Gemini 4 Neon

https://storage.googleapis.com/gweb-uniblog-publish-prod/images/gemini_4_neon_table.width-1200.format-webp.webp
Gemini 4 Neon is a new AI model that offers frontier-level intelligence and fast task completion across complex software engineering, knowledge work, and science workflows. It ranks #3 overall with a score of 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level.

AI Will Not Make You Rich (2025)

https://colossus.com/wp-content/uploads/2025/08/AI-Will-Not-Make-You-Rich-Artwork.webp
Generative AI may not create new wealth for investors, as it is likely to be captured by dominant ICT companies. Instead, opportunities will arise in sectors that benefit from AI's efficiency gains and lower costs.

Unauthenticated remote code execution in Nginx

https://www.datocms-assets.com/75231/1794936719-img_7875.png?fm=webp
Researchers from Wiz Research and other teams discovered two critical remote code execution vulnerabilities in NGINX, a widely used open-source web server. The vulnerabilities, CVE-2026-123633 and CVE-2026-123007, can allow attackers to execute arbitrary code on the server without authentication.

Gemini 4 Neon looks like an Opus 5.5 competitor

https://micjm.com/images/gemini-4-neon-looks-like-a-top-model/calibration.png
Gemini 4 Neon's performance is projected to be equivalent to Opus 5.5, with a win rate of around 48.3% against Opus in Agent Duel leaderboard games. This surpasses Gemini 4 Carbon's Elo rating of 1,066 when assuming the same performance as Carbon and benchmark scores of 127 tokens per second.

Multiple vulnerabilities in Regexp and String case conversion

https://www.ruby-lang.org/images/og-image.png
Multiple vulnerabilities have been discovered in Ruby's regular expression engine and string case conversion, allowing remote code execution. Users are strongly advised to upgrade to Ruby 4.0.10, 3.4.13, or 3.3.14 to fix the issues immediately as they are no longer supported by newer versions of the software

Seven new models at scalable price-performance

https://microsoft.ai/wp-content/uploads/2026/11/Custom-Models-no-text.jpg
MAI is releasing a family of seven new world-class models at exceptional price-performance, including Cyber 2 Flash, Code 2, Image 2.7 and more. These models offer high-quality performance at significantly lower costs than competitors like GPT-6 and Opus.

Anthropic prices initial public offering

https://assets.claude.com/6d4a0d28992ade92d6fa63646fd9c9d318245c6c.jpg
Anthropic priced its IPO at $83.00 per share, offering 980M shares to the public. The offering is expected to close on November 20, 2026, and will trade under the symbol "ANT" on Nasdaq Global Select Market.

LLM assisted reverse engineering

https://www.slowchronicle.com/og-default.png
The user has been reverse engineering software and hardware for 2 decades, using AI tools like Deepseek, Codex, and Luna to speed up the process. They describe their workflow as finding binary type and platform information, documenting findings in MD files, analyzing basic structure of the binary.

AI agents ask better questions than KPMG accountants in audit trial

https://europeanbusinessmagazine.com/wp-content/uploads/2026/06/ai5555-scaled.jpg
KPMG's AI experiment shows agents can investigate financial issues more deeply and ask better questions than human audit teams, raising concerns about the future of auditing. The technology has potential to fundamentally change how complex financial information is examined, but its limitations and challenges must be addressed.

Inside Deutsche Bank

https://www.inside-db.com/content/images/2026/11/BAD-BANK-1.png
Deutsche Bank leaked documents reveal the bank's involvement in various scandals, including money laundering and arms sales to conflict-ridden countries. The bank facilitated transactions worth €172 million linked to Jho Low and supplied weapons used against civilians in Palestine.

DNS resolvers are the wrong place to fight piracy

https://techstackups.com/img/articles/quad9-leaves-france/cover.jpg
Quad9, a Swiss non-profit DNS resolver, will stop answering queries from France due to court fines for refusing to block pirate sports-streaming domains. Quad9's departure is seen as a canary in the coal mine by Cloudflare and others.

Fable: Written, painted and directed by Claude Fable 5.2

https://claudia.gallery/films/fable/stills/01.s.webp
Claudia reads bedtime stories to the AI, including classic tales like Snow White and Pinocchio, with a request to keep the book. The AI creates a short film based on these stories, using 61 stills and $7.93 at list for production costs.

DaisyRose-27B

https://www.radixark.com/_next/image?url=%2Fimages%2Fdaisyrose-27b%2Fhero.png&w=3840&q=75&dpl=dpl_2WZt4617DGLJ3EDD7HwJmGV2VG4H
RadixArk released DaisyRose-27B, a model built on Qwen3.8-27B with reinforcement learning using Miles, achieving significant gains on various benchmarks. The model's success is attributed to high-quality training data, efficient infrastructure and robust sandboxes.

Previous-Token Prediction Based LLM Inversion for Prompt Reconstruction

https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png
This paper introduces a black-box method for inverting LLMs by training an inverse model on synthetic data generated from the target model. The approach uses previous-token prediction to create a generative link, enabling faithful prompt reconstruction and outperforming prior methods.

FailTrace – Reproduce a failed GitHub Actions run locally from its URL

https://raw.githubusercontent.com/Ri4dH/failtrace/main/docs/demo.gif
FailTrace identifies the exact step and commit causing a GitHub Actions failure. It provides a reproduction plan and runs it in Docker to diagnose issues locally.

Dumping an AES key from a MyQ garage opener to stop rebooting when offline

https://kusha.me/imgs/145d4d43b2f05e68a7ca.png
A myQ device on an IoT VLAN with no WAN access constantly attempts to connect to Chamberlain's cloud endpoints, causing reboots and unresponsiveness in HomeKit. To remediate this, the TLS-PSK AES-128 key must be dumped and a custom MQTT server setup using the device's CPU debug pads.

We need smart people who give a shit

The world relies on complex systems held together by smart people who genuinely care, but a shift to generative AI threatens this balance. As AI replaces human expertise, the number of smart individuals giving a shit decreases, leading to more system failures and potentially catastrophic consequences for capitalism.

Advancing chip design with OpenAI models

https://images.ctfassets.net/kftzwdyauwt9/1FaKIo74WEGULSjb4ooTrV/718f3bc75e7b2ddd276d43a0fc97760a/academic-researchers-art-card.png?w=3840&q=90&fm=webp
OpenAI's models are improving chip design capabilities, achieving significant gains on third-party hardware with AMD. The models have demonstrated expertise-level design improvements in accelerating processor designs and optimizing production chips.

The Cambridge Dictionary has named ‘human’ as the Word of the Year 2026

https://www.cam.ac.uk/sites/default/files/styles/rectangle-349x183/public/2026-11/supplychainWEB.jpg.webp?h=0e196f1f&itok=L4S_aCzb
Global debates on AI's impact on humanity have intensified, with many defending human qualities against machine replication. The Cambridge Dictionary has named "human" its Word of the Year, sparking discussions about what it means to be human in an age of AI.

Genie 4 is now available

https://storage.googleapis.com/gweb-uniblog-publish-prod/images/hero_ZCQxfQQ.2e16d0ba.fill-300x300.format-webp.webp
Google is releasing an updated Project Genie experience for eligible Google AI Ultra subscribers, offering more control and social features. The update includes new world models, a community gallery, and real-time native audio that adapts to the scene unfolding.

Cuttlefish on Mac: Android's Virtual Device Framework on Apple Silicon

https://engineering.fb.com/wp-content/uploads/2026/11/Cuttlefish-on-Mac-lead.png
Meta's Reality Labs ported Cuttlefish, an Android emulator framework, to macOS and Apple silicon. This unified platform reduces maintenance requirements and enhances fidelity for developers working on the AOSP project.

Show HN: ExitPath – Open-source ways out of Slack, Notion and Microsoft 365

https://exitpath-ruddy.vercel.app/og.png
Find, compare and switch to open-source alternatives for the tools you depend on. Honest ratings, migration difficulty scores and step-by-step checklists. Nothing you do here leaves your browser.

A Collage Showing 49 Drill Holes Made by NASA's Curiosity Mars Rover

https://assets.science.nasa.gov/dynamicimage/assets/science/psd/photojournal/pia/pia26804/PIA26804.jpg?w=1536&h=1536&fit=clip
NASA's Curiosity Mars rover has captured 49 drill holes on the Martian surface, each 0.63 inches in diameter, to collect rock samples for analysis. The images were taken by the Mars Hand Lens Imager (MAHLI) camera on the rover's robotic arm.

Show HN: Take Voyager 1 apart, down to a single memory chip

https://rshekhar.com/voyager/og.jpg
Two spacecraft launched in 1977, still talking to Earth from interstellar space. This is where they are right now.

Show HN: Open-source, real-time object tracking for live video feeds

https://raw.githubusercontent.com/aryamandarda/tamserve/main/docs/media/camera-wall.gif
Tamserve is a software that uses Meta's EdgeTAM and SAM 2.1 models to track objects in video feeds, allowing for fast and accurate tracking of people, objects, and vehicles across multiple cameras. It achieves this by using continuous batching, GPU acceleration, and optimized memory management.

Launch HN: Maritime (YC F26) – A managed cloud for production AI agents

Ary and Maria, co-founders of Maritime, are building an agent-native cloud to handle AI agent lifecycle management. They aim to simplify the infrastructure problem for AI agents by providing a comprehensive solution with built-in services and cost optimization.

Collisions in Hash Functions

A full break requires a collision across all rounds, standard padding, and output. It must include both inputs and demonstrate broken collision resistance, not just a compression function collision.

Show HN: ADE Index – a sourced and ranked index of 975 agentic coding tools

https://adeindex.com/media/shots/paseo/1-1600.webp
Updated 18 Nov 2026. 165 tools shipped a stable release in the last 24 hours. The apps where you run, watch and steer coding agents. Plus the 587 agents, orchestrators, editors and builders around them. 975 listed tools (970 current), plus 24 adjacent tools. One listing per product, a source for every fact. Four forms of agent workspace, each showing its highest Index scoreHow the score is ...

Palo Alto Networks to Acquire Applied Compute for $4.5B

Palo Alto Networks acquires Applied Compute, a platform for post-training and inference of open-weight AI models, to help enterprises train, adapt and deploy AI securely. The acquisition is valued at $4.5 billion and aims to bring security built-in to open-source and open-weight AI models in public clouds or sovereign environments.

Software companies are trying to kill home computing?

https://l-files.livejournal.net/og_image/2602278/2055?v=1784261761
Software companies aim to kill personal and home computing by making it less appealing, as they can't make money from selling software that people already own. They're shifting focus to services, like cloud computing, where users pay monthly fees for access to software and data storage.