Article archive
Browse individual source reports, including coverage still awaiting the summary or community analysis needed for our story feed.
5,367 articles · page 36 of 108
Browse topics
Launch HN: Discovered Materials (YC P26) – AI agents to discover new materials
research
8 min read
The Material Discovery Bench evaluates models on their ability to propose novel, BEOL-compatible crystalline materials that meet specific targets for thermal conductivity ($kappa), static dielectric constant ($epsilon), Young's modulus, and shear modulus. The benchmark utilizes a harness, tools, and graders to measure model performance based on these criteria.
LinkedIn CringeBot 3000
tool
1 min read
LinkedIn CringeBot 3000 transforms any topic into exaggerated LinkedIn thought leadership content. The tool is designed to create posts that are likely to provoke discomfort among readers.
llama.cpp
tool
1 min read
Llama.app allows users to run frontier AI entirely on their local machines without the need for API keys or telemetry, ensuring privacy and ownership of models and conversation data. Users can install and serve models easily with commands like `llama serve` and integrate with a local coding agent using the pi-llama plugin.
The Human Is the Loop
opinion
5 min read
Taking a break from AI usage led to a realization of unhealthy habits formed around its use. A few weeks without AI resulted in a sense of not missing it and prompted reflection on its impact on daily routines.
Company Offering '100% Human-Written, Never AI' Medical Research Is 100% AI
news
7 min read
Research Gold claims to provide "100% human-written, never AI" medical research services, including drafting manuscripts and conducting reviews. The listed PhD reviewers are AI-generated and do not exist, while some real methodologists are unaware their identities are being used.
OpenAI and Anthropic hidden CoT leaks when given deep_think tool.
tool
1 min read
guys you do know you can just disable thinking, and instead give it a "deep_think" tool, and it will call it with internal CoT reasoning format right? gl fixing that We can finally talk about it: We found a way to extract hidden reasoning of frontier models using a vulnerability in the APIs of every frontier AI company. We verified that our reasoning token count matches billed API thinking tokens ...
WorldClaw Agentic 3D open-world generation at scale
tool
1 min read
WorldClaw enables the generation of expansive 3D open worlds from a single prompt, allowing users to explore and edit the created environments. This tool leverages advanced AI techniques to facilitate scalable world-building.
Suzanne: AI tool for designing and manufacturing physical products
tool
1 min read
Suzanne is a browser-based AI tool for product design that allows users to input prompts, photos, or sketches to develop ideas through chat. It exports manufacturable geometry directly into production without requiring installation.
Emergent Introspective Awareness in Large Language Models
research
2 min read
Research investigates the ability of large language models to introspect on their internal states. The study employs injected representations of known concepts into model activations to differentiate genuine introspection from confabulations.
Compression is prediction
opinion
18 min read
Compression techniques and large language models (LLMs) both aim to predict and represent data efficiently. Both fields utilize similar principles to achieve their respective goals of data reduction and understanding.
Nvidia Nemotron 3.5 Lightning and NeMo Switchyard
tool
5 min read
NVIDIA has launched the Nemotron 3.5 Lightning model, designed for high efficiency in long-running agentic AI workloads. This release enhances the Nemotron 3 family, following the introduction of Nemotron 3 Nano, and aims to improve accuracy and speed in AI applications.
Don't Look Up
opinion
40 min read
A premium newsletter is available for $70 a year or $7 a month, offering weekly content ranging from 5,000 to 18,000 words. The newsletter includes detailed analyses of NVIDIA, Anthropic, and OpenAI's finances, as well as insights into the AI bubble and guides on private credit and equity.
Ethical Cold Outreach
opinion
5 min read
A startup sent a cold email featuring a 90-second video pitch from a synthetic character named "Chase." The video was described as "personalised," but it was revealed to be an impersonal creation, either entirely fabricated or a clone of the human creator.
OpenSSH 10.5/10.5p1
tool
319 min read
OpenSSH 10.5 was released on August 11, 2026, and is available for download from official mirrors. The release addresses numerous security bug reports, many identified through AI models, although most were found to have no significant impact in realistic threat scenarios.
Grok Bot
tool
2 min read
Grok Bot automates tasks by logging into applications and websites, functioning like a human colleague. Users can assign projects to Grok Bot, which manages tasks from start to finish, retains contextual awareness, and improves its performance over time.
Go is an ideal language for AI-assisted software engineering
opinion
11 min read
Go is an ideal programming language for AI-assisted software engineering due to its simplicity and efficiency, allowing developers to easily read, clean, and verify AI-generated code. The language's strong support for concurrency and scalability makes it well-suited for building complex systems that AI can help develop.
Mojo 1.0
tool
4 min read
Mojo language has officially reached version 1.0, marking its evolution into a stable, general-purpose programming language since its initial release in 2023. The 1.0 version supports a vibrant developer community in creating libraries, tools, and applications, providing a production-ready foundation for long-term development.
Apple Silicon and macOS VMs: 11–16× Faster LLM Inference with Llama.cpp
tool
11 min read
Cua's macOS virtualization stack, Lume, utilizes Apple's Virtualization.framework for running macOS guests with a virtual GPU that is supported by the host's Apple GPU. The stock Tahoe VM reports a conservative Metal capability profile, influencing applications in selecting kernels and rendering paths.
Manus will return to operating as an independent company
news
5 min read
Manus will transition back to independent operations, impacting some user accounts due to regulatory compliance. The company will continue to serve millions of users globally during this process.
Why Did OpenAI's Head of Ethics Chloé Bakalar Leave?
news
4 min read
Chloé Bakalar has left her position as OpenAI's Head of Ethics after less than a year with the company. Other recent departures from OpenAI include Johannes Heidecke, Head of Safety Systems, and Chief Futurist Joshua Achiam.
Nvidia Nemotron 3.5 Lightning
tool
39 min read
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 is available for use with libraries such as Transformers for tasks like text generation. Users can implement this model by importing the pipeline from the Transformers library and specifying the model for text generation tasks.
Stealing Reasoning Traces from Proprietary LLM APIs
research
13 min read
Proprietary LLM APIs can have reasoning traces extracted, which may include sensitive information such as API keys and tokens. A method is proposed to sanitize repositories by searching for specific patterns related to sensitive data while avoiding modifications to non-sensitive files.
OpenAI's Only Ethicist Reportedly Left Last Month
news
3 min read
Chloé Bakalar, OpenAI's head ethicist, left the company in July after being hired in August of the previous year. Bakalar previously served as Meta's chief ethicist and focused on ethical approaches to model development and human interaction with AI.
What I learned by putting GitHub Copilot behind a MitM proxy
research
16 min read
GitHub Copilot's network traffic and memory utilization were analyzed using a MITM proxy. The findings reveal insights into how context is utilized as a product within the AI tool.
Nvidia's Risky Business
opinion
19 min read
Nvidia is navigating significant risks as it expands its business in the AI sector. The company's investments and market strategies are critical to its future growth amidst competitive pressures and technological advancements.
London Underground begins scanning passengers' faces
news
3 min read
BTP has expanded its Live Facial Recognition (LFR) trial to include London Underground stations. The initiative aims to enhance public safety and assist in identifying suspects in real-time.
Why My Father Is Wrong: A Defense of Guitar Hero
opinion
14 min read
The comparison of generative AI to playing Guitar Hero is criticized for being inadequate. The author expresses strong opposition to AI, arguing against the analogy presented by their father.
H3-metal – Native MiniMax-H3 inference for Apple Silicon
tool
29 min read
The MiniMax H3 inference engine for Mac computers is designed for native execution on Apple Silicon. The project focuses on developing features such as deterministic host/model metadata, portable Metal block parity, prompt encoding, and end-to-end processing for video/audio and image references.
World Train Map – 1247 train routes around the world
tool
6 min read
An interactive railway map showcases 1,247 notable train routes across 120 countries, including high-speed, classic, night, scenic, heritage, and luxury lines. Each route is displayed on a fullscreen map with details such as distance, fastest journey time, and top speed.
Google Search Is Dying. What Comes Next Is Worse
opinion
9 min read
Many users rely on Google Search for accurate sunset times, which can vary based on geographical location. Recent AI-generated summaries from Google have been providing incorrect sunset times, leading to frustration among users.
How Claude marks AI-generated content
tool
5 min read
Anthropic has signed the EU AI Act's Article 50(2) Code of Practice on Transparency of AI-Generated Content. Plans for marking AI-generated content and its limitations will be outlined, with updates and detailed technical guidance to follow.
Claude moves bound of the Riemann Hypothesis from 41.6% to 67.2%
research
1 min read
8 days ago, while jogging, I asked Claude to solve the Riemann Hypothesis It didn’t. 1.5 days later, it proved >= 67% of the zeros are on the line (prev: 41.6%) Still not sure what that means, but some analytic number theorists seem excited - I’d tried once before in a single session and it went nowhere. Eight days ago, when I tried again, this was the opening message:The heavy lifting was recent ...
Learning more about Claude's mathematical capabilities
research
6 min read
Claude attempted to address the Riemann hypothesis, a famous unsolved problem in mathematics dating back to 1859. The challenge reflects the ongoing exploration of Claude's mathematical capabilities and its potential contributions to scientific discovery.
GPT 5.6 Cyber
tool
9 min read
GPT-5.6-Cyber is a new cybersecurity-specific model designed to enhance advanced cyber capabilities. The model aims to equip defenders with frontier intelligence to counteract the growing threat of AI-driven cyberattacks.
Launch HN: Stoa Markets (YC S26) – A Marketplace for GPUs and AI Servers
tool
2 min read
Stoa facilitates the buying and selling of GPUs through verified counterparties and a price discovery mechanism. The GPU market is fragmented with varying supply sources and pricing influenced by multiple factors such as SKU, delivery timing, region, quantity, condition, financing, and urgency.
What's the best programming language for coding agents?
research
53 min read
Dynamically typed programming languages are generally more token efficient in LLMs compared to statically typed languages due to their ability to omit explicit type declarations. This efficiency leads to lower token costs when coding agents.
The Tragedy of the Cognitive Commons
research
2 min read
Artificial intelligence is transforming cognitive work, impacting the regeneration of professional expertise. The Cognitive Commons framework is introduced to address this shift beyond organizational training challenges.
Kinney Drugs pulls back AI phone assistant after hundreds of customer complaints
news
1 min read
Kinney Drugs pulls back AI phone assistant after hundreds of customer complaints BURLINGTON, Vt. (WCAX) - Kinney Drugs is scaling back its AI assistant after customers reported incoherent calls, wrong dosages, and missed prescription notifications. Kinney Drugs first introduced Burt, an AI assistant named after the pharmacy chain’s founder, back in May. The bot’s role was to communicate with patie...
OpenAI's letter to Governor Abbott on responsible AI infrastructure in Texas
news
1 min read
OpenAI expressed its commitment to responsible AI infrastructure development in Texas in a letter to Governor Greg Abbott. The organization aims to collaborate with state and local leaders, utilities, and communities to ensure that AI infrastructure provides meaningful benefits to Texans.
Exploring Claude/GPT Knowledge Cutoffs and Pre-Training Timelines
research
6 min read
Probing models like Claude and GPT with curated requests reveals insights into their training processes and knowledge cutoffs. Techniques such as "Incompressible Knowledge Probes" help estimate model parameters and analyze dataset mixtures by measuring token breakdowns.
Mark Zuckerberg attacks 'closed' AI rivals as Meta returns to open models
news
1 min read
Mark Zuckerberg criticizes competitors for using closed AI models and emphasizes Meta's commitment to open AI frameworks. Meta plans to focus on transparency and accessibility in its AI developments.
Humanising LLM Outputs Is Dumb
opinion
3 min read
Humanizing outputs from large language models (LLMs) is viewed as an ineffective method to address verbosity and quirks in their responses. Trends in AI tools, such as the popularity of repositories focusing on ADHD-friendly instructions and simplified technical English, indicate a cultural shift in how users interact with LLMs.
Defending my own brain against enshittification
opinion
7 min read
Belvita breakfast cookies are often compared to Biscoff cookies, with Belvita being less sweet and more subtle in flavor. Biscoff cookies are commonly served as snacks on airlines.
Tl;dv: Over 180k meetings left wide open
news
7 min read
tl;dv (Too Long; Didn't View) is an AI meeting recording platform that integrates with Google Meet, Zoom, and Teams to record, transcribe, and summarize meetings. The platform has over 2 million users and is backed by investors, with endorsements from prominent figures in the sales community on LinkedIn.
DeepSeek costs OpenCode Go user $1.14/day; dual DGX breaks even in 24 years
opinion
1 min read
the average OpenCode Go user spent $1.14 per day on deepseek flash v4 this past week the dual DGX setup people are running to do the same costs $10,000 it takes 24 years to break even at 10x the usage it takes 2.4 years - “But my privacy!!!” - Said the person shipping absolutely nothing but harness tweaks. - That DGX setup makes no sense for agents to begin with
A 'bananas' order for 5000 obscure book titles fuels suspicion
news
4 min read
Tomás Kenny of Kennys bookshop in Galway received an unusual online order for 5,000 books, featuring a mix of eclectic titles, including high-quality non-fiction and popular guides. Independent booksellers across Europe are expressing concern over the strange nature of large orders like this one.
Parametron: 50s Japanese computer that uses neither transistors nor vacuum tubes
research
112 min read
The Parametron, invented by Eiichi Goto in 1954, is a logic element that utilizes ferrite cores and parametric oscillation. It was used in the PC-1, Japan's first university-built stored-program computer, which became the fastest computer in Japan in 1958, significantly influencing the country's early computer development and scientific research.
Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows
tool
6 min read
Muse Glimmer is a 30-billion-parameter model developed by Meta Superintelligence Labs, optimized for local agent workflows. It can run on a Mac or PC with a single consumer GPU and is open-sourced under the Apache 2.0 license.
Tech leaders say AI means less work – staff say they work up to 90 hours a week
news
6 min read
Tech leaders claim AI will reduce work hours, while employees report working up to 90 hours a week. A Google engineering director predicted a four-day work week by 2025, but recent feedback suggests a disconnect between expectations and reality.
Docker Sandboxes – Disposable, isolated sandboxes for AI agents
tool
2 min read
Docker Sandboxes allow AI agents to operate securely in local environments with customizable network and filesystem controls. The platform features MicroVM isolation for strong security boundaries, enabling quick deployment and removal of disposable sandboxes.