Start your morning scandalously informed.

A daily curated digest with the tech news that matter + community vibes, delivered daily, in tabloid style. Like you always wanted.

← Back to Home

Thursday, April 23, 2026

Firefox Flaw Tags Tor Users In Secret!

Firefox Flaw Tags Tor Users In Secret!

Today’s tech world stares hard at privacy as a hidden Firefox bug quietly links supposedly separate Tor sessions across tabs... Users learn that “anonymous” browsing is more fragile than it looks while Apple races to patch an iPhone hole that let police tools pull back “deleted” messages... Developer trust shakes as GitHub CLI starts sending new telemetry by default, and a California neighborhood pushes back on the data center next door, reminding us the cloud has a loud, diesel heartbeat... Meanwhile, AI power plays escalate: Google unveils a monster TPU chip, OpenAI pushes Workspace Agents into the office, and Meta staff revolt over invasive workplace surveillance for model training... New EV batteries promise near‑instant charging, the Zed editor turns swarms of AI agents into a coding team, and startups brag they spend more on GPUs than people as we edge toward offices run by scripts in suits.

Tech Clashes With Privacy And The Real World

  • Firefox Bug Exposes Hidden ID Linking Tor Sessions

    Researchers found a nasty Firefox quirk that lets websites create a stable ID from how browser processes lay out memory, quietly tagging you across tabs. That even hits hardened setups like Tor, which is supposed to keep identities separate. It’s subtle, clever, and makes anonymous browsing feel a lot shakier than most of us assumed.

  • Apple Plugs iPhone Hole Cops Quietly Exploited

    Apple pushed an iOS update that fixes a bug letting police tools pull back messages that were deleted or set to auto‑disappear in apps like Signal. Users are relieved but annoyed it existed this long, and it’s a blunt reminder that when the system keeps stray copies, the word “deleted” is doing a lot of marketing work.

  • GitHub CLI Starts Phoning Home With New Telemetry

    GitHub quietly turned on “pseudoanonymous” telemetry in its CLI tool, sending usage data back by default. They insist it’s harmless and helps improve the product, but command‑line diehards hate surprise tracking. Yes, you can opt out, but the move pokes directly at the fragile trust developers place in their day‑to‑day tools.

  • Neighbors Unite To Kill Giant Backyard Data Center

    Residents in Monterey Park learned a huge data center with diesel backup generators was about to land just 500 feet from homes, so they organized and killed it. Locals worried about fumes, noise, and power use, and they actually won. It’s a warning shot: the physical footprint of the cloud is finally meeting real‑world pushback.

  • Chinese Battery Promises Near Empty To Full In Minutes

    CATL showed off a new LFP car battery it says can jump from 10% to 98% charge in under seven minutes and still work in freezing weather. EV fans love the idea, but everyone wants to see real‑world tests, grid impact, and how long these packs actually last before crowning this the end of range anxiety.

AI Labs Race To Automate Every Last Task

  • Google Unveils Monster TPU Chip Built For AI

    Google pulled apart its eighth‑gen TPU design, built to train gargantuan models more efficiently and cheaply. The chip leans hard into mixture‑of‑experts and scale, screaming “we can still play with Nvidia.” It’s impressive silicon, but also a reminder that only a handful of labs can even afford to use toys this big.

  • OpenAI Agents Aim To Live Inside Your Workplace

    OpenAI launched Workspace Agents, shared bots that roam your tools, updating tickets, editing documents, and sending messages without you babysitting every step. It sounds like the end of busywork and the start of “who approved this bot to touch our CRM.” Office life just took one step closer to being quietly run by scripts in suits.

  • Meta Staff Revolt Over New Spyware For AI Training

    Meta is reportedly installing software that logs keystrokes and mouse moves on employee PCs to feed its AI push, under banners like “Model Capability Initiative.” Workers are furious at being turned into lab rats on corporate hardware. Coming from the company that tracks the whole world, this level of internal surveillance still manages to feel like a new low.

  • Code Editor Zed Turns AI Agents Into Team

    The Zed editor now lets you run multiple AI agents in parallel, each with scoped access to specific folders and repos. It’s like having a tiny team of junior devs living inside your editor. People love the power but worry that at some point, the human is just supervising bots arguing over what to refactor next.

  • Startups Boast They Spend More On GPUs Than People

    A new crop of founders proudly says they’re pouring cash into GPU time and LLM calls instead of salaries, “tokenmaxxing” every process they can. It’s equal parts clever efficiency and dystopian brag—like boasting that your company is mostly scripts with a few humans stapled on. Great for margins now, maybe, but what happens when the compute bill comes due?

Nerdy Delights From Fonts To Fake Influencers

  • Right Wing Bombshell Influencer Turns Out To Be AI

    A wildly popular bikini‑clad MAGA influencer turned out to be an AI‑generated persona run from India, using tools like Google Gemini and Grok AI to churn out patriotic thirst‑trap content. Fans feel duped, but honestly the whole thing just proves how easily clout, politics, and deepfakes mix when there’s subscription money on the table.

  • Designers Say Adobe Finally Reaps Years Of Greed Back

    A blistering takedown of Adobe argues that endless subscriptions, heavy apps, and pushy upsells have driven creatives to look elsewhere just as new tools arrive and regulators sniff around. For many designers who already rage‑quit Creative Cloud, the essay feels less like hot take and more like a long‑overdue “told you so.”

  • Sony Ping Pong Robot Starts Beating Serious Human Players

    Sony AI built a ping‑pong robot named Ace that now beats top‑level human players under official rules. It tracks the ball, moves with inhuman consistency, and calmly sends back shots most people can’t even see. It’s thrilling tech, but also a little eerie watching a machine turn a friendly table sport into a one‑sided clinic.

  • New Rip Language Promises Easier JavaScript With Reactivity

    Rip is a new language that compiles to modern JavaScript, adding extra operators and built‑in reactivity so you can write fewer hooks and boilerplate. It has strong CoffeeScript vibes: some devs are intrigued by the cleaner syntax, others are groaning “not another compile‑to‑JS toy” and waiting to see if it survives the hype cycle.

  • How Shazam Hears A Song And Knows It Instantly

    A deep dive on Shazam shows how it turns music into sparse audio fingerprints using spectrograms and FFTs, then matches those patterns in a massive database. It’s surprisingly elegant: instead of magically “knowing” songs, the app is basically doing a super‑fast connect‑the‑dots trick that makes your phone feel smarter than it really is.

Top Stories

Google rolls out monster TPU 8 chips

Artificial Intelligence

Google lifted the curtain on its eighth‑generation TPU design, a custom AI chip meant to power gigantic models and keep Google in the hardware arms race with Nvidia and other cloud giants.

Read more →Comments

OpenAI wants agents living inside your job

Artificial Intelligence

OpenAI’s new Workspace Agents promise bots that roam across tickets, docs, and chats doing tasks on your behalf. It sounds wildly convenient and just as wildly ripe for permission nightmares at work.

Read more →Comments

Meta staff revolt over spyware for AI training

Technology

Meta is reportedly pushing keylogging software onto employee machines to harvest data for AI training. Even at the company built on surveillance, workers are treating this as a creep‑out moment.

Read more →Comments

Firefox flaw quietly linked Tor identities together

Security

Researchers exposed a Firefox bug that lets sites derive a stable ID from how memory is laid out, potentially tying all your Tor sessions together. For privacy die‑hards, this lands like a gut punch.

Read more →Comments

Apple closes iPhone hole cops used for 'deleted' chats

Security

A new iOS update finally fixes a bug that let law enforcement tools recover supposedly deleted or disappearing messages. It’s a harsh reminder that ‘delete’ on your phone doesn’t always mean gone.

Read more →Comments

Worldcoin cozies up to Zoom and Tinder

Technology

Sam Altman’s eyeball‑scanning Worldcoin is cutting deals with Zoom and Tinder, pushing its ID system deeper into everyday apps. People are seriously asking whether logging in now means scanning your soul.

Read more →Comments

Startups brag they spend more on GPUs than people

Business

A new breed of founders claims they’d rather burn cash on AI compute than salaries, openly ‘tokenmaxxing’ instead of hiring. It feels like both the future of work and the setup for a nasty hangover.

Read more →Comments
All stories (84)
Complete list of news articles from this day

Garbage Collection Without Unsafe Code

by foota

The article presents safe-gc, a Rust garbage collection library that avoids unsafe code entirely in both its API and implementation, enforced via a forbid(unsafe_code) pragma. It contrasts this with p...

Key Points

  • safe-gc implements GC in Rust with zero unsafe code and a safe Trace trait.
  • Objects are allocated in per-heap arenas and accessed via heap indexing.
  • The design prioritizes soundness under Rust’s borrow rules over performance.

Contact Lens Uses Microfluidics to Monitor and Treat Glaucoma

by pseudolus

A research team has created an electronics-free, all-polymer smart contact lens that addresses two major challenges in glaucoma care: continuous measurement of intraocular pressure (IOP) and patient a...

Key Points

  • Electronics-free, all-polymer lens uses microfluidics for IOP sensing and closed-loop drug delivery.
  • CNN-powered phone app reads microchannel changes, achieving 94% pressure-readout accuracy.
  • Design enables tunable, staged release of two drugs and up to two weeks of medication.

XOR'ing a register with itself is the idiom for zeroing it out. Why not sub?

by ingve

This article examines why the x86 ecosystem predominantly uses the instruction idiom “xor r, r” to zero registers instead of alternatives like “sub r, r.” It starts from Matt Godbolt’s observation tha...

Key Points

  • “xor r, r” is a compact, widely used x86 zeroing idiom; “mov r, 0” is larger.
  • Intel front-ends special-case both “xor r, r” and “sub r, r,” effectively making them zero-cost and dependency-breaking.
  • Itanium’s NaT bit prevents the xor trick; a zero register is available instead.

Rock carving facts – Tanum Sweden

by janandonly

The Tanum World Heritage area in Sweden hosts more than 600 rock carving sites dating from about 1700 to 300 BCE. Created by hammering with pounding stones, these carvings emphasize themes of power, r...

Key Points

  • Over 600 Bronze/Iron Age rock carving sites exist in Tanum, Sweden.
  • Dating uses artifact comparison, radiocarbon methods, landscape analysis, and overcut chronologies.
  • Post-glacial rebound shifted many panels inland and higher; ships are a common motif across sites.

What Async Promised and What It Delivered

by zdw

The article traces how asynchronous programming evolved to tackle the C10K problem—serving thousands of concurrent connections without a thread per connection. It explains why OS threads are ill-suite...

Key Points

  • Event-driven callbacks replaced thread-per-connection to solve C10K-scale performance limits.
  • Callbacks introduced complexity in control flow, error handling, and cancellation.
  • Promises/futures standardized in ES2015 and Java 8 improved composition and error propagation.

All your agents are going async

by zknill

The article argues that AI agents are transitioning from chat-style, synchronous interactions to asynchronous, background workflows that operate across devices and channels. It highlights new capabili...

Key Points

  • Agents are moving from synchronous chat to asynchronous, background operations.
  • HTTP/SSE streaming is ill-suited for long-running, decoupled agent workflows.
  • Vendors and tools (e.g., Anthropic, OpenClaw, ChatGPT, Cursor, Temporal/Vercel WDK/Relay.app) are adding async capabilities.

MuJoCo – Advanced Physics Simulation

by modinfo

MuJoCo (Multi-Joint dynamics with Contact) is a fast, accurate physics engine for simulating articulated bodies across domains such as robotics, biomechanics, graphics, and machine learning. Maintaine...

Key Points

  • Maintained by Google DeepMind, MuJoCo offers C/Python APIs, OpenGL GUI, and a Unity plug-in.
  • Quick start via native ‘simulate’ viewer and Colab tutorials covering topics from basics to MJX/JAX and differentiable physics.
  • Cross-platform binaries on GitHub and pip-based Python installation; monthly releases with modified SemVer.

Ask HN: Why Opus4.6 was silently removed from Claude Code?

by t0duf0du

An Ask HN post reports that Claude Opus 4.6, which the author says was functioning after prior cache issues were resolved, appears to have been removed from Claude Code following the release of Claude...

Key Points

  • User claims Opus 4.6 was removed from Claude Code after Opus 4.7 launched.
  • The removal is described as silent; no official rationale is provided in the post.
  • The post seeks community explanations for the model version change.

Meta employees are up in arms over a mandatory program to train AI on their

by latein

Meta is rolling out a data collection tool on US employees’ work computers to help train its AI models on real-world computer interactions, according to an internal announcement obtained by Business I...

Key Points

  • MCI captures inputs and screen context on work computers to train AI on real user workflows.
  • Data collection is limited to specified work apps/URLs and excludes phones; no opt-out on work laptops.
  • The program fits Meta’s larger AI push, including MSU and launch of the Muse Spark LLM.

Windows 9x Subsystem for Linux

by sohkamyung

Hailey unveiled “Windows 9x Subsystem for Linux” (WSL9x), a project that runs a modern Linux kernel cooperatively with the Windows 9x kernel to allow Windows and Linux applications to operate side-by-...

Key Points

  • WSL9x runs a modern Linux kernel cooperatively with the Windows 9x kernel without hardware virtualization.
  • Demonstrated on Windows 95 with Linux tools (fastfetch, ps, ls /dev, mount, uname -a, tty) running via MS-DOS prompts.
  • Source and details are available on Codeberg at codeberg.org/hails/wsl9x.

How the Heck Does GPS Work?

by alfanick

The article explains how GPS determines location by translating signal travel time into distance and combining measurements from multiple satellites. Each satellite broadcasts a timestamped signal at ...

Key Points

  • GPS computes distance from satellites by timing signals traveling at light speed.
  • Trilateration with at least three satellites determines a unique 3D position.
  • A fourth satellite measurement corrects the receiver’s clock bias for accuracy.

Context Is Software, Weights Are Hardware

by maxaravind

This article examines how large language models learn and adapt, contrasting two mechanisms that shape transformer activations: in-context learning via the KV cache and fine-tuning via weight updates....

Key Points

  • Context (KV cache) and weights both modulate activations; only weight updates persist.
  • Theoretical results equate in-context learning with a gradient step in linear attention cases.
  • Longer context helps, but weight changes add capabilities beyond what prompts can provide.

Irony as Meta staff unhappy about running surveillance software on work PCs

by jjgreen

Meta plans to install a data-collection tool, the “Model Capability Initiative,” on employee work computers to capture keystrokes, mouse activity, and occasional screenshots. Reuters and Business Insi...

Key Points

  • Meta will record certain employee device activities to train AI models via the “Model Capability Initiative.”
  • Monitoring applies to specified work-related apps/URLs such as Gmail, GChat, VCCode, and Metamate.
  • The effort aligns with broader industry moves toward AI agents by Anthropic, OpenAI, and Microsoft.

Prefill-as-a-Service:KVCache of Next-Generation Models Could Go Cross-Datacenter

by matt_d

The article presents Prefill-as-a-Service (PrfaaS), a cross-datacenter architecture for serving large language models that tackles the longstanding constraint of KVCache transfer in prefill-decode (PD...

Key Points

  • PrfaaS offloads long-context prefill and ships KVCache over commodity Ethernet to local decode clusters.
  • Combines model-side KV efficiency with bandwidth-aware scheduling and cache-aware placement.
  • Delivers up to 54% throughput gain over homogeneous PD with modest cross-datacenter bandwidth.

CATL's new LFP battery can charge from 10 to 98% in less than 7 minutes

by PotatoNinja

CATL has introduced the third-generation Shenxing lithium-iron phosphate (LFP) battery, emphasizing ultra-fast charging and robust cold-weather performance. The company claims the pack charges from 10...

Key Points

  • Shenxing 3.0 claims 10–98% charge in 6:27 and 10–80% in 3:44.
  • Performs at -22°F with a 10–98% charge in 9 minutes.
  • Compared against BYD Blade 2.0 and 800 V NMC systems from Hyundai and Porsche.

Iran claims US exploited networking equipment backdoors during strikes

by pseudolus

Iranian state media alleges that networking gear from Cisco, Juniper, Fortinet, and MikroTik malfunctioned during recent U.S. and Israeli military operations, implying hidden firmware or backdoors ena...

Key Points

  • Iran alleges backdoors in U.S.-made networking gear caused device failures during strikes; claims are unverified.
  • U.S. acknowledges cyber and space operations disrupted Iranian networks ahead of attacks.
  • Historical vulnerabilities at Cisco, Juniper, Fortinet, and MikroTik provide context for the allegations.

Nobody Got Fired for Uber's $8M Ledger Mistake?

by ohduran

Alvaro Duran examines Uber’s decision to build its 2017 payment platform on Amazon DynamoDB and the subsequent shift away from that architecture due to cost at Uber’s scale. He emphasizes that DynamoD...

Key Points

  • DynamoDB’s consumption pricing drove high costs for Uber’s ledger workload at scale.
  • Uber shifted to hot/cold data tiering and later migrated to LedgerStore from DynamoDB.
  • DynamoDB’s consistency model can fit payments but not ledger requirements, per the author.

3.4M Solar Panels

by marklit

Version 2 of the Ground-Mounted Solar Energy in the United States (GM-SEUS) dataset has been released, expanding coverage from 2.9 million to over 3.4 million solar panels and adding a new rooftop arr...

Key Points

  • GM-SEUS v2 expands to over 3.4M panels and adds a rooftop arrays dataset.
  • Workflow uses GDAL, DuckDB (with extensions), and QGIS for processing and mapping.
  • Data is fetched from Zenodo; GPKG projection is Albers Equal Area; conversion to Parquet required DuckDB v1.4.4 for stability.

GitHub CLI now collects pseudoanonymous telemetry

by ingve

GitHub’s announcement details that GitHub CLI (gh) collects pseudoanonymous telemetry to inform product decisions and evaluate real-world feature usage. The team uses aggregated usage signals—such as ...

Key Points

  • Logging mode lets users inspect telemetry payloads before any data is sent.
  • Opt-out can be done via environment variables or CLI config, with env vars taking precedence.
  • Telemetry applies to GitHub CLI only; extensions and Copilot products handle data collection separately.

Our eighth generation TPUs: two chips for the agentic era

by xnx

Google announced the eighth generation of its Tensor Processing Units (TPUs), introducing two specialized chips—TPU 8t for training and TPU 8i for inference—built for the demands of AI agents and larg...

Key Points

  • Two specialized TPU chips: 8t for training and 8i for inference
  • TPU 8t superpod: 9,600 chips, 2 PB shared HBM, 121 exaflops
  • Near-linear scaling via Virgo Network + JAX + Pathways; TPUDirect accelerates data access

Kernel code removals driven by LLM-created security reports

by edward

Linux kernel maintainers are advancing proposals to prune legacy networking-related code to address a growing influx of security-bug reports attributed to large language models and persistent findings...

Key Points

  • Proposed removals focus on legacy networking subsystems and drivers (e.g., AX.25, ATM, ISDN).
  • Amateur radio protocols (AX.25/NET/ROM/ROSE) and drivers would be moved out of the mainline kernel tree.
  • The rationale centers on high bug report volume from syzbot and LLM-generated submissions and lack of maintainers.

Columnar Storage Is Normalization

by ibobev

The article explores how column-oriented storage fits within the relational database paradigm by contrasting it with row-oriented storage and framing it as an extreme form of normalization. In row-ori...

Key Points

  • Columnar storage can be viewed as per-attribute tables joined on a shared key.
  • Row-oriented excels at whole-row operations; columnar excels at column-specific scans.
  • Ordinal position can serve as an implicit key for reconstructing rows in columnar formats.

I don't chain everything in JavaScript anymore

by AllThingsSmitty

The article cautions against overusing chained JavaScript methods for both array transformations and promise-based async flows. While concise pipelines can appear elegant, they demand mental simulatio...

Key Points

  • Break long chains into explicit steps to improve readability and debugging.
  • Use early-exit patterns instead of processing entire arrays when only one result is needed.
  • Separate async control flow from transformations; prefer sequential awaits before array operations.

The eighth-generation TPU: An architecture deep dive

by meetpateltech

Google introduced its eighth-generation Tensor Processing Units, splitting capabilities into two systems—TPU 8t and TPU 8i—to address diverging needs across the AI lifecycle. As models shift toward Mi...

Key Points

  • Two specialized TPU systems (8t for pre-training, 8i for serving) underpin Google Cloud’s AI Hypercomputer.
  • TPU 8t adds SparseCore, native FP4, and balanced VPU/MXU scaling to boost training efficiency.
  • Virgo Network and large-scale 3D torus topology (9,600 chips per superpod) increase throughput and networking capacity.

Another Day Has Come

by ndr42

Apple announced a planned leadership transition: CEO Tim Cook will move into the role of executive chairman, and John Ternus will become CEO. The article contrasts this change with Apple’s 2011 transi...

Key Points

  • Tim Cook moves to executive chairman; John Ternus becomes Apple CEO.
  • Cook will engage with policymakers worldwide in his new role.
  • Apple’s product lineup is described as strong, including iPhone 17 and a $600 MacBook Neo.

Books Are Not Remotely Too Expensive

by gHeadphone

The article examines claims that new hardcover books are too expensive by comparing historical and current prices in real terms. Using examples like Harper Lee’s To Kill a Mockingbird ($3.95 in 1960) ...

Key Points

  • Inflation-adjusted comparisons show today’s hardcovers are cheaper than historical equivalents.
  • BLS data indicate slight deflation for recreational books since 1997.
  • Other sectors have outpaced inflation, making books relatively cheaper.

Monitor your Pi / OMP sessions

by ankitg12

PI Dashboard is a web-based interface for monitoring and controlling pi agent sessions. It delivers real-time session mirroring, bidirectional command input, and workspace organization with session st...

Key Points

  • Real-time monitoring and control of pi sessions with PromptBus-based unified routing
  • Integrated terminal, pi-flows dashboard, and Flow Architect for end-to-end flow management
  • Built-in provider auth, package management, OpenSpec tools, and mDNS discovery in a Node.js/React architecture

Treetops glowing during storms captured on film for first time

by t-3

Penn State researchers have documented the first natural observations of treetop corona discharges during thunderstorms, a phenomenon hypothesized for decades. In June 2024, after weeks chasing Florid...

Key Points

  • First direct natural observation of treetop corona discharges during thunderstorms
  • Captured with a Newtonian telescope–UV camera system near UNC Pembroke
  • UV emissions from corona can generate hydroxyl, contributing to atmospheric cleansing

Show HN submissions tripled and are now mostly vibe-coded

by hubraumhugo

The article examines a surge in Show HN submissions—described as tripled in recent years—and explores whether many of these projects share a similar, AI-influenced visual style. The author compiles a ...

Key Points

  • 500 Show HN pages were scanned with Playwright and CSS/DOM checks.
  • 21% heavy, 46% mild, 33% clean pattern prevalence.
  • HN restricted new-account Show HN submissions amid submission growth.

DuckDB 1.5.2 – SQL database that runs on laptop, server, in the browser

by janandonly

DuckDB 1.5.2 is a patch release focused on stability and performance that also broadens DuckDB’s data lakehouse capabilities. It introduces support for DuckLake v1.0, a backward-compatible, production...

Key Points

  • DuckDB 1.5.2 adds production-ready DuckLake v1.0 support and expands Iceberg features.
  • Jepsen collaboration yielded a discovered bug and fix; test suite is public.
  • New WebAssembly shell enables in-browser file workflows; benchmarks show ~10% TPC-H gain after OS upgrade.

Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model

by mfiguiere

The article announces Qwen3.6-27B, a new 27-billion-parameter dense multimodal model from the Qwen team that is fully open-source. Positioned at the community’s preferred scale, it is designed to offe...

Key Points

  • Qwen3.6-27B is a fully open-source 27B-parameter dense multimodal model optimized for high-level agentic coding.
  • It outperforms the prior Qwen3.5-397B-A17B MoE flagship and peer-scale dense models on major coding and several reasoning benchmarks.
  • The model is easier to deploy than MoE systems and is available via Qwen Studio, API access, and open weights on Hugging Face and ModelScope.

The best time to post on Hacker News

by alcazar

This article examines the best times to post on Hacker News and stresses that timing helps but does not guarantee success due to the platform’s multifactor ranking system (flags, anti-abuse mechanisms...

Key Points

  • Start with Tuesday–Thursday, 14:00–17:00 UTC for broad technical posts.
  • Separate goals: more active readers vs fewer competing submissions; choose timing accordingly.
  • HN ranking factors beyond votes/time mean timing is only a marginal advantage.

Top MAGA influencer revealed to be AI

by java-man

A 22-year-old Indian orthopedic surgery trainee told Wired he engineered an AI-generated persona, “Emily Hart,” tailored to appeal to U.S. conservatives after using Google’s Gemini AI for audience and...

Key Points

  • AI guidance (Google Gemini) informed creation of a conservative-targeted influencer persona.
  • Monetization via merchandise and Fanvue subscriptions used AI-generated images (Grok AI).
  • Instagram removed the account for fraudulent activity; a related Facebook page was also taken down.

Windows 9x Subsystem for Linux

by luu

Windows 9x Subsystem for Linux (WSL9x) is a project that embeds a modern Linux kernel, noted as version 6.19, cooperatively within the Windows 9x kernel. This approach brings key OS features—paging, m...

Key Points

  • Linux kernel runs cooperatively inside Windows 9x to enable side-by-side app use.
  • Build requires i386-linux-musl cross toolchain and Open Watcom v2; patched kernel from win9x-um-6.19.
  • Run via 'wsl' in MS-DOS; optional ANSI colors with nnansi.com; GPL-3 license.

Bodega Cats of New York

by zdw

“Bodega Cats of New York” is a forthcoming photo-and-story collection documenting the working cats of NYC bodegas. Opening with a vignette about a polydactyl cat, the article outlines how these animal...

Key Points

  • Photo book with 120 images and 60+ stories releases October 2026 via Quarto Publishing.
  • Legal reforms (Int. 1471, A08341) are proposed to legalize bodega cats; 14,000-signature petition backs them.
  • Brand placement services and historical walking tours expand the project beyond the book.

Sam Altman's Creepy Eyeball-Scanning Company Gets in Bed with Zoom and Tinder

by speckx

Tinder and Zoom have partnered with World, the biometric identity company associated with Sam Altman, to introduce human verification features. Tinder, which previously piloted the process in Japan, w...

Key Points

  • Tinder and Zoom announced partnerships to use World’s biometric verification.
  • Tinder expands a Japan pilot globally and offers incentives for verified users.
  • Zoom will use World ID Deep Face’s three-step face matching for “Verified Human” badges.

The Tech Oligarch's Republic

by speckx

This article situates readers in a tense moment for the U.S./Israel–Iran ceasefire. It reports that U.S. Marines seized an Iranian-flagged container ship bound for Bandar Abbas and that Iran had not s...

Key Points

  • Ceasefire tension escalates after a U.S. seizure of an Iranian-flagged ship near Bandar Abbas.
  • Talks in Islamabad to extend/expand the ceasefire stall as Iran declines to send a delegation.
  • Palantir publishes a manifesto-like summary of CEO Alexander Karp’s The Technological Republic.

Show HN: Broccoli, one shot coding agent on the cloud

by yzhong94

Broccoli OSS is introduced as a self-hosted AI coding agent that automates software delivery by transforming Linear issues into reviewable GitHub pull requests. The tool is designed for privacy and co...

Key Points

  • Self-hosted AI agent connects Linear issues to GitHub PRs on Google Cloud.
  • Production-grade stack: Cloud Run, Secret Manager, Postgres; customizable prompts.
  • Guided deployment covers GCP, GitHub App, Linear bot, API keys, webhook, and secrets.

Youth Suicides Declined After Creation of National Hotline

by marojejian

A research letter in JAMA reports that the introduction of the 988 Suicide & Crisis Lifeline in 2022 was followed by a notable decline in suicide deaths among U.S. adolescents and young adults. Betwee...

Key Points

  • Youth suicides were 11% below projections after the 988 rollout, with 4,372 fewer deaths estimated.
  • States with higher increases in answered 988 calls saw an 18.2% reduction versus expected suicides.
  • Findings were published in JAMA by Harvard Medical School researchers.

Martin Fowler: Technical, Cognitive, and Intent Debt

by theorchid

Martin Fowler reflects on a half-hour interview with Kent Beck at the first Pragmatic Summit, hosted by Gergely Orosz, where they discussed AI’s growing role in software development alongside lessons ...

Key Points

  • LLMs can incentivize overcomplicated code unless guided by constraints and clear principles.
  • Applying YAGNI can significantly simplify implementations and improve maintainability.
  • TDD-style processes can extend to AI agents by pairing clear instructions with verification.

5x5 Pixel font for tiny screens

by zdw

This article presents a compact 5x5 pixel bitmap font designed for use on tiny screens and embedded systems, rendered on a 6x6 grid. The author argues that 5x5 is the smallest size that preserves legi...

Key Points

  • 5x5 (on a 6x6 grid) is positioned as the smallest legible bitmap font for tiny screens.
  • The font uses 350 bytes and targets constrained microcontrollers and small OLED displays.
  • Fixed-width design eases programming and layout, with predictable 6-pixel character widths.

Startups Brag They Spend More Money on AI Than Human Employees

by SLHamlet

A subset of AI-native startups is publicly celebrating heavy spending on AI compute—sometimes eclipsing payroll—as a deliberate strategy to scale with minimal headcount. This practice, labeled “tokenm...

Key Points

  • Startups are reallocating headcount budgets to AI compute as a scaling strategy.
  • Enterprises are tracking and evaluating AI usage with metrics like Meta’s “Claudenomics” and Salesforce’s “Agentic Work Units.”
  • Examples show minimal-staff companies claiming large-scale outputs and revenue via heavy AI spend.

MythosWatch: Tracking who has access to Anthropic's Mythos AI

by clauderx

The article describes a tracking initiative, informally referred to as MythosWatch, which compiles publicly disclosed information about who can access Anthropic’s restricted Claude Mythos Preview AI m...

Key Points

  • A tracking site is cataloging publicly disclosed access to Anthropic’s restricted Claude Mythos Preview, with sources linked for each entry.
  • Anthropic is investigating reported unauthorized access to Claude Mythos Preview via a third-party vendor environment and says its core systems were not impacted.
  • Media reports indicate a small unauthorized group demonstrated live access to the model, while the NSA has reported access and the Pentagon is legally barred.

Surveillance Pricing: Exploiting Information Asymmetries

by cainxinth

The article traces the evolution of retail pricing from 19th‑century fixed price tags to contemporary data-driven “surveillance pricing.” It credits John Wanamaker’s Grand Depot in Philadelphia with p...

Key Points

  • Surveillance pricing uses personal data to individualize and vary prices.
  • Concrete examples span tickets, ride-hailing, travel booking, tutoring, retail, and grocery delivery.
  • Underlying data and market structures drive these practices beyond any single ban.

Anker made its own chip to bring AI to all its products

by Brajeshwar

Anker introduced “Thus,” a custom neural‑net compute‑in‑memory AI audio chip designed to bring local AI to compact, low‑power devices such as earbuds, mobile accessories, and IoT hardware. The company...

Key Points

  • Anker debuts “Thus,” a compute‑in‑memory AI audio chip for local AI on small devices.
  • First integration is planned for Soundcore flagship earbuds, targeting better noise handling and call clarity.
  • Full details and features will be announced at Anker Day on May 21; leaked models and prices surfaced.

Parallel Agents in Zed

by ajeetdsouza

Zed’s latest release introduces Parallel Agents and a Threads Sidebar, enabling developers to orchestrate multiple AI agent threads concurrently within one window. The open-source update focuses on co...

Key Points

  • Parallel Agents let multiple AI agent threads run at once in a single Zed window.
  • The Threads Sidebar controls per-thread access, groups by project, and offers quick management actions.
  • New default layout docks Threads and Agent Panel left; Project and Git Panels move right, and users can reconfigure.

Coding Models Are Doing Too Much

by pella

The article examines a recurring issue in AI-assisted software development called “Over-Editing,” where large language models modify substantially more code than necessary when fixing defects. While t...

Key Points

  • Over-Editing: models often change more code than needed for a fix.
  • Example shows a one-line bug fix expanded into a full rewrite by a model.
  • Dataset approach: 400 BigCodeBench tasks are programmatically corrupted to evaluate minimal edits.

Who Killed the Florida Orange?

by danso

Florida’s once-dominant orange industry is in freefall, as outlined at the 2026 Florida Citrus Show. USDA projects only 12 million boxes of oranges this season—down from 242 million in 2003—marking th...

Key Points

  • USDA forecasts 12 million boxes of Florida oranges in 2026, a >95% drop from 2003.
  • HLB (citrus greening), spread by the Asian citrus psyllid, is the primary cause of long-term decline.
  • Tariffs, a government shutdown, and a historic freeze intensified recent losses.

Workspace Agents in ChatGPT

by mfiguiere

OpenAI introduced workspace agents in ChatGPT, shared agents that automate complex, long‑running team workflows under organizational permissions and controls. Positioned as an evolution of GPTs, these...

Key Points

  • Shared, cloud‑run workspace agents launch in ChatGPT (research preview).
  • Agents work in ChatGPT and Slack, can be scheduled, and follow org processes with approvals.
  • Powered by Codex, agents run code, use connected apps, and support templates and examples.

OpenAI: Workspace Agents for Business

by tosh

OpenAI’s Workspace Agents for Business are introduced as AI agents designed to automate routine work across existing organizational tools. The agents can gather information and take actions—such as up...

Key Points

  • Agents act across tools to update tickets, edit documents, and send messages without step-by-step guidance.
  • Designed for multiple departments, they automate research, routing, reporting, and lead outreach.
  • Use case: researching inbound leads, scoring prospects, and routing accounts to the right sales team.

We found a stable Firefox identifier linking all your private Tor identities

by danpinto

Researchers uncovered a privacy flaw in Firefox-based browsers where the order of database entries returned by the IndexedDB API produced a stable, deterministic identifier tied to the lifetime of the...

Key Points

  • IndexedDB result ordering leaked a stable process-scoped identifier.
  • Mozilla fixed the issue in Firefox 150 and ESR 140.10.0 (Bug 2024220).
  • The vulnerability defeated Private Browsing and Tor Browser “New Identity” unlinkability.

Website streamed live directly from a model

by sethbannon

Flipbook is an experimental interface that reimagines the web as a sequence of AI-generated images instead of traditional HTML. Each “page” is a full image; clicking anywhere in the scene triggers a n...

Key Points

  • Pages are generated images with clickable exploration; no HTML or overlays.
  • Content is grounded via agentic web search and model knowledge, with accuracy akin to major AI assistants.
  • An experimental live video stream animates pages and transitions; future plans include interactivity and data storage.

Scores decline again for 13-year-old students in reading and mathematics

by u1hcw9nx

NCES reports that 13-year-old students’ performance on NAEP Long-Term Trend (LTT) assessments declined again in the 2022–23 school year. Compared with the 2019–20 administration, average scores fell b...

Key Points

  • Average scores fell 4 points in reading and 9 points in math vs. 2019–20.
  • Declines occurred at all selected percentiles; math drops were largest for lower performers.
  • Findings are based on nationally representative samples with significance-based comparisons.

Anonymous credentials: an illustrated primer (Part 2)

by kkl

This article advances a primer on anonymous credentials by moving from theory to implementation. It recaps how these systems let users authenticate without linking to their identity, detailing the Iss...

Key Points

  • Privacy Pass is a simple, widely deployed anonymous credential protocol standardized in IETF RFCs 9576–9578.
  • Limiting credential reuse is essential to prevent cloning and bot abuse in anonymous systems.
  • Major tech firms (Cloudflare, Apple, Google, Brave, Microsoft) deploy Privacy Pass or its branded variants.

Homegrown – An interactive map of every 2025 FBS college football player

by brockbedard

This BETA web-based tool maps the hometowns of all 2025 FBS college football players, providing an interactive way to explore where athletes come from. Users can tailor the display by selecting a team...

Key Points

  • Maps the hometowns of all 2025 FBS players.
  • Filter by team, conference, or state.
  • '+ Compare' lets you view two selections side-by-side.

New study compares growing corn for energy to solar production. It's no contest

by dotcoma

A PNAS study led by Cornell researchers examines whether U.S. farmland currently dedicated to growing corn for ethanol could be more efficiently used for solar power. Mapping ethanol cropland nationwi...

Key Points

  • 3.2% of ethanol cropland near transmission could match all U.S. corn ethanol energy and lift solar’s share to 13%.
  • Solar PV is ~31× more land-efficient than corn ethanol for energy production.
  • Partial conversion yields environmental benefits and potential 3–4× higher farm lease income.

The great Scouse pasty war

by DamonHD

This piece contrasts the trajectories of two British bakery chains: Liverpool’s Sayers and Newcastle’s Greggs. Sayers, founded in 1912 in Old Swan, long served as a local mainstay for sausage rolls, p...

Key Points

  • Sayers’ decline involved job cuts, a bakery closure in 2008, and partial rebranding to Poundbakery.
  • Greggs grew to a £2bn turnover and strong cultural presence, surpassing McDonald’s for breakfast.
  • Locals cited Sayers’ heated counters and product quality, but also criticized cost-cutting and portion shrinkage.

Ping-pong robot beats top-level human players

by wslh

Sony’s AI research division has developed Ace, an autonomous table tennis robot that competes under International Table Tennis Federation rules and has sometimes beaten top-level human players in Toky...

Key Points

  • Ace reached expert-level play and sometimes defeats top human players under ITTF rules.
  • The robot combines nine cameras, three vision systems, and learning-based control on an eight-joint platform.
  • Results published in Nature include wins over elite and, later, professional players.

The Illuminated Man: an unconventional portrait of JG Ballard

by agronaut

This article reviews The Illuminated Man, Christopher Priest’s biography of JG Ballard, which intertwines the biographer’s terminal illness with a reassessment of Ballard’s life and oeuvre. It outline...

Key Points

  • Priest’s biography integrates his terminal illness into Ballard’s story.
  • Ballard’s life experiences directly shaped his fiction’s themes and imagery.
  • ‘Ballardian’ has entered common usage, with real-world examples cited.

Adobe Is Cooked

by arecsu

This opinion article contends that Adobe’s transition to its Creative Cloud subscription model has undermined innovation and user trust. The author argues Adobe’s size and bureaucracy slow its ability...

Key Points

  • The article criticizes Adobe’s Creative Cloud subscription model and cancellation practices.
  • It claims Adobe was fined $150 million and settled after government action.
  • It argues the subscription model reduced innovation incentives and fostered a walled garden.

Apple fixes bug that cops used to extract deleted chat messages from iPhones

by cdrnsf

Apple has issued a software update for iPhones and iPads to fix a privacy bug that cached notification text for up to a month, enabling the recovery of content from messages users had deleted or set t...

Key Points

  • Apple fixed a bug that retained notification content, exposing deleted/disappearing messages.
  • 404 Media reported the FBI recovered deleted Signal messages due to cached notifications.
  • The fix is available for iPhones and iPads and was backported to older iOS 18 versions.

A Vompeccc Case Study: Spotify as Pure ICR in Emacs

by chiply

This article presents a practical case study of building a Spotify client (“spot”) in Emacs as a pure Incremental Completing Read (ICR) application using the VOMPECCC suite of completion packages. Rat...

Key Points

  • A 493-line shim integrates Consult, Marginalia, and Embark to provide the app’s UI.
  • Async Spotify Web API queries stream results into the minibuffer for interactive search.
  • Consult’s split style supports local narrowing without additional API requests.

Olive CSS: Lisp powered vanilla CSS utility-Class A la Tailwind

by jjba23

Olive CSS is a utility-class, vanilla CSS framework written in Guile Scheme, designed as a Tailwind-inspired, drop-in alternative. The project emphasizes hackability and customization, with the abilit...

Key Points

  • Tailwind-inspired, Guile Scheme–based utility CSS framework with drop-in usage.
  • Production optimization via custom builds, reduced breakpoints/colors, and optional dark mode.
  • Licensed LGPLv3+ for code and FDL 1.3+ for docs; emphasizes hackability and minimal dependencies.

The Neon King of New Orleans

by renameme

The article profiles neon artist and shop owner Nate Sheaffer, who founded Big Sexy Neon in 2020 and recently relocated his workspace to Metairie. Sheaffer is dedicated to preserving New Orleans’ neon...

Key Points

  • Big Sexy Neon restores historic New Orleans signs and creates new neon works.
  • Tujague’s historic sign was saved and now operates in the Southern Food and Beverage Museum.
  • Neon techniques remain largely unchanged, sustainable, and yield long-lasting signage amid LED competition.

Books are not too expensive

by herbertl

Joel J. Miller examines whether books are too expensive by comparing historical hardcover prices to present-day costs after adjusting for inflation. Using examples like To Kill a Mockingbird at $3.95 ...

Key Points

  • Inflation-adjusted 1960s book prices would be ~$43–$54 today, above current hardcover prices.
  • CPI data shows “recreational books” have slightly declined in real price since 1997.
  • Other consumer categories have outpaced inflation, while books have remained flat.

First per-image PCA decomposition of Kodak suite reveals deliberate curation

by PearsonZero

A newly published repository delivers a comprehensive, per-image Principal Component Analysis (PCA) for all 24 images in the Kodak Lossless True Color Image Suite. Framed as baselines approximating th...

Key Points

  • First per-image PCA statistics for the 24-image Kodak suite
  • Systematic variability found: condition numbers 7.55–1739.16; blue independence 2.3%–52.0%
  • Baselines intended for benchmarking compression, redistribution, and PSNR/SSIM stability

Show HN: Ghost Pepper Meet local meeting transcription and diarization

by MattHart88

Ghost Pepper is a free, MIT-licensed macOS app that provides on-device voice dictation and meeting transcription, emphasizing privacy by keeping all data local. Designed for Apple Silicon Macs running...

Key Points

  • All processing is on-device; no cloud, accounts, or telemetry.
  • Multiple local ASR and LLM model options with speed/quality trade-offs.
  • macOS 14+ on Apple Silicon is required (some models need macOS 15+).

How to Stop a Data Center in Your Backyard

by cdrnsf

Residents in Monterey Park, California, mobilized to oppose a proposed 250,000-square-foot data center slated for a site near homes. SGV Progressive Action, drawing on its existing volunteer network a...

Key Points

  • Public records revealed reliance on the developer’s impact assessment, not a fuller state review.
  • Outreach broadened beyond limited notifications, yielding large turnout and a 150-person teach-in.
  • The developer withdrew the application after sustained community organizing.

Approximating Hyperbolic Tangent

by jtomschroeder

This article reviews practical techniques for approximating the hyperbolic tangent (tanh) function to meet stringent performance demands in neural networks and real-time audio processing. It explains ...

Key Points

  • Rust examples show Taylor and [7/6] Padé tanh approximations with defined saturation ranges.
  • Spline-based piecewise cubics (three segments over [0, 18]) prioritize speed for neural network use.
  • IEEE‑754-based tricks are highlighted as another promising avenue for fast approximations.

Bring your own Agent to MS Teams

by umangsehgal93

This article presents a practical method for integrating existing conversational agents with Microsoft Teams using the Teams TypeScript SDK. By leveraging the SDK’s HTTP server adapter, developers can...

Key Points

  • Use the Teams TypeScript SDK to inject the /api/messages endpoint into an existing Express server.
  • Co-host Slack (Bolt) and Teams bots on the same process, sharing common agent logic.
  • Forward Teams messages to LangChain or Azure AI Foundry agents and relay their responses.

Effectful Recursion Schemes

by marvinborner

This article by Marvin Borner demonstrates how to implement recursion schemes—specifically catamorphisms—using the Effekt programming language’s effects and handlers. Instead of relying on the traditi...

Key Points

  • Effekt replaces functor-based recursion schemes with a refunctionalized effect interface (TermF[T,R]).
  • A generic cata function with effect signature enables catamorphisms over terms.
  • Handlers with resume implement folds by mapping sym, lam, and app to results.

Rip language. Compiles to ES2022. Built-in reactivity

by perfunctory

Rip is introduced as a modern, CoffeeScript-inspired language that compiles to ES2022 JavaScript, delivering native classes, optional chaining, nullish coalescing, and modules in its output. The proje...

Key Points

  • Compiles to ES2022 with built-in reactivity operators (:=, ~=, ~>) and new concise operators like ! and |>.
  • Self-hosting, zero-dependency compiler (including parser generator) with installation and tooling via bun.
  • Optional types compile to .js with .d.ts emission for IDE support through the TypeScript Language Server.

The handmade beauty of Machine Age data visualizations

by benbreen

This article examines William James’s visual creativity through archival research at Harvard’s Houghton Library and situates it within the 19th‑century emergence of modern data visualization. It docum...

Key Points

  • Archival evidence shows William James’s sustained visual practice and art training.
  • James’s era bridged early pioneers (Playfair, Nightingale) and modern data visualization.
  • James’s network included Galton and Du Bois; he is credited with an early neural network schematic.

Tempest vs. Tempest: The Making and Remaking of Atari's Iconic Video Game

by mwenge

“TEMPEST vs TEMPEST” is a free, book-length technical exploration of two landmark arcade titles: Dave Theurer’s Tempest (1981) and Jeff Minter’s Tempest 2000 (1994). The book aims to reveal how numero...

Key Points

  • In-depth, code-level analysis of Tempest (1981) and Tempest 2000 (1994).
  • Focus on 6502 and Motorola 68K assembly implementations.
  • Free downloads (standard and high-res) with GitHub repository for more details.

Verus is a tool for verifying the correctness of code written in Rust

by fanf2

Verus is a formal verification tool designed to prove the functional correctness of low-level Rust systems code through static analysis and theorem proving, avoiding runtime checks. It combines a math...

Key Points

  • Verus statically verifies Rust systems code correctness without runtime checks.
  • It generates SMT-solvable verification conditions and uses linear type checking for memory/aliasing.
  • Specs and proofs are written in Rust with macro extensions and mathematical types.

OpenAI's response to the Axios developer tool compromise

by shpat

OpenAI detailed a software supply chain incident in which a malicious version of the Axios library (v1.14.1) executed within its GitHub Actions workflow for macOS app signing on March 31, 2026. The wo...

Key Points

  • Malicious Axios package entered OpenAI’s macOS signing workflow; certificate likely not exfiltrated.
  • OpenAI rotating and revoking macOS code-signing certificate; users must update by May 8, 2026.
  • Root cause: GitHub Actions misconfiguration (floating tag, no minimumReleaseAge), now fixed.

How the Heck does Shazam work?

by datadrivenangel

This explainer describes how apps like Shazam identify songs from brief, noisy recordings. Rather than analyzing melody or lyrics, the phone captures audio via its microphone and digitizes it into a w...

Key Points

  • Waveforms are converted to spectrograms via FFT for analysis.
  • Only the loudest spectrogram peaks are kept as a sparse fingerprint.
  • Peak-based fingerprints enable fast, noise-robust song matching.

Borrow-checking without type-checking

by jamii

The article introduces a demonstration language, informally called “zest,” that aims to combine dynamic typing with robust mutable value semantics, single ownership, and a limited borrowing model. Unl...

Key Points

  • Borrow rules are enforced dynamically with stack-based, thread-local reference counts used only in dynamic frames.
  • Zest uses explicit annotations to switch between dynamically typed (interpreted) and statically typed (compiled) modes.
  • The language supports value semantics, references via a ‘box’ function, and aims for low overhead with clear error reporting.

Flow Map Learning via Nongradient Vector Flow [pdf]

by E-Reverance

This paper introduces SGFlow (StopGrad Flow), a method for accelerating sampling in diffusion and flow-based generative models by directly learning the ODE flow maps governing probability flow traject...

Key Points

  • SGFlow learns ODE flow maps without invertibility or nested backpropagation.
  • The true flow map is a unique stationary point of SGFlow’s training dynamics.
  • On CIFAR, SGFlow shows a favorable FID–step trade-off versus strong baselines.

A True Life Hack: What Physical 'Life Force' Turns Biology's Wheels?

by Prof_Sigmund

This article explores how evolution solved the challenge of microbial movement through the bacterial flagellar motor—an electric, self-assembling rotary machine that spins at hundreds of revolutions p...

Key Points

  • Studies since 2020 resolved structures of key flagellar motor components, capped by findings in March 2026.
  • The flagellar motor’s CCW vs. CW rotation governs running and tumbling in bacterial movement.
  • The proton motive force powers the motor and is emphasized as a fundamental cellular energy source.

It's time to reclaim the word "Palantir" for JRR Tolkien

by IdahoSpring

This piece is the opening talk by Juan Sebastián Pinto at the launch of Medact’s briefing on Palantir’s role in the UK’s National Health Service (NHS), delivered alongside Amnesty International, the G...

Key Points

  • Talk launches Medact’s briefing on Palantir in the NHS, with major civil society groups involved.
  • Author urges reclaiming AI language so non-engineers can scrutinize cloud surveillance platforms.
  • Cites alleged real-world harms and defense use of AI tools to contextualize NHS concerns.

I am building a cloud

by bumbledraven

A Tailscale co‑founder introduces exe.dev, a new cloud venture announced alongside fundraising, and explains the motivation for building it. The post argues that today’s cloud platforms constrain user...

Key Points

  • exe.dev is introduced with a fundraising announcement.
  • Current cloud VM and PaaS abstractions are criticized as restrictive and provider‑specific.
  • The author advocates resource‑centric control (CPU, memory, disk) and questions remote storage defaults like S3.

A Boy That Cried Mythos: Verification Is Collapsing Trust in Anthropic

by taejavu

This article critically evaluates Anthropic’s claims about the Claude Mythos Preview’s cybersecurity capabilities. The author reviews the 244-page system card and asserts that only about seven pages a...

Key Points

  • System card devotes limited space to cybersecurity claims and lacks key vulnerability details, per the author.
  • Demonstration involved already-patched bugs in a de-sandboxed environment, questioning real-world impact.
  • “Thousands of zero-days” is cited in marketing, not substantiated in the system card, according to the article.

The Onion to Take over InfoWars

by lxm

The Onion has revived its effort to control Infowars by proposing a licensing agreement instead of a purchase. A court filing by Gregory Milligan, the court-appointed manager of Infowars, asks Judge M...

Key Points

  • The Onion seeks a court-approved license to operate Infowars.com for $81,000 per month.
  • Judge Maya Guerra Gamble must approve; Alex Jones could appeal any decision.
  • The plan follows a previously blocked sale amid Sandy Hook-related defamation and bankruptcy cases.
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.