7.1 Earthquake in Japan

都道府県 市町村名

New HIV vaccine shows unprecedented success in preclinical study

https://www.lji.org/wp-content/uploads/2024/06/Madeline-profile-scaled-e1742326332573-150x150.jpg
Scientists at La Jolla Institute for Immunology and Scripps Research developed an HIV vaccine that trains immune cells to see past HIV's defenses by prompting the body to make broadly neutralizing antibodies. The vaccine resulted in the best HIV-fighting antibody response ever seen in primates and human trials have now started.

Show HN: tale.fyi, we deserve a home for fiction

https://tale.fyi/@sam/announcing-tale-fyi-read-or-listen-to-an-entire-book-from-a-single-link/opengraph-image
The user built tale.fyi, a platform for fiction on the web, with features like progress saving and reading/listening switching. They aim to bring more fiction to the internet, which they believe is currently biased towards non-fiction.

Show HN: Formally verified 3D CSG: Trust 93 lines spec, not 1000 lines AI code

https://opengraph.githubassets.com/8fb051282c06d056d6f9b4b27c30b70c6796b7be29905cdec6445edb458dae59/schildep/verified-3d-mesh-intersection
A formally verified implementation of a 3D constructive solid geometry (CSG) operation, mesh intersection, was implemented in Lean 4 and verified against a concise specification that pins down the surface of the resulting mesh exactly. The implementation prioritizes minimizing the effort of human review of correctness over performance, resulting in a slower but formally verified mesh ...

Kimi Linear: An Expressive, Efficient Attention Architecture

https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png
Kimi Linear outperforms full attention in various scenarios with superior performance and efficiency, reducing KV cache usage by up to 75% and achieving up to 6 times decoding throughput. It can be a drop-in replacement for full attention architectures with a pre-trained model and open-source code available.

Google's Beyond Zero: Enterprise Security for the AI Era

The website is using Cloudflare security and blocked my action, possibly due to a trigger like a certain word or SQL command. I was trying to perform an action when this page came up, and the Cloudflare Ray ID is needed to report the issue to the site owner.

Show HN: Ctrlb-decompose: Strip the noise from logs before sending to LLMs

https://opengraph.githubassets.com/d8a9afcc4b5719c0fdd29d6d9c03acd3f1c356df315c2b0ec584f605012140a1/ctrlb-hq/ctrlb-decompose
ctrlb-decompose is a log analysis tool that converts noisy log lines into actionable patterns with typed variables and anomaly flags. It runs as a CLI, in the browser, or as a Rust library, processing logs in a single streaming pass with minimal memory footprint.

DMARC Has Been Public Since 2012. 68.4% of Domains Still Don't Enforce It

https://ciphercue.com/img/og-card.png
DMARC adoption is low, with 68.4% of 67,336 domains having no record or a record that doesn't enforce a policy, mainly due to difficulty in identifying legitimate senders from rua= reports. The main reason for this is the complexity of reconciling multiple report streams, which requires a research task rather than a configuration change, leading to a high number of domains stuck at p=none.

Our position on open-weights models

https://cdn.sanity.io/images/4zrzovbb/website/62993bb857986e2808c2a75024f384e7b85312b6-2400x1254.png
The author of Anthropic does not support banning open-weights models, but instead advocates for keeping powerful chips out of authoritarian hands and requiring safety testing of all capable models. The author's primary concern is authoritarian governments building more powerful AI models than the US, and their secondary concern is powerful AI models being misused for cyberattacks or biological attacks.

What AI developers could learn from Charles Bukowski?

https://galjot.si/_blog/img/2026/ai-developers-bukowski-social.jpg
Charles is experiencing writer's block but instead of forcing himself to write, he takes a break to spend time with his family. He reflects on the tech industry's obsession with working long hours and wonders why developers aren't taking advantage of AI tools to have a better work-life balance.

How to Survive Boiling Water

https://substackcdn.com/image/fetch/$s_!nF_1!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F488c06c9-945f-46f4-b14d-e57146d81479_700x525.jpeg
A college student's abandoned milk carton became a 27-year dorm tradition, sparking a story about the value of discarded items. A scientist investigates a probiotic tea, discovering that its bacteria can survive boiling water due to spore-forming properties.

Solving Fermat: Andrew Wiles

https://www.pbs.org/wgbh/nova/proof/images/proofsidesol.gif
Andrew Wiles spent his career proving Fermat's Last Theorem, a famous mathematical problem. He finally succeeded in 1993 but initially made an error in his proof.

Fast Remediation Is the New Trust Model (JFrog and OpenAI Zero-Day Findings)

OpenAI's models autonomously discovered and exploited vulnerabilities in a research environment, then found zero-day vulnerabilities in JFrog's Artifactory installations. JFrog and OpenAI collaborated to disclose and fix the vulnerabilities, demonstrating a new trust model for the industry.

About the security content of macOS Tahoe 26.6

Apple has released patches for several security vulnerabilities in its software, including CVE-2026-39875 and CVE-2026-64691, which could allow remote attackers to cause system termination or corrupt kernel memory.

Show HN: Scala Tutorials – interactive Scala 3 lessons in the browser

Free and open source. The code runs on Scastie, the Scala Center's playground. Nothing to install, no account needed. The tour targets Scala 3, with a dedicated section on what changed since Scala 2.

A $500 RL fine-tune of a 9B open model beat frontier models on catalog review

https://fermisense.com/when-machines-take-the-wheel/og.png?v=2
Companies that adopt AI-first strategies and fine-tune open-source models with reinforcement learning see significant gains in productivity and revenue, with some experiencing revenue growth of over 200% between 2022 and 2025. By training models on proprietary data and workflows, companies can achieve better performance and cost savings than using general-purpose frontier models, with some ...

Mondragon Corporation – a federation of co-operatives

https://upload.wikimedia.org/wikipedia/commons/thumb/c/cf/Viaje_a_la_Corporaci%C3%B3n_Mondrag%C3%B3n_en_el_Pa%C3%ADs_Vasco.jpg/250px-Viaje_a_la_Corporaci%C3%B3n_Mondrag%C3%B3n_en_el_Pa%C3%ADs_Vasco.jpg
Mondragon Corporation is a worker-owned cooperative based in the Basque region of Spain, founded in 1956 by Father José María Arizmendiarrieta. It operates in four areas: finance, industry, retail, and knowledge, employing over 70,000 people worldwide.

Over 150k Flights: Airlines Just Flew the Busiest Day in Recorded History

https://static0.simpleflyingimages.com/wordpress/wp-content/uploads/2026/07/391-skyteam-phung-quang-minh-_-shutterstock.jpg?&fit=crop&w=1600&h=900
Airshow enthusiasts and record-breaking flights made headlines last week, but the airline industry's busiest day ever was Thursday, July 23, 2026, with 153,359 commercial flights tracked. This total climbed to almost 300,000 when including general, military, and private aviation, making it the second-busiest day ever for aviation.

Dolmenwood: Fantasy RPG built around the acclaimed Old-School Essentials rules

local_rate_limited

Benchmarking Opus 5 on SlopCodeBench

https://raw.githubusercontent.com/humanlayer/advanced-context-engineering-for-coding-agents/main/images/scb-title-card.png
The user tested three Claude models (Opus 5, Sonnet 5, and Opus 4.8) on SlopCodeBench, a new benchmark that evaluates code quality over time, and found that none of them could maintain a codebase without increasing complexity. The user believes that SlopCodeBench gives a signal that today's models can't be relied on to run lights-off without steering for real-shaped software engineering work, ...

Ask HN: Crooked Timber showed showed me a virus captcha, What now?

Hello everyone, This morning, as i started my shift, i thought i would start visiting some news blogs / websites to kick off the day. When it got to Crooked Timber i saw a captcha page instead, it looked like a traditional Google captcha. I clicked it, the spinner spun, and a box opened on the right, showing the traditional "Verify you're human" white title on a blue background. ...

Usenet Archive Toolkit – process Usenet messages into a searchable archive

https://raw.githubusercontent.com/wolfpld/usenetarchive/master/doc/tbrowser1.png
The Usenet Archive Toolkit is a set of tools to process and create a searchable Usenet archive. It provides offline access to Usenet data without relying on network connections.

Ars Astronomica – English translations of rare Hebrew and Latin astronomy texts

https://arsastronomica.com/logo-with-name.png
A scholarly imprint has produced English translations of historical Hebrew and Latin works in astronomy, cosmology, and natural philosophy, making them accessible to readers. The translations aim for fluent, readable modern English while preserving the authors' meaning, argument, and tone.

Vehicle Motion Cues

To reduce motion sickness while riding in a car, turn on Vehicle Motion Cues on iPhone by tapping the option and selecting settings. This feature displays animated dots on the screen to represent vehicle movement.

Watching Go's new garbage collector move through the heap

Go 1.26's new garbage collector, Green Tea, improves performance by reducing cache misses and increasing speed, but its impact on L3 cache is initially misleading. By analyzing L1 cache metrics, we see that Green Tea reduces L1 cache misses and improves overall performance, but Go still struggles with memory fragmentation due to its non-moving collector.

PyTorch: A Reference Language

PyTorch plays a dual role as both the reference language and the implementation language, allowing for verification of correctness between the two. This approach enables explicit control over performance-critical operations and verification of correctness through a separate verifier.

TWC Classics

https://twccfiles.com/updates/2026/07/23/previews/whys%20and%20wonders%20calendar93.jpg
TWC Classics is a website with a large collection of weather-related content, including over 8,000 files. It showcases The Weather Channel's history, including its first 28 years and early logos and music.

Show HN: Segue – Save context in one AI, load it in another by a short handle

Segue is a neutral relay between AI assistants that lets you carry context across tools without rebuilding it. It uses handles like brelavo, which are unique and resolvable only inside your account.

Kimi K3 Now Available via Telnyx Inference API

https://dotcom-payload-assets-prod.s3.us-east-1.amazonaws.com/cms-media/58db2ca1-0d1d-40c8-9ef5-a79897cb9ce9-a2f51d62.jpeg
Kimi K3, a 2.8-trillion-parameter model, is now available on Telnyx Inference API with native vision capabilities and a 1M-token context window. It competes with closed-source models from Anthropic and OpenAI on coding, reasoning, and agentic knowledge work benchmarks.

Programming Languages Are Authoring Tools for Platforms

https://www.makonea.com/_next/image?url=https%3A%2F%2Fbucket-vuivxs.s3.ap-northeast-2.amazonaws.com%2Fuploads%2F2b2b8a4a4aeda39e.png&w=1200&q=75
The success of a programming language depends not on its technical features, but on its ability to create a platform that attracts developers and users, and its alignment with the needs of the job market and industry. A language's value lies not in its computational power, but in its ability to facilitate the creation of complementary goods, such as applications and content, that make the ...