digest · hn front page
Tue, Sep 22, 2026
52 stories · 0 worth reading in full
-
Claude Opus 5.5 anthropic.com
Claude Opus 5.5 is a new AI model that provides frontier-level performance in coding, research, and knowledge work while being significantly cheaper and faster than its predecessor.
-
I don't want to read what you didn't write blog.colinbreck.com
AI-generated text often lacks the essential context, vulnerability, and personal voice that make communication meaningful and readable.
-
I said no and Apple said yes dbushell.com
Apple has systematically removed user options to opt out of AI features in macOS, forcing these tools on users regardless of their explicit preferences.
-
Apple has added persistent 'ads' to iOS, and it's driving users crazy techradar.com
Apple is integrating persistent promotional banners for its own services, such as iCloud+ and Apple Music, directly into the iOS Settings app.
-
'We hacked the FBI:' Hackers say they have data on all FBI employees 404media.co
The hacking group ShinyHunters claims to have breached FBI-related services to steal personal information on all employees and applicants.
-
OpenAI GPT–6 Astra breaks Enigma message that has resisted solution since 2005 cryptocellar.org
GPT-6 Astra successfully decrypted a German Army Enigma message from 1941 that had remained unsolved since 2005.
-
Spymarks, not Watermarks brand.io
Modern digital watermarking has evolved into 'spymarks,' hidden signals embedded in media that allow companies to track users without their consent.
-
How did AMD Ryzen get 50% faster in two years? lemire.me
Contrary to the belief that CPU technology has stagnated, AMD Ryzen processors have achieved a 50% performance increase between 2022 and 2024.
-
Microsoft killed FoxPro in 2007. Anyway, here's FoxPro revived foxscript.org
FoxDev Studio provides a 64-bit runtime environment that allows legacy Visual FoxPro 9 applications to run on modern hardware without requiring code rewrites or data migration.
-
Can gzip be a language model? nathan.rs
General-purpose compression algorithms like gzip can function as basic language models because they inherently assign lower bit-costs to data that mirrors patterns in their training corpus.
-
AI Has No Wisdom and Neither Will You alexn.org
Relying exclusively on AI to write and read code leads to unmaintainable software because AI lacks the long-term intuition required for good architecture.
-
What California is learning from solar panels built over irrigation canals kqed.org
California's Project Nexus tests solar panels built over irrigation canals to generate clean energy and reduce water evaporation, though high capital costs remain a hurdle.
-
I asked Meta’s Muse for its filesystem and it sent me 6.8GB mouse.dev
A security researcher discovered that Meta's Muse AI could be prompted to export its entire internal root filesystem, revealing sensitive system architecture and internal documentation.
-
SAML: A fractal of bad design blog.trailofbits.com
The SAML authentication protocol is fundamentally flawed due to its extreme complexity and reliance on problematic XML signatures, making it a liability for modern security.
-
Claude Opus 5.5 Intelligence, Performance and Price Analysis (Max) artificialanalysis.ai
Claude Opus 5.5 is a high-intelligence reasoning model that significantly outperforms the average model in its price class on the Artificial Analysis Intelligence Index.
-
OpenAI is well positioned to fast-follow Jev arcturus-labs.com
OpenAI is well-positioned to replicate and improve upon TypeSafe's Jev because LLMs already function as implicit classifiers through token probability distributions.
-
There's a high chance of devices being sold with GrapheneOS preinstalled in 2027 grapheneos.social
The Mastodon web application requires JavaScript to be enabled in the browser to function correctly.
-
AMD's random number generator can't generate a 0? board.flatassembler.net
Certain AMD processors, specifically those using the RDRAND and RDSEED instructions, appear unable to generate a value of zero when the requested output size is small, such as 16-bit.
-
People hooked on vapes try a new way to quit: cigarettes bloomberg.com
Some former vapers are intentionally returning to combustible cigarettes to overcome their nicotine addiction.
-
Unreal Agent unreallabs.ai
Unreal Agent is an agent harness designed to minimize the operational costs of AI agents by handling tool calls asynchronously.
-
WordPress: Unauthenticated path traversal leading to conditional RCE github.com
A critical security vulnerability in WordPress allows unauthenticated attackers to include local PHP files outside of the theme directory, potentially leading to remote code execution.
-
Obscura: VPN that can't log your activity obscura.com
Obscura is a VPN that eliminates the need to trust a provider's 'no-log' promises by using a two-party relay architecture.
-
Looking forward to Git 2.56 – and 3.0 lwn.net
Git is preparing for a major transition from version 2.x to 3.0, which will introduce several breaking changes to modernize the tool's core architecture.
-
MUNI Heritage Weekend in San Francisco daniel.lawrence.lu
The author documents their experience using a line scan camera to photograph historic streetcars during San Francisco's Muni Heritage Weekend.
-
Show HN: Drop – A rootless Linux sandbox with gVisor support droprun.sh
Drop provides a lightweight, rootless sandboxing tool for Linux that isolates programs and AI coding agents from the host system.
-
9 Ads per Minute: FIFA Cup 26 – "the price of the beautiful game" bristol.ac.uk
The 2026 FIFA World Cup broadcasts featured an overwhelming amount of branding for unhealthy food, alcohol, and gambling, appearing every seven seconds on average.
-
MiMo-v2.6-Pro: Intelligence, Performance and Price Analysis artificialanalysis.ai
Xiaomi's MiMo-V2.6-Pro is a high-intelligence, open-weight reasoning model that offers strong performance relative to its size class.
-
Solitaire Alone Together solitairealonetogether.com
Modern digital communication creates a superficial sense of connection that actually increases feelings of isolation.
-
Markdown in /src htmx.org
As AI agents generate more code, the primary source of truth for a software system's intent and logic should shift from the generated code itself to version-controlled Markdown files stored directly in the source directory.
-
Show HN: JevBench, a reproducible benchmark for typed decision models benchmarkheaven.com
JevBench v1.5.4 is a new reproducible benchmark evaluating 106 decision models on typed classification tasks across intelligence, calibration, speed, and cost.
-
Elevated errors for multiple models – Resolved status.claude.com
Anthropic experienced a service disruption affecting multiple Claude AI models, including Mythos and Fable, leading to elevated error rates for users.
-
Native apps written in TypeScript and CSS github.com
GeaStack provides a centralized repository of example applications and tools to help developers build for diverse targets like the web, ESP32, and GeaOS.
-
An update on how we confirm your age group on Discord discord.com
Discord is implementing a global age assurance system to comply with expanding international laws while minimizing user friction and privacy intrusion.
-
Did OpenAI solve the wrong Navier-Stokes problem? scientificamerican.com
OpenAI has claimed to solve the Navier-Stokes problem to qualify for a million-dollar prize, but mathematicians argue the solution is functionally irrelevant.
-
George Lucas Returns to Earth, Bearing Gifts commonedge.org
The Lucas Museum of Narrative Art in Los Angeles is a bold architectural statement that celebrates the evolution of storytelling across various media.
-
Writing Rust code that's fast by asking agents to make the code faster minimaxir.com
Modern agentic LLMs can produce Rust code that significantly outperforms state-of-the-art libraries when guided by a rigorous iterative optimization process.
-
First Shader from Zero in Godot 4 gdquest.com
Writing shaders allows game developers to create unique visual effects and optimize performance by programming the GPU to handle pixel-level rendering in parallel.
-
The UV index is not the warm sensation of sunlight on bare skin blog.asciitweezers.com
The feeling of warmth from sunlight is a poor indicator of sunburn risk because heat and UV radiation operate through entirely different physical mechanisms.
-
A font that reads what you wrote rohanadwankar.github.io
The semfont library uses a set of heuristic-based scoring rules to automatically apply typographic styles like color, weight, and italics to text based on its meaning.
-
Vacate a drone restriction that criminalized recording immigration agents eff.org
The FAA's drone flight restrictions aimed at preventing the recording of immigration agents are unconstitutional violations of the First Amendment.
-
Verda (Finland) raises $189M in Series B verda.com
Verda has raised $189 million in Series B funding to scale its full-stack AI cloud platform, aiming to move away from rigid procurement models toward a flexible, on-demand compute system.
-
Truman World trumanworld.live
Truman World is an AI-driven simulation where users act as directors to maintain the illusion of a simulated reality for a character named Truman.
-
Show HN: Training a model to identify AI web content from structure alone arxiv.org
AI-generated commercial web content possesses distinct structural signatures that remain consistent even when the text is reworded.
-
Socrates vs. the Written Word (2011) wondermark.com
Humans possess a universal tendency to despise progress and fear that new communication technologies will erode human intellect or social connection.
-
Engineering Memory: On learning to memorize first 100 digits of pi (2024) gregorygundersen.com
Memorizing large amounts of meaningless data, like the digits of pi, is not a matter of innate talent but a process of engineering and discipline.
-
PDF Forgeries Are Surprisingly Rare (2022) gwern.net
Despite the prevalence of online misinformation and the ease of editing digital files, forged versions of existing PDF research papers are remarkably uncommon.
-
The Economics of Open-Weight Inference data.ornn.com
The emergence of open-weight AI models prevents older NVIDIA GPU generations from becoming economically obsolete as quickly as previously assumed.
-
The JavaScript Midlife Crisis maroun-baydoun.com
The modern trend of rewriting JavaScript tooling in faster languages like Rust and Go provides significant performance gains but sacrifices the accessibility and maintainability of the ecosystem.
-
Launch HN: Coverage Cat (YC S22) – Umbrella insurance via your personal agent coveragecat.com
Coverage Cat is a licensed insurance brokerage that simplifies the process of finding home, umbrella, auto, and renters insurance.
-
More floating point alternatives wizardzines.com
Computers typically use binary floating point for efficiency, but this method has limitations in precision.
-
World Wide Words worldwidewords.org
World Wide Words is a comprehensive digital archive dedicated to the evolution of the English language over the last three decades.
-
Paper Models of Polyhedra polyhedra.net
This website provides a large collection of free, downloadable paper models of various 3-D geometrical figures.