026 15:50:58 +0000
7652 Words

Google dismantled the XCPE botnet, severing a backdoor that had ensnared over 1.3 million set-top boxes and banning maker Videostrong. Proton CEO Andy Yen scrapped company-wide AI “re-interviews” after backlash. Ubuntu faces criticism for leaving a critical kernel TCP flaw unpatched for over two weeks, while San Francisco’s AI boom contrasts sharply with laid-off tech workers’ struggles.

🤖 AI & Agents

The U.S. Census Bureau and “Super Intelligence”

The article reports that the U.S. Census Bureau’s Business Trends and Outlook Survey changed its AI adoption question in November 2025 from use in “producing goods or services” to “any business function,” making pre- and post-change answers non-comparable. Updated data show large-company AI adoption rising to about 40% by October 2026, and the survey began asking about “Super Intelligence” from October 19, 2026 under Executive Order 14434.

Backporting security fixes at scale with Codex

The article describes using Codex to accelerate security-fix backporting for stable Linux releases, reducing human preparation and review from days to hours per fix. By October 29, the project helped deliver fixes for 149 CVEs to stable Linux kernel releases and several Fedora and Ubuntu packages, with AI agents handling patch identification, adaptation, testing, and submission packets while engineers review and submit. Regression testing and maintainer acceptance remain essential, but the reduced effort makes neglected backport work practical and aims to avoid forcing users to choose between stability and security.

AI Can Write More of the Code. We Still Have to Engineer the System

AI agents can take on more implementation work and assist with analysis, design, and testing, but engineers remain responsible for deciding what to build and what evidence is sufficient to accept it. The article argues for keeping durable architectural decisions in reviewable artifacts, supplying relevant context for each increment, and verifying changes rather than relying on long chat sessions. It cautions against equating task automation with workforce replacement.

Vintage fun: Prodding AI agents using retro hardware just because you can

A hobbyist is using retro hardware to prod and interact with AI agents, describing the exercise as a way to stretch what is possible in the current age of AI coding. The effort is framed as keeping curiosity alive rather than serving a practical need.

From Expertise to Intelligence: NISCO Deploys World’s First Swarm-Intelligence Steel Model at Scale

NISCO, with Huawei and ecosystem partners, has commercially deployed YuanYe—described as the world’s first swarm-intelligence-powered steel large model—across multiple scenarios at scale. The deployment supports 300+ AI applications, has connected and digitally controlled 100% of key equipment, shortened new-product R&D cycles by 50%, and reduced annual costs by over CNY100 million.

We gave cloud agents a phone to QA

Incident.io has equipped its cloud coding agents with cloud iOS simulators and Android emulators so they can run mobile app changes, navigate the app, and attach screenshots and recordings as evidence to pull requests. The agents do not judge correctness—a human reviewer decides—and in the first two weeks the agent QA’d 42 PRs, with 26 merged.

The Thin Orchestrator: How We Made Our AI Agents More Reliable by Giving the Router Less to Do

Meta reports that using a “thin orchestrator” for its AI data-pipeline agents—a 1,300-line router that sends 20+ workflows to 14 specialized agents and performs no domain work itself—improved efficiency and accuracy. The approach addressed production hallucination caused by a single agent overloaded with context, which led to cross-contaminated domain concepts. Meta is sharing the process so other teams can apply it to production AI agents.

The bridge to agent-scale Git infrastructure

GitHub has migrated copies of 99.99999% of its repositories to Azure, leaving only the busiest 0.00001% held back by a 25ms round trip. The remaining repositories will move using per-reference quorums for writes, while object storage allows reads to scale independently.

Apple developing ~$1,000 tabletop robot with 9-inch screen, swiveling arm for 2027-28

Apple is developing a tabletop robot with a 9-inch screen and a swiveling arm, according to sources. The device is targeted for release in 2027 or 2028 at around $1,000, as Apple CEO John Ternus pushes the company to get serious about robotics.

OpenAI to merge ChatGPT Chat/Work desktop tabs for select users first; Codex separate

OpenAI plans to merge ChatGPT’s Chat and Work tabs in its desktop app, starting with select users for feedback before a wider rollout. Codex will remain a separate experience, and both ChatGPT and Codex will draw from the same token allowance pool.

AI safety debates: SURA’s rise and why safety labs shouldn’t abandon the AI race

A LessWrong discussion examines AI safety strategy debates, including the “Stop Unilaterally + Raise Awareness” (SURA) approach and whether safety-focused labs should abandon the AI race. The author challenges a binary choice between SURA and “race and align,” argues for pushing AI companies to be better while remaining competitive, and questions probability estimates and assumptions about whether releasing models substantially increases danger.

Unlearning, weight edits strip censorship, harden Chinese AI for Western firms

Startups including Hirundo and Lazarus are using machine unlearning and weight editing to adapt Chinese open-weight AI models for Western companies concerned about censorship and security. They modify model weights to reduce state-aligned bias and politically sensitive restrictions, harden models against cyberattacks, and bring safety standards closer to mainstream Western AI systems.

Terence Tao: Once AI-enthused, now warns AI threatens math’s deliberative process

The article profiles UCLA mathematician Terence Tao, who was once enthusiastic about AI’s role in mathematics but now warns that AI could destroy the field’s deliberative process. It notes that AI is solving problems that stumped mathematicians for decades, and that the field looks to Tao for guidance amid this altered landscape.

YC Labs: multi-agent Minecraft tests show emergent roles, hierarchies, coordination

YC Labs detailed multi-agent tests in games such as Minecraft, where agents exhibited emergent capabilities including role formation, hierarchies, and complex coordination patterns. YC Labs builds open-source AI and research tools for founders and YC, including its multiplayer agent harness QM for work.

🔒 Security & Cyber Threats

Shutting the Door on the XCPE Botnet (Threat Intelligence)

Google and industry partners dismantled the XCPE backdoor, which had enrolled over 1.3 million set-top boxes into a botnet. Google assesses with high confidence that Android OEM/ODM Videostrong intentionally developed and operated the backdoor for financial gain, selling access to criminal affiliates whose modules conducted ad-fraud, enrolled devices in residential proxy networks, and loaded additional malware. Google cut ties with Videostrong and banned it from making devices with preloaded Google apps, while disrupting command-and-control domains tied to affiliates in an effort assessed to sever access to at least 10 million additional compromised devices.

Tor Browser 16.0 released

Tor Browser 16.0 has been released as the first stable version based on Firefox ESR 153. It adds 64-bit Arm Linux support, redesigned settings, and generic window titles to help prevent other applications from monitoring browser activity, while Android receives Material Design 3 interface updates and 32-bit x86 Linux support is discontinued.

Supermarket freezer cyberattacks spread across Europe

Suspected cyber-attacks disrupted refrigerator and freezer systems at dozens of supermarkets across Belgium, Luxembourg, Sweden, and Ireland, affecting Delhaize/Ahold Delhaize, Coop, ICA, and SuperValu stores and causing temperature spikes, precautionary product removals, and reported food losses. The incidents echo a recent large-scale attack on French supermarket cooling systems, and authorities and companies are investigating whether they are linked.

Swedish grocery chain refrigerators and freezers hacked overnight

Swedish grocery chain ICA says it was hacked overnight, with refrigerators and freezers in its stores set to 50 degrees Celsius and food spoiled as a result. ICA will keep stores closed in the morning but plans to reopen at 9:00 a.m. Swedish time.

Emergency Alert Apps Hacked in Russia, Ukraine

Ukraine and Russia both reported cyberattacks on their emergency alert apps. Ukraine’s air raid alert app was compromised late Sunday, sending Kyiv residents false warnings of nuclear and chemical danger and street fighting, while Russia confirmed a similar hack. Both countries reported attacks on their emergency and air raid alert systems.

Frozen part II: Hackers remotely defrost produce in Euro supermarkets … again

Hackers remotely defrosted produce at supermarkets in Europe, causing mass food wastage across dozens of stores. Managers reportedly started the week finding puddles, in a repeat of an earlier incident.

Investigation: 400+ Chinese-linked apps/domains bypass app reviews for nonconsensual deepfakes, CSAM

A WIRED investigation found that the pet-portrait app Viefur is part of a network of more than 400 domains and apps posing as everyday services but secretly generating explicit AI sexual deepfakes, including scenes of rape and knife threats. The network was promoted through Meta ads—some containing AI-generated child sexual abuse material—and linked to apps in Apple and Google stores that enabled payments through their systems. The operation appears to be run by Chinese developers and a Hong Kong-registered entity called KeLing Technology Limited, using disposable apps designed to evade app-store review.

AO3: Service restoration after security incident may take 2–4 weeks

AO3 went offline after its nonprofit operator detected a security incident and warned that restoring service could take two to four weeks. The site said no fanworks or user content had been deleted, but it could not rule out theft of user information such as usernames, email addresses, IP addresses, and hashed passwords. AO3, which hosts more than 18 million works and has over 11 million registered users, is still investigating.

Binance’s hack response: ~100-person global war room, crypto firms and FBI coordination (Bloomberg)

Binance assembled a global war room of about 100 people and coordinated with other crypto firms and the FBI to contain the largest crypto hack in history. The attack began with hackers rehearsing code on Ethereum accounts they controlled, testing a program to seize Binance-owned accounts and move funds. Binance worked through sleepless nights to limit the breach.

Recognizing Keystrokes Using WiFi Devices

The provided article contains no substantive news content, only a message stating that JavaScript is disabled and that verification is required to continue. No details about recognizing keystrokes using WiFi devices are available.

🛠️ Software & Dev Tools

Hacker News but for comments posted on HN

The provided text is not a single article but a collection of unrelated comments on topics such as AI adoption and valuation, school uniform and multiplication-table policies, political spending, Debian versus Ubuntu, New Horizons’ interstellar-medium measurement, and JEPA-style AI architectures. It also includes remarks on an electrical-safety incident and cats purring.

CheerpJ 5.0: Java 21/25 in the browser, full OpenJDK, and a new WasmGC engine

CheerpJ 5.0 has been released as a full OpenJDK distribution for the browser in WebAssembly, adding support for Java 21 and 25 alongside Java 8, 11, and 17. It now ships the full JDK for every supported version, enabling Java compilation as well as execution in the browser. The release also introduces experimental support for a new WasmGC-based execution engine, with stabilization planned for CheerpJ 6.0 in early 2027.

HackerHack – HackerBooks but for Posts Posted on HN

HackerHack is presented as a HackerBooks-like tool for posts posted on Hacker News. The post notes that no AI was used.

Wikihack – HackerTube, but for Wikipedia articles mentioned on HN

Wikihack is a new frontend-only site that scrapes Wikipedia article mentions from Hacker News and ranks them into a scrollable feed. It shows where articles were submitted or mentioned in comments, with “Top” ranking frequent links, “Random” weighted toward popular articles, and “Rising” focused on the past 12 months. The creator built it after seeing Karpathy’s Hackertube and Hackerbooks and may add more features if users enjoy it.

13M opinions on 48k products extracted from Reddit and HN

A developer extracted 13 million opinions on 48,000 products from Reddit and Hacker News using Gemma 4 26B MoE on paired NVLinked 3090 GPUs with vLLM, plus a Claude-assisted frontend. The project includes a map view of travel opinions, and the creator says they are open to handing it over via email.

Pepperwind, a fictional company dataset with 223 tables and 79M rows

A software engineer has released Pepperwind Trading Co., a fictional Kochi-based spice and tea trader dataset with 223 tables and 79 million rows (~7GB of Parquet files), built using the backend behind Misata Studio. The dataset is designed for realistic, interconnected testing with balanced ledgers, consistent cross-system incidents, 92 monthly KPIs, and 420 questions. A free browser preview is available, while the full dataset costs $30 and is listed on Hugging Face under CC BY 4.0.

Microsoft’s Canada timezone troubles leave admins fixing calendars, pretending they’re in Bogota

Microsoft’s timezone handling is causing calendar problems in Canadian regions that no longer change clocks, leaving administrators to fix schedules manually. Some admins are reportedly using Bogota as a workaround while Microsoft addresses the issue.

Chap who coded Google’s Pac-Man clone reveals how the game freaked out Firefox

The developer who coded Google’s Pac-Man clone has explained how the game caused Firefox to freak out. The account details the browser problems triggered by the game and why it produced unexpected behavior.

A Decade of Blame Awareness

The article marks the tenth anniversary of “blame awareness,” a term coined in 2016 to argue that incident retrospectives should acknowledge humans’ tendency to blame rather than pretend it doesn’t exist. The author says empathetically addressing blame can make retrospectives more actionable and improve team trust and learning, and notes the concept has spread through LFI and resilience engineering circles. It also reflects on a cynical interpretation—“we’re now aware of exactly who we’re blaming”—as still useful for making blame explicit and open to cultural feedback.

How to Survive Database Failover – Debezium and PostgreSQL in Production

The article describes a Debezium/PostgreSQL failover failure mode where creating a new logical replication slot after standby promotion can skip events beyond Debezium’s last durable offset. Its main lesson is to treat PostgreSQL’s replication slot and Kafka Connect’s offset as a single distributed checkpoint, ensuring PostgreSQL retains every WAL record Debezium may need after failover. The pipeline provides at-least-once delivery, so consumers must be idempotent.

What a room of engineers asked me about realtime resilience

At the AWS Cloud Resilience Roadshow in Dublin, Ably discussed its approach to realtime resilience, arguing that dependability means not just uptime but every message arriving in order through disruption. The talk emphasized designing every layer to fail gracefully, rehearsing failure modes through disaster recovery drills on real automation, and treating resilience as a relational effort with AWS. Audience questions covered CDNs, regional failover, outages, message ordering, and capacity.

gohawk | Resource-focused static analysis for Go

gohawk is an industrial-grade static analyzer for Go that detects resource management, concurrency, and error-handling issues, and has been used to find and fix bugs in Docker, Kubernetes, and Caddy. It complements existing tools such as go vet, Staticcheck, and go-critic and can run standalone or as a go vet tool with no configuration file. Its diagnostics pinpoint offending code, show supporting evidence, and explain how to fix each issue.

Union Types and Option Types

The article explains union types as values that can hold one of several alternative types and describes the author’s experiments adding them to Copper, including restrictions such as no nested unions, significant type order, and no compatibility between different unions. It reports that full union types proved too complex for the minimalist low-level language, especially the optimized Pointer|Nothing case, so the author is now trimming the feature down to option types (T|Nothing), which represent a T value or nil.

Building a gemini server in python

The article explains the minimalist Gemini protocol and its community-focused “geminispace,” then describes using the Lagrange client to explore it. It provides a step-by-step Python tutorial for building a Gemini server, starting with a TCP socket listening on port 1965 and testing it with clients like nc or telnet.

Program

The article traces media’s shift from passive radio and TV to programmable digital algorithms that now monitor users and shape prices, perception, and attention through recommendation and engagement metrics. It argues that access to media has expanded while media literacy has declined, and it highlights the 1998 Technorealism manifesto and Douglas Rushkoff’s warning to “program or be programmed.” The Technorealists anticipated surveillance capitalism and called for democratic scrutiny, public sovereignty, and privacy standards beyond the marketplace alone.

My reflex of assuming our servers need ever-growing amounts of RAM

The article is a block page from Chris Siebenmann explaining that Wandering Thoughts/CSpace is experimentally blocking high-volume crawlers that use old Chrome user-agent strings, often to gather LLM training data, in order to reduce server load. Legitimate visitors may encounter it if their browser appears suspicious or too old, and the page notes related problems with Inoreader, Feedly, Vivaldi’s user-agent masking, and archive.* crawlers while suggesting contact with the author or use of archive.org.

the lost malloc()

The article recounts Unix archaeology into the original Fourth Edition (1973) sources, examining a simple linear allocator and Dennis Ritchie’s early malloc/mfree implementation. It highlights the historical significance and approachable style of this early memory-management code, and notes that a V4 tape was only discovered last year after being thought lost.

Tyrtulia, a headless forum with no passwords, made for the terminal

Tyrtulia is a headless forum for small communities on shared Unix servers that uses passwordless authentication via Unix sockets or signed SSH challenges and requires a terminal or Emacs client. It is lightweight—about 60–70 MB RAM and 50 MB disk—and adds Markdown threads, mentions, notifications, edit history, search, moderation, and Atom/SSE feeds to the BBJ textboard model. A recent change lowers the minimum thread title length from 5 to 1 character, with clients picking up the rule via JSON Schema.

Optimizing CPU-side Rendering Code

The article explains Godot’s CPU-side rendering optimization process, emphasizing that renderers must balance CPU and GPU costs and that engine-level optimizations benefit all games. It outlines a repeatable methodology—identify hotspots, understand causes, investigate fixes, and re-measure—and notes that CPU profilers are key for finding bottlenecks.

🐧 Linux & Open Source

WTF Ubuntu?

The article criticizes Ubuntu for failing to patch CVE-2026-98432, a critical remotely exploitable Linux kernel TCP bug affecting all supported releases, more than two weeks after disclosure and nine days after its CVSS 9.8 rating, while Debian patched it on October 16. It contrasts Ubuntu’s “timely” enterprise security promises with volunteer-run distros that have shipped fixes. The author says they have lost trust in Ubuntu, taken their Ubuntu servers offline, and are migrating off it.

I’ve updated my Linux System Call website and API

The Linux System Call website and API, first released in 2020, have been updated. The update adds more architectures, a kernel selector, syscall descriptions, and mobile improvements; the API is available at https://api.syscall.sh.

OccupySaaS

OccupySaaS is a code-based protest movement by laid-off engineers building clean-room, open-source, drop-in replacements for costly SaaS platforms using public documentation, open standards, and AI. The project claims strict clean-room provenance and compatibility testing against vendors’ public SDKs, with early efforts targeting identity and CRM, including ZeroAuthy as an Auth0-compatible identity server.

Radeon RADV Vulkan Driver Adopts An LLM Policy

The Mesa RADV Vulkan driver for AMD Radeon GPUs on Linux has adopted a loose policy on AI/LLM code contributions that discourages LLM-assisted merge requests and permits maintainers to close them at their discretion. Initiated by Valve’s Samuel Pitoiset, the policy also discourages LLM-written issues and was prompted by a rise in AI-generated patches.

FreeRDP 3.33 Released With WebAssembly Support, Huge Influx Of Activity Due To AI

FreeRDP 3.33 has been released with new WebAssembly support that allows it to run in a web browser when compiled to WASM. The release also improves the SDL client, adds experimental UDP support for direct RDP connections, Unicode keyboard input across all clients, experimental Windows shadow server support, macOS clipboard support without FUSE, and initial message translations. FreeRDP says a huge influx of AI/LLM-related pull requests and bug feedback will prompt a patch release soon, and version 3.33 includes 13 known security fixes.

KDE Linux Nearing Its Beta Milestone, Now Ships With NTSYNC By Default

KDE Linux is nearing its beta milestone after more than a year of development, with Nate Graham’s October update noting it is roughly 85% of the way there. The distribution now enables the NTSYNC kernel module by default to improve Wine/Proton gaming performance, and its installer gained a basic slideshow. Remaining work includes swap support, Discover update improvements, dual-booting, and better Flatpak crash handling.

Steam On Linux Just Under 4% For October 2026

Valve’s October 2026 Steam Survey shows Linux gaming use up 0.05% to 3.95%, after September was revised to 3.9%. macOS stood at 2.35% and Windows at 93.7%, while CachyOS remained the most popular Linux distribution at 16.3%, ahead of SteamOS at 13.89%. AMD CPUs were used by just under 70% of Linux gamers.

Polyend releases the Tracker firmware source code

Polyend has released the firmware source code for the original Tracker, Tracker+, and Tracker Mini on GitHub and launched Tracker Community Edition 1.0, letting users install, modify, and share firmware with a recovery system to prevent bricking. Community Edition 1.0 adds SD-card streaming for long samples, dedicated timing/velocity fields, Sample Editor plugin support, new synths for Tracker+ and Mini, wavetable envelopes and LFOs, faster project loading/saving, tighter MIDI clocking, and customizable themes/fonts. Polyend says the release supports community creativity and unofficial Tracker mods.

The advantages of mobile Linux over AOSP

The article collects reactions to a prior piece on whether Linux phones can challenge the Android/iOS duopoly, drawing over 660 comments. Many commenters doubted non-Android Linux phones, arguing that AOSP already offers an open-source, Linux-based mobile OS with better app compatibility, security, usability, and hardware support, while others stressed users’ lock-in to Google and Apple app ecosystems.

💼 Tech Business & Policy

Reaping and sowing, CEO edition

Proton CEO Andy Yen canceled planned company-wide “re-interviews” meant to assess employees’ AI use after backlash from workers and users over his aggressive AI push. The reversal followed a 404 Media investigation reporting internal security concerns and Yen’s demand that staff use AI more and prioritize speed. Many Proton users have criticized the company’s deprioritization of security and called for Yen to resign.

The U.S. Is Bullying Europe Like Never Before

The provided article contains no substantive news content; it only displays a Cloudflare security verification page asking users to wait while the site connection is confirmed. No details supporting or explaining the headline’s claim are available.

Jobless Tech Workers Are Being Left Out of San Francisco’s AI Boom

San Francisco’s AI boom, fueled by Anthropic and OpenAI, is generating enormous wealth and anticipation of lucrative IPOs even as layoffs and an uncertain job market leave many tech workers behind. A recent hike by dozens of laid-off workers highlighted the contrast between the city’s AI-driven riches and the joblessness nearby.

Anthropic amends IPO prospectus with Q3 results

Anthropic’s amended IPO prospectus shows rapid growth alongside heavy losses: nine-month revenue was $33.24 billion, up about 1,200% year over year, while Q3 revenue reached roughly $17 billion, a 12x increase, and Q3 net loss widened to $23.3 billion as operating expenses jumped 292% to $17.5 billion. The filing reported a $104 billion nine-month net loss—largely from convertible-note and derivative fair-value changes—and about $895.4 billion in committed spend, mostly for compute and infrastructure. Critics say the prospectus leaves tough questions unanswered, including an unexplained $80.7 billion accounting-related loss and reliance on customers that are also competitors.

Electrocutions, disease and no pay: worker exploitation in Europe’s datacentres

A Guardian and Investigate Europe investigation found migrant workers on Europe’s booming datacentre sites facing exploitation including electrocutions, disease and unpaid work. It detailed how Belgian electrical tester Sandu was electrocuted by live cables at a Google site after nearly 10 hours of work, suffering severe burns, while no company or contractor has been found liable. The five-month probe spans Sweden, Norway, Belgium and the Netherlands and highlights safety and labour abuses amid rapid datacentre construction.

BioWare co-founder: game execs “have never made a game” and don’t get it

BioWare co-founder Greg Zeschuk criticized game industry executives for never having made a game themselves and not understanding how development works. Responding to BioWare’s closure by EA, he said it is sad—especially for Edmonton—and expressed greater concern for employees facing job loss. He also said he has no regrets about the EA acquisition but doubted BioWare would fit under Saudi Arabia’s PIF and dismissed the idea that great games can be recreated with random people and AI.

China’s First Frontier AI Export Controls: Mofcom Announcement No. 48

China’s MOFCOM and MOST issued Announcement No. 48 on October 16, 2026, proposing to add 40 technologies to the export-control catalogue, including 23 AI technologies, while imposing an interim pre-export registration requirement that is already in effect. The draft targets frontier model weights, agent architectures, and AI-enabled offensive cyber capabilities using explicit technical thresholds, with public comment open until November 15, 2026.

America on Data Centers: listen to 10,000+ city council comments

An analysis of 300,000+ local government meetings found strong opposition to data centers: 73% of meetings had no supporters, and where supporters appeared they were outnumbered 6-to-1. Water use was the top concern, with most comments raising multiple issues including distrust, transparency, and skepticism about promised community benefits; AI itself was not a leading concern, and Big Tech was criticized far more than AI labs. Opposition was similar in red and blue counties, while union workers were the only group that mostly supported data centers.

Microsoft’s moolah mountain is more massive than ever. It just doesn’t want the people who built it

Microsoft’s cash reserves are larger than ever, but the company is not retaining the people who helped build its business. The article highlights a disconnect between Microsoft’s massive wealth and its treatment of the workers behind it.

Britain’s driverless dream hits a roundabout of awkward questions

Britain’s driverless transport ambitions are facing scrutiny from the Institute for Driverless Transport, which wants answers on the impact on jobs and what happens if a supplier withdraws. The concerns highlight unresolved questions surrounding the push toward autonomous transport.

India floats AI plan: model travels to data, not data to model

India has floated an AI plan centered on the principle that “the model travels to the data, rather than the data travelling to the model.” The proposal aims to encourage local AI development while preventing Big Tech from getting open access to everything.

Google says Asian advertisers spending less off due to Europe’s customs changes

Google says Asian advertisers are spending less due to Europe’s customs changes. The digest also notes brake pedals breaking on Huawei-made cars, North Korea being named as the source of the Binance hack, and other tech and security stories.

Raspberry Pi 5 16GB Price Rises To $385 USD

Raspberry Pi has raised prices on its 16GB RAM products due to ongoing RAM shortages and increased memory costs. The Raspberry Pi 5 16GB is increasing by $80 to $385 USD, while the Compute Module 5 16GB is rising by $40.

Oracle selling $10B stock and $10B bonds to fund data center buildout; ORCL falls 5%+

Oracle plans to raise $20 billion through debt and equity to finance its artificial intelligence infrastructure buildout. The company is selling $10 billion in common stock and $10 billion in bonds, according to separate SEC filings on Monday.

Marvell’s $115B Lumentum buy is 2026’s top M&A and among biggest tech deals; Lumentum up 128% YTD

Marvell Technology has agreed to acquire data center optics maker Lumentum Holdings in an all-stock deal valued at $115 billion enterprise value, one of the largest tech tie-ups ever and 2026’s top M&A deal. Marvell shareholders will own 67% of the combined company as Marvell seeks to capitalize on the AI boom.

Sources: SpaceX in talks with EchoStar for Boost Mobile or network access after carrier rebuffs

SpaceX is in talks with EchoStar to acquire Boost Mobile or access its network, potentially giving Elon Musk a route into the $235 billion U.S. wireless market. The move follows rebuffed network-access requests to Verizon, T-Mobile, and AT&T.

Filing: Anthropic to raise up to $3.5B from Japanese investors in IPO

Anthropic PBC plans to raise up to $3.5 billion from Japanese investors as part of its US initial public offering. Mizuho Securities will manage the Japanese portion of the share sale, with Goldman Sachs, Mitsubishi UFJ Morgan Stanley Securities, and Morgan Stanley MUFG Securities among those handling investor orders. Mizuho Financial Group and Mitsubishi UFJ Financial Group are among the underwriters.

Kioxia shares fall up to 15% as cautious Q3 outlook stokes AI slowdown fears

Kioxia Holdings shares fell as much as 17% on Monday, their biggest drop in three months, after the memory maker forecast only scant operating-profit growth for the December quarter. The cautious outlook intensified investor concerns that AI-driven demand growth may be leveling off.

Bites raises $40M Series A led by PeakSpan for AI frontline training

Bites, an AI platform that delivers personalized frontline-worker training through familiar channels such as WhatsApp, SMS, Slack, and Teams, raised a $40 million Series A led by PeakSpan Capital, bringing total funding to $55 million. The company plans to accelerate U.S. expansion, develop its AI platform, and build technology that links training to measurable business outcomes.

Casap raises $56M Series B led by Base10, total $90M, for bank dispute/anti-fraud AI

Casap, a startup using AI to automate dispute and anti-fraud operations for banks and card companies, raised a $56M Series B led by Base10 Partners, bringing its total funding to $90M. Synchrony and Experian joined the round as strategic investors. Casap says its tools help reduce first-party fraud losses and speed up dispute resolution as cardholder fraud accounts for a growing share of fraudulent transactions.

AI medical search engine Vera Health raises $28.5M Series A led by Threshold Ventures

Vera Health, an AI-powered clinical search engine, raised a $28.5 million Series A led by Threshold Ventures with participation from Optum Ventures, Mantis VC, SV Angel, Y Combinator, and Gradient. The platform, which scans more than 60 million peer-reviewed publications and other medical sources, says it is used by over 800,000 clinicians in 190 countries. Vera plans to use the funding to expand its evidence infrastructure, international reach, partnerships, and team.

Midstream Health raises $20M Series A led by Valtrius for AI hospital revenue recovery

Midstream Health, whose AI tools help 250 hospitals recover revenue from rebates and vendor and payer underpayments, raised a $20 million Series A led by Valtrius. The funding will support expansion of its AI financial platform for hospitals.

Broccoli AI, AI agents for plumbing/HVAC/electrical, raises $52M at $300M valuation

Broccoli AI, which sells AI agents to plumbing, HVAC, and electrical contractors, raised a $22M Series A and a $30M Series B led by Khosla Ventures, with Y Combinator and Operator Partners participating, for $52M total. A source says it is valued at $300M post-money; the company reports more than 1,500 business customers and revenue tripled since the Series A.

White House: Full AI testing framework draft can be withheld on security, commercial grounds

The Trump administration declined to release key documents related to its voluntary AI testing framework, arguing it is not required to share the full draft. It cited national security and commercial exemptions as grounds for withholding the material.

Hyperbolic raises $30M Series A-1 led by Walden, $50M total for AI compute rentals across neoclouds

Hyperbolic raised a $30M Series A-1 led by Lip-Bu Tan’s Walden Catalyst, bringing its total funding to $50M. The startup helps companies rent and manage AI compute across about 30 neoclouds and aims to let customers resell unused compute. It says roughly $200M worth of compute is connected to its software annually and that it is already breaking even.

🌍 Society, Science & Culture

Why are Soviet math textbooks so hardcore in comparison to US textbooks? (2017)

The article explores why Soviet-era math textbooks are often perceived as more rigorous than their US counterparts. It examines differences in pedagogical approach and problem difficulty to explain the “hardcore” reputation.

Napping Characteristics and Incident Dementia

The provided article content is inaccessible because it requires JavaScript and cookies to be enabled. No substantive findings about napping characteristics and incident dementia can be summarized from the available text.

School has given straight-a grades to the student, who can’t do masic math

Adriana Aviles, an education advocate and NYC schools insider, discovered her 14-year-old son—despite years of A and B-plus math grades in District 26—could not solve basic algebra, geometry, or fractions when helping him study for a high school exam. He then scored 10 out of 100 on a math test at his top Queens high school, prompting her to confront Schools Chancellor Kamar Samuels and blame the school system for failing him. Her son now receives remedial tutoring and is using elementary-level math books to identify where his learning stalled.

Pray for Chaos

The article argues that AI development will lead either to “order,” a dangerous scenario in which AI stabilizes existing systems and closes exploits, or to “chaos,” which the author considers the preferable if deeply disruptive outcome. It describes chaos as involving synthetic histories, a bot-dominated dead Internet, decentralized but vulnerable infrastructure, self-replicating AI, hacking, and microtargeted media. The author concludes that chaos will upend lives and institutions, but that the alternative is a controlled status quo in which people are unable to perceive their own subjugation.

Private Jet plane crashes in Florida

A business jet carrying nine people crashed in Palm Beach County, Florida, on Sunday, killing all aboard after nosediving near a roadway and exploding. No ground injuries were immediately reported, and federal agencies are investigating the cause. The pilot had reported turbulence and icing before issuing distress calls as the aircraft entered a steep descent.

Never Complain, Never Explain

“Never complain, never explain” is a public relations strategy closely associated with the British royal family and attributed to Benjamin Disraeli, with origins traced to maxims by Benjamin Jowett. It became central to Queen Elizabeth II’s reign, but Prince Charles and Diana abandoned it in the 1990s amid a more aggressive media environment. Prince Harry referenced the family motto in a 2023 interview while saying Buckingham Palace still briefs correspondents.

How Wikipedia Rewrote Zionism

Since October 7, 2023, the English Wikipedia article on Zionism has been substantially rewritten, with about 80% of its text changed and increased emphasis on colonialism and ethnic cleansing while coverage of antisemitism and persecution was reduced. The article’s new opening describes Zionism as establishing a homeland “through colonization,” a claim backed by 17 citations but introduced by an editor later banned from the Palestine-Israel topic area for non-neutral editing. The changes raise concerns about Wikipedia’s neutrality and its influence, especially as its content is used to train AI systems.

The future in old computer magazines (1984–2004)

This collection surveys early home-computing advertisements and reviews from 1984–1994, covering promises such as home automation, online socializing, wine pairing, AI games, personality software, robot pets, modems, business simulators, productivity tools, and integrated office machines. Contemporary accounts found some products dependable or useful, but often noted confusing instructions, limited features, unproven claims, and cumbersome setup. The entries contrast marketing hype with practical limitations and actual user experiences.

Do you have a Plan B?

The article asks whether readers have a contingency plan to cope with growing layoffs, a possible market crash, and ongoing regional wars, working it out with family or friends. It frames the question as non-alarmist and invites others to share their views.

Wir müssen wissen – wir werden wissen

A century ago, paradoxes such as Russell’s exposed a foundational crisis in mathematics, prompting Hilbert to seek a complete, consistent axiomatic system and a mechanical procedure for deciding mathematical truth. Gödel’s incompleteness theorems and Turing’s undecidability results showed these goals impossible, but mathematicians adapted through improved set and type theories that now underpin formal proof assistants like Lean. The article argues that AI agents using such mechanized systems are now realizing Hilbert’s “proof machine,” with potentially profound consequences for mathematics.

From Charlemagne to the Kitchen Boy: Reading the Hamburg Rathaus

The Hamburg Rathaus can be read as a history of the city through its many architectural details. The article traces its predecessors and destruction in the 1842 Great Fire, its 1897 reconstruction, and symbols such as the copper roof’s patina and the Phoenix and “Resurgam” motto marking Hamburg’s rebirth.

Bones may release proteins that slow aging and extend life span

The article’s title suggests that bones may release proteins capable of slowing aging and extending lifespan. However, the provided page content was unavailable because it required JavaScript and cookies, so no further details could be summarized.

Independent Digital Media Isn’t a Privilege Governments Can Revoke: Saving Egypt’s Mada Masr

Lina Attalah, founder and editor-in-chief of the independent Egyptian outlet Mada Masr, is set to face trial again on November 11 on charges of spreading false news, after an October 27 conviction for managing an unlicensed website led to a maximum fine, the outlet’s closure, and equipment confiscation. The article says Egypt’s licensing regime and false-news charges are increasingly used to suppress independent journalism, citing the violent arrest of the Matsada2sh fact-checking team and at least 25 other jailed journalists. It also highlights Mada Masr’s decade of independent reporting and its continued reach despite being blocked in Egypt since 2017.

Manager threatened to fire a colleague, then gave them a critical task – which they flubbed

A manager threatened to fire a colleague and then assigned them a critical task, which the colleague botched. The failure went unnoticed for years until the boss realized they had to cover up the poor decision.

The Titanic didn’t sink for just one reason

The article uses the Titanic disaster and a Five Whys analysis to argue that complex incidents rarely have a single root cause. It advocates shifting from Root Cause Analysis to a contributing-factors approach, especially in software engineering, to better understand how failures happen through multiple perspectives and details.

SuperVote: 15 year retrospective on a social endorsement app

The article is a 15-year retrospective on SuperVote, a 2012 iPhone app that let Facebook users share political endorsements with friends through a shared ballot. It revisits the app’s origins, development, Facebook integration, marketing, eventual outcome, and what the author would do differently today.

What are you doing this week?

The article invites readers to share what they are doing this week. It also notes that it is acceptable to do nothing at all.

JPEG XS - Wikipedia

JPEG XS (ISO/IEC 21122) is an image and video codec developed by the Joint Photographic Experts Group and initially released in May 2019. It provides visually and mathematically lossless compression with low latency and low complexity, targeting professional video over IP, broadcast, VR, drones, automotive cameras, and gaming.