digest · hn front page
Tue, Sep 8, 2026
51 stories · 1 worth reading in full
-
Navier-Stokes – Tristan Buckmaster [pdf] cims.nyu.edu
Mathematician Tristan Buckmaster alleges that OpenAI rushed an internal AI model to duplicate his unreleased Navier-Stokes work after a leak, then pressured him to strip his co-author of credit.
-
On the Navier–Stokes Millennium Prize Problem openai.com
OpenAI has used a highly advanced internal AI system to resolve the Navier-Stokes existence and smoothness problem, proving that fluid dynamics can develop a singularity in finite time.
-
Mistral raises €3B mistral.ai
Mistral AI has raised €3 billion in a Series D round to advance its goal of providing high-performance, open-weight AI models that ensure organizational sovereignty.
-
Jellyfin 12.0 jellyfin.org
Jellyfin 12.0 is a major stable release that focuses on database performance, expanded media support for books and comics, and a modernized user interface.
-
AlphaGenome Atlas: a high-resolution map of human DNA blog.google
Google DeepMind has created AlphaGenome Atlas, a massive database predicting the effects of all 9 billion possible single nucleotide variants in the human genome.
-
We built our house for LAN parties (2024) lanparty.house
The authors designed and built a custom home in Austin, Texas, specifically optimized for hosting large-scale local area network (LAN) parties.
-
LG TVs caught spying even when offline or on standby theverge.com
LG smart TVs are aggressively collecting personal data, including audio recordings and network scans, even when the devices are offline or in standby mode.
-
There's a new "Google Jail" for independent wikis weirdgloop.org
Google implemented a change in early 2024 that effectively penalizes brand-new root domains by preventing their internal pages from appearing in search results, creating a 'Google Jail' for new websites.
-
I-have-ADHD: A skill to stop coding agents from burying the answer github.com
AI coding assistants often bury critical information under polite filler and conversational fluff, which can be distracting and inefficient for users with ADHD.
-
I've factored the RSA keys of a Certificate Authority from the 90s mcpherrin.ca
Early web security standards were often lax, allowing the use of 512-bit RSA keys that are now trivial to factor on modern consumer hardware.
-
Tao: Open math problems being non-renewably mined by AI mathstodon.xyz
The Mastodon web application requires JavaScript to function.
-
How to build a printer nishantjosh.dev
The author transforms a programmable e-ink display into a functional network printer by implementing the Internet Printing Protocol (IPP) and optimizing memory usage.
-
We Must Return to the Office to Use AI in Person mcsweeneys.net
The author uses a satirical own-account of working for a dystopian corporate entity called Mondo Mayo to mock the irrationality of mandatory return-to-office mandates.
-
Among European Companies That Use a CDN, Nearly 9 in 10 Use Cloudflare ciphercue.com
Nearly 90% of European companies that use a content delivery network are dependent on Cloudflare, creating a significant single point of failure for the region's digital infrastructure.
-
ChatGPT Images 2.5 openai.com
OpenAI has launched ChatGPT Images 2.5, a new image generation model designed to provide sharper details, faster processing, and more precise editing capabilities.
-
DHS 'Predictive Policing' Unit Is Analyzing Americans' Financial Habits 404media.co
The Department of Homeland Security's Border Patrol operates secret predictive policing units that monitor the financial habits of Americans to identify targets for police stops.
-
TALA Is Open-Source d2lang.com
TALA is a new open-source autolayout algorithm specifically designed to create software architecture diagrams that look like natural whiteboard drawings.
-
We have a year to fix security everywhere jyn.dev
The release of high-capability, open-weight LLMs like GLM 5.3-flash allows almost anyone to run powerful hacking tools locally without safety restrictions.
-
Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapses quesma.com
Quantizing the Qwen3.8 27B model down to 4-bit (Q4_K_M) maintains performance nearly identical to the full BF16 version across science, instruction-following, and coding benchmarks.
-
Kimi K3 (2.8T) at 1 token/s on a MacBook Pro, streamed from four SSDs github.com
Deltafin allows the full, unpruned 2.8-trillion-parameter Kimi K3 model to run on consumer hardware by streaming expert weights from SSDs rather than relying on VRAM.
-
Getting your hands dirty is good for you bbc.com
Direct physical contact with soil, plants, and natural materials increases the diversity of microbes on human skin, which strengthens the immune system.
-
Show HN: Copperhead – Cursor for circuit boards copperhead.sh
Copperhead is an open-source AI engineering platform designed to automate the creation, verification, and documentation of printed circuit boards.
-
The two Christian saints who are the Buddha signoregalilei.com
The story of Saints Barlaam and Josaphat is actually a Christian adaptation of the life of Gautama Buddha, illustrating how religious narratives migrate and morph across cultures.
-
Mercury 2.5 inceptionlabs.ai
Mercury 2.5 is a high-performance diffusion language model designed to balance intelligence and extreme speed for latency-sensitive applications.
-
Antiquated HTML Snippets and Artefacts vale.rocks
The web's constant evolution leads to the creation of non-standard HTML snippets designed to combat browser competition, vendor extensions, and platform-specific limitations.
-
The Helicopter with Radioactive Blades hackaday.com
The CH-53 Sea Stallion helicopter utilizes a clever mechanical system called IBIS to detect structural cracks in its rotor blades during flight.
-
Large language models develop novel social biases through adaptive exploration openreview.net
The page serves as a security gate requiring users to complete a verification check before accessing OpenReview.
-
How well do agents use test/verification techniques? danluu.com
AI coding agents currently struggle to use advanced testing and verification techniques effectively, often applying them superficially or incorrectly regardless of the specific tool or library mentioned.
-
My Feed, My Way pm.gov.au
The Australian government is introducing draft legislation to hold digital service providers accountable for user safety through a mandatory 'Duty of Care'.
-
Show HN: Stuxnet – A reconstructed source code of the infamous cyber-weapon github.com
This repository provides a reverse-engineered reconstruction of the Stuxnet worm to facilitate academic study and defensive security research.
-
Show HN: LLM Attention Visualization ishamf.dev
Transformer-based large language models use an attention mechanism to selectively draw information from previous tokens to generate the next token.
-
FreeBSD 14.5-Release freebsd.org
FreeBSD 14.5-RELEASE is a maintenance-focused update to the stable/14 branch designed to improve system reliability and compatibility.
-
Replacing a Rust Enum with a 64-Bit Word Made My Interpreter 17% Faster pointersgonewild.com
Replacing a 16-byte Rust tagged enum with a 64-bit tagged word representation significantly improves interpreter performance by reducing memory traffic and allowing values to fit into CPU registers.
-
John Margolies' photographs of roadside America publicdomainreview.org
John Margolies documented thousands of unique, homemade buildings and signs along American roads to preserve a vanishing form of vernacular architecture.
-
I tested 10 model/harness combinations on the same Three.js task alvins82.github.io
The author tests various AI models and harness configurations to determine which combination produces the best results for creating a complex Three.js sci-fi hangar scene.
-
Multi-Agents LLM Financial Trading Framework github.com
TradingAgents is an open-source framework that simulates a professional trading firm by using specialized LLM-powered agents to collaboratively analyze financial markets.
-
ZX Spectrum: Experimenting with 1-Bit Sound bumbershootsoft.wordpress.com
The author explores various methods for generating complex audio, such as chords and PCM playback, using the ZX Spectrum's limited 1-bit beeper.
-
Connecting the machines herdr.dev
Herdr 0.9 introduces multi-machine support to solve the friction of managing AI agents across different servers and local hardware.
-
AlphaGenome Atlas predictive map of every DNA letter change in the human genome deepmind.google
Google DeepMind has created AlphaGenome Atlas, a massive dataset providing AI-driven predictions for the molecular effects of all 9 billion possible single-letter DNA changes in the human genome.
-
Arm Mali G2-Ultra NX GPU: desktop-class mobile gameplay with AI-native graphics newsroom.arm.com
The Arm Mali G2-Ultra NX GPU integrates neural accelerators directly into its graphics pipeline to achieve desktop-class visual fidelity on mobile devices.
-
C*: Unifying Programming and Verification in C (2025) arxiv.org
C* is a proof-integrated language that merges C programming and formal verification into a single environment.
-
Implementation of GCC's Nested Functions (vs. C++ Lambdas) uecker.codeberg.page
GCC implements nested functions by gathering all parent variables accessed by the child function into a single synthetic structure and passing a pointer to this structure as a hidden argument.
-
End-to-end infrastructure for training and inferencing open weight models docs.appliedcompute.com
Applied Compute Agent Cloud (AC2) provides a comprehensive researcher cockpit for the end-to-end development, inference, and continual learning of open weight models.
-
The VMs Powering Mobile Agents (Instinct, Claude Code) rohanadwankar.github.io
AI agents are shifting from local execution to cloud-based virtual machines (VMs) to enable mobile access and better persistence.
-
Understanding Computer Memory Architecture and SSD Internals codingpirate.com
Modern computing relies on a sophisticated memory hierarchy that balances speed, capacity, and cost by layering different technologies from fast CPU caches to dense non-volatile storage.
-
Animation in Bevy: The Big Picture glocq.com
Animating 3D models in Bevy requires a specific mental model of how components and hierarchies are used to control movement.
-
Extracting Steering Vectors from J space darshanmakwana412.github.io
The author investigates whether steering vectors for large language models can be derived by inverting the Jacobian lens, which maps internal activations to vocabulary tokens.
-
Picolibrary: A Small Press novalis.org
Republishing a rare book in the modern era is a manageable task for a single individual if they leverage LLMs for contract drafting, OCR cleanup, and formatting.
-
Tracing np.add, all the way down blog.veitheller.de
The simple call to np.add(a, b) triggers a sophisticated sequence of operations that transitions from high-level Python to highly optimized C and SIMD kernels.
-
Extinct Tasmanian tiger's 'snap' unlike any living mammal's bite cnn.com
Recent analysis of Tasmanian tiger skulls reveals that these extinct marsupials had specialized feeding adaptations for small prey, contradicting the historical belief that they were wolf-like predators of livestock.
-
Colorlight 5A-75B: A dive into a popular low-cost ECP5 (FPGA) development board blog.yosyshq.com
The Colorlight 5A-75B, originally designed as an LED receiver card, serves as an affordable and powerful FPGA development board for hardware hackers.