Article archive
Browse individual source reports, including coverage still awaiting the summary or community analysis needed for our story feed.
5,369 articles · page 42 of 108
Browse topics
Toolcraft
tool
1 min read
Toolcraft is an open-source starter kit and UI library designed for developing custom design applications that incorporate AI features. It provides essential tools and components to streamline the creation of AI-driven design apps.
Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the Incident
research
27 min read
A detailed timeline of the July 2026 Frontier Lab Agent intrusion outlines the attack chain and phase activities over a 4.5-day campaign. An interactive replay is available for visualizing the attack across trust boundaries, emphasizing the emerging capabilities of frontier agents and the need for improved defense strategies.
Pacing the frontier
opinion
9 min read
1,134 employees from frontier AI companies assert that while AI has the potential to create a significantly improved future, there are risks associated with rapid capability development. They indicate that the automation of AI research could accelerate progress beyond current understanding or control.
Deflock Casa Grande
news
1 min read
Casa Grande's surveillance camera system monitored a student protest at City Hall and assisted police in locating missing individuals. The system's capabilities include identifying involved parties and monitoring activities, despite residents being informed that the cameras track vehicles only.
Discovering Cryptographic Weaknesses with Claude
research
18 min read
Researchers at Anthropic have utilized Claude Mythos Preview to uncover vulnerabilities in cryptographic algorithms. One attack significantly compromises the HAWK digital signature scheme, while another reveals a new method to exploit round-reduced AES, a widely used symmetric cipher.
You Could Have Come Up with Kimi Delta Attention
research
13 min read
Kimi Delta Attention utilizes bra-ket notation for clarity in mathematical derivations, where column vectors represent states, row vectors represent duals, numbers represent scalars, and matrices represent linear transformations. In this notation, vectors are oriented to the right by default, while keys are oriented to the left within the linear attention structure.
Kimi K3 Architecture Overview and Notes
tool
2 min read
The Kimi K3 architecture is a scaled-up production version of the Kimi Linear model, increasing from 48 billion parameters to 2.8 trillion parameters. A new component in Kimi K3 is the LatentMoE, which enhances its capabilities as the largest open-weight model currently available.
Anthropeum – Where in the world, and when, does this human artifact belong?
tool
1 min read
Anthropeum is a daily game that tests knowledge of human cultures through ten artifacts spanning 5,000 years of history. Players must guess the location and time of each object's creation and can see their ranking among participants.
Now is the time to give LLMs access to the ACM digital library
opinion
4 min read
The Association for Computing Machinery (ACM) is considering how to engage with AI technologies while maintaining the integrity of the ACM Digital Library. ACM has prioritized understanding the perspectives of authors and volunteers over monetizing content.
Una GPS smart watch – Repairable, USB-C charging, developer-friendly
tool
1 min read
UNA is a modular GPS watch designed for repairability and upgradability, allowing users to replace individual components instead of discarding the entire device. It features advanced heart rate monitoring and is built with sustainability in mind, promoting longevity in use.
What AI developers could learn from Charles Bukowski?
opinion
3 min read
AI developers can draw inspiration from Charles Bukowski's perspective on patience and the importance of waiting for the right moment to act. Emphasizing the need for creativity and alternative pursuits during periods of inactivity can enhance innovation in AI development.
Fast Remediation Is the New Trust Model (JFrog and OpenAI Zero-Day Findings)
news
3 min read
JFrog and OpenAI are collaborating to address zero-day security findings, emphasizing the importance of fast remediation in building trust. AI models are revealing new types of vulnerabilities, necessitating higher security standards for software products.
Kimi Linear: An Expressive, Efficient Attention Architecture
research
2 min read
Kimi Linear is a hybrid linear attention architecture that outperforms full attention in short-context, long-context, and reinforcement learning scenarios. It utilizes Kimi Delta Attention to achieve this efficiency.
Google's Beyond Zero: Enterprise Security for the AI Era
research
20 min read
Enterprise security is evolving to a high-frequency, AI-mediated defense model known as Beyond Zero. This model implements a contextual, risk-based, resource-level authorization system that operates at machine speed to enhance security for both humans and AI agents.
About the security content of macOS Tahoe 26.6
news
1 min read
macOS Tahoe 26.6 includes security updates that address vulnerabilities related to system integrity, memory management, and application sandboxing. The update enhances overall system security and protects user data from potential exploits.
Ars Astronomica – English translations of rare Hebrew and Latin astronomy texts
research
11 min read
Ars Astronomica produces the first English translations of historical Hebrew and Latin works in astronomy, cosmology, and natural philosophy. These texts, which include Gersonides’ 136-chapter mathematical astronomy, have never appeared in English and allow for comparative study of cosmological questions across different authors.
PyTorch: A Reference Language
tool
4 min read
PyTorch serves as a reference language for deep learning, providing a common framework of APIs and conventions. Reference implementations prioritize clarity over performance, leading to some confusion regarding the term "reference language" in the context of PyTorch.
"Opus 5 is a really bad model"
opinion
1 min read
- its /awful/, Opus-5 makes me feel like setting my hair on fire out of frustration. it's annoying to work with for high tasks: it constantly makes bad decisions and large mistakes so any token-usage you saved by not using Fable instead is useless/worse. It's a slop-cannon. - i feel like i’m being lied to here. i’ve been the biggest gpt > claude guy. but sol is pathetic. it makes every task into a...
Using an open model feels surprisingly good
tool
1 min read
Using an open model for AI inference allows users to have greater control over their data and deployment. Setting up an open code inference endpoint can provide a sense of ownership and freedom in managing AI tasks.
EYG: A Programming Language for Humans
tool
6 min read
Eat Your Greens (EYG) is a statically typed functional programming language designed to improve programming experiences. The language incorporates selected features aimed at enhancing usability and addressing common user frustrations in programming.
A $500 RL fine-tune of a 9B open model beat frontier models on catalog review
research
2 min read
An open-source model combined with proprietary task data and reinforcement learning has been implemented in various real-world scenarios. Bridgewater Associates utilizes AI to analyze a continuous influx of documents to determine their relevance to investment strategies.
Teacher Arrested for Clapping in Support of Anti-Data Center Activists
news
3 min read
High school physics teacher Lux Claridge was arrested by police for clapping in support of an anti-data center activist during a public meeting of the Emporia City Commission in Lyon County, Kansas. The incident raises concerns about the exercise of First Amendment rights in protests against data centers.
Don't ask an LLM for a confidence score
opinion
16 min read
LLM confidence scores often misrepresent the certainty of generated outputs. Relying on these scores can lead to misunderstandings about the reliability of AI responses.
Kimi K3 Now Available via Telnyx Inference API
tool
1 min read
Kimi K3, the first 2.8T parameter open-source model, is now on Telnyx Inference. 1M context window, native vision, configurable reasoning. Input $2.70/1M tokens.
Our position on open-weights models
opinion
8 min read
Dario Amodei, CEO of Anthropic, addresses concerns regarding open-weights models, particularly those from China. There are discussions among US officials about potentially banning the use of these models by US companies, prompting support from various tech companies for open-weights models.
My current strategy is to not read any of the code written by my agents
opinion
1 min read
I am trying to use Claude to help me write something, but I just don't feel comfortable letting it edit my files. Does anybody else feel the same? If I am responsible for code, I NEED to understand it, psychologically if for no other reason. Started programming in 1983. Old? - I’m significantly older than you. I started coding in the late 60s. My current strategy is to not read any of the code wri...
Professor's invisible prompt trap catches 32/35 students cheating with AI
news
2 min read
A professor at Alcorn State University used a hidden white font in an assignment prompt to instruct AI to generate irrelevant responses. This method successfully identified 32 students who attempted to cheat on their midterm exam.
MAI-Cyber 1
tool
4 min read
MAI-Cyber-1-Flash is integrated into MDASH, providing multi-agent vulnerability identification and remediation. This solution delivers performance at 50% of the cost of leading models.
Nvidia's $750B in Deals Reignite Circular AI Fears
news
1 min read
Nvidia Corp. is pursuing AI deals exceeding $750 billion, raising concerns about inflated demand and valuations in the industry. A partnership with SK Group will involve over $500 billion in business between the two companies.
Kimi-K3 Technical Report [pdf]
tool
1 min read
Kimi-K3 is a project hosted on GitHub by MoonshotAI that focuses on collecting and integrating user feedback. The documentation provides details on available qualifiers for user input processing.
Apple Will 'Watch Everything Burn' When the AI Bubble Bursts
opinion
9 min read
Memory prices have doubled, leading to increased costs for Macs and iPads, with iPhones expected to follow. Ed Zitron argues that the investments driving these costs will not yield sufficient returns, suggesting potential negative consequences for Apple when the AI bubble bursts.
AI companies spend record sums on Washington lobbying
news
1 min read
AI companies are spending unprecedented amounts on lobbying efforts in Washington, D.C. to influence regulations and policies affecting the industry.
Should you wash your solar panels?
research
4 min read
Cleaning solar panels can result in a 2%-5% increase in power output. Dust accumulation over time can significantly impact the efficiency of solar energy generation.
AI companies are shredding rare books
news
1 min read
🦔AI companies are bulk-buying rare books, scanning them through high-speed machines that cut the spines off, and shredding the originals. A service called ISBNdb facilitates orders of up to a million books and keeps buyers anonymous. Pre-2022 books are premium because they're free of AI-generated text. A federal judge ruled the practice is fair use because eliminating the original means only one c...
How is the Bun Rewrite in Rust going?
opinion
5 min read
Bun is undergoing a rewrite in Rust, which aims to leverage AI capabilities for the work typically done by open-source maintainers. Anthropic's AI tools, including a C compiler, are contributing to this rewrite process.
Why I Left Google DeepMind
opinion
61 min read
Cherished memories of the LessWrong community include acts of defiance in pursuit of goodness, such as challenging prestigious dogma, helping others despite social pressure, seeking truth beyond self-expectations, and resisting threats from powerful adversaries. These themes emphasize the importance of individual agency and truth-seeking in the face of societal and personal challenges.
Chinese chipmaker shares surge 470%
news
2 min read
ChangXin Memory Technologies (CXMT) shares surged nearly 470% during its debut on the Shanghai Stock Exchange's Star Market. This increase raised its market valuation to approximately 3.3 trillion yuan ($487 billion), making it the most valuable listed company in mainland China.
Kimi-K3 Releases on HuggingFace 7/27
tool
1 min read
Kimi-K3 is the latest model from Moonshot AI, set for release on July 27, 2026. It is the world's first open 3T-class model, featuring a new architecture based on Kimi Delta Attention and Attention Residuals, with native agentic capabilities including tool calling, browsing, and multi-step reasoning.
Scriptc by Vercel: TypeScript-to-Native compiler, no JavaScript engine in binary
tool
2 min read
scriptc compiles TypeScript into small, fast native executables without the need for Node, V8, or a JavaScript engine. It allows users to run TypeScript code directly and generates self-contained native binaries with minimal startup time.
We have proof automation now
opinion
18 min read
ImperialViolet discusses the advantages of dependently-typed languages like Coq, Rocq, and Lean for enforcing complex invariants through their type systems. These languages help prevent misunderstandings and integration issues in large teams by formally encoding requirements that might otherwise be lost.
What does GitHub's security team even do?
opinion
4 min read
Thousands of repositories on GitHub are currently distributing malware, accessible through the standard search function without special knowledge. Despite having a substantial budget, a dedicated security team, and AI resources, GitHub has not resolved this issue over the past two years.
Introduction to Data-Oriented Design [pdf]
tool
1 min read
Data-oriented design focuses on optimizing data structures and memory access patterns to improve performance in software development. This methodology emphasizes the organization of data to enhance cache efficiency and processing speed, particularly in game development.
It's not empowering to hand off the details
opinion
1 min read
AI enthusiasm often stems from a desire to manifest ideas without delving into intricate details. However, achieving quality and innovation requires a deep understanding and meticulous attention to the complexities involved.
How AST-grep Rewrote Tree-sitter in Rust and Made It 30% Faster
tool
16 min read
ast-grep rewrote Tree-sitter's C core in Rust, resulting in a 30% increase in parser speed and a 22% improvement in overall performance for ast-grep. The new core enhances parsing efficiency and accelerates tree reading.
This July I Was Fired from Simple AI (A Deeply YC Company)
opinion
5 min read
An individual was fired from Simple AI, a company associated with Y Combinator, in July. The post reflects on the implications of writing publicly in the context of LLMs and human-machine interactions.
An Inside Look at the Token Reseller Market
opinion
8 min read
Token fraud is a significant issue in the relay market, impacting companies that lose millions of dollars daily due to abuses such as free-credit exploitation and support chatbot manipulation. Software engineers working on AI gateways are increasingly addressing these challenges as token resellers exploit vulnerabilities.
The New AI Superpowers: Focus and Followthrough
opinion
4 min read
Burnout is increasing despite advancements in AI that can enhance task completion speed significantly. The complexity and overwhelming nature of managing multiple AI-driven tasks may contribute to feelings of burnout rather than alleviating them.
Google Discloses $94.1B in SpaceX Stock, Marking 6% Stake
news
1 min read
Google holds $94.1 billion in SpaceX shares, representing approximately a 6% stake in the company. Of this amount, $80 billion is subject to short-term sale restrictions and $14.1 billion is subject to long-term restrictions until the third quarter of 2027.
Terence Tao: Mathematics in the Age of AI [pdf]
research
1 min read
Terence Tao explores the intersection of mathematics and artificial intelligence, emphasizing how AI can enhance mathematical research and problem-solving. He discusses the implications of AI for the future of mathematical discovery and education.
Elevated Errors for Opus 5
news
1 min read
Users can subscribe to receive email and text message notifications for updates on elevated errors for Opus 5. SMS updates require phone number verification, while email subscriptions can be activated without additional steps.