KDZONE 情報站

Article URL: https://fazamhd.com/mental-models/networking/ Comments URL: https://news.ycombinator.com/item?id=48871470 Points: 179 # Comments: 47

It's unclear how the planet avoided its star's bloated red giant stage.

For the first time since 1995, the Bureau of Land Management is rewriting its grazing regulations.

本週美國網路與基礎設施安全局（CISA）再度更新已遭利用漏洞清單（KEV），加入開放原始碼內容管理系統Joomla!兩款外掛套件的漏洞，分別是：iCagenda的CVE-2026-48939，以及Balbooa Forms的CVE-2026-56291，皆因CVSS嚴重度評分為10分漏洞出現遭到積極利用的活動，所以被列入K

Reinforcement learning uses error information to adjust control algorithms.

More police and firefighters use drones to catch and deter illegal fireworks.

https://www.science.org/doi/10.1126/science.aei1285 Comments URL: https://news.ycombinator.com/item?id=48866134 Points: 238 # Comments: 82

"Clearly, they admire the work that's being done by SpaceX and are trying to replicate it."

The Quest shipwreck is in worse shape than expected, but it has turned into a thriving marine ecosystem.

Six years in prison for man who "sold out the very victims he was hired to represent."

Article URL: https://lwn.net/SubscriberLink/1080822/990a8a5e2d379085/ Comments URL: https://news.ycombinator.com/item?id=48864252 Points: 210 # Comments: 215

Article URL: https://www.nyc.gov/mayors-office/news/2026/07/mayor-mamdani-announces-landmark--click-to-cancel--consumer-prot Comments URL: https://news.ycombinator.com/item?id=48864158 Points: 181 # Comments: 2

The children who get a combination shot are some of the most vulnerable.

Article URL: https://casp.ac/reports/ai-enabled-terrorism Comments URL: https://news.ycombinator.com/item?id=48863707 Points: 215 # Comments: 185

https://www.nyc.gov/mayors-office/news/2026/07/mayor-mamdani... Comments URL: https://news.ycombinator.com/item?id=48863464 Points: 515 # Comments: 250

Article URL: https://www.zdnet.com/home-and-office/networking/spacex-wants-to-launch-100000-more-starlink-satellites/ Comments URL: https://news.ycombinator.com/item?id=48863064 Points: 211 # Comments: 691

Dozens of titles too taxing for Steam Deck are still unrated for the new hardware.

Article URL: https://vfxblog.com/2017/08/23/the-tech-of-terminator-2-an-oral-history/ Comments URL: https://news.ycombinator.com/item?id=48862365 Points: 216 # Comments: 76

Article URL: https://www.smithsonianmag.com/smart-news/spider-silk-loses-top-spot-natures-strongest-material-snails-teeth-180954346/ Comments URL: https://news.ycombinator.com/item?id=48862252 Points: 210 # Comments: 158

Affected users will have to configure their lights and settings all over again.

The Earth may not be that massive, but it still distorts space-time.

Article URL: https://www.jeffgeerling.com/blog/2026/quadrf-can-spot-drones-and-see-wifi-through-my-wall/ Comments URL: https://news.ycombinator.com/item?id=48861717 Points: 578 # Comments: 192

How migrating Copilot code review to shared Unix-style code exploration tools reduced review cost by reshaping agent workflows around pull request evidence. The post Better tools made Copilot code review worse. Here’s how we actually improved it. appeared first on The GitHu…

Digital Services Act may force Meta to make big changes on its platforms.

Cloudflare has recently introduced temporary accounts that let AI agents deploy Cloudflare Workers immediately, without first creating or authenticating with a permanent account. If left unclaimed, the accounts and their deployments expire automatically after 60 minutes. By Renat…

"Industry finally knows what NASA is asking of them."

VW's plan calls for half as many models but didn't mention closures or job cuts.

Bryan Oliver discusses the frontier of AI infrastructure: chaos engineering for large-scale GPU clusters. He shares how engineering leaders can handle complex topologies, network protocols like RDMA, and NUMA misalignments. Discover seven practical fault-injection strategies to m…

Article URL: https://unstack.io/write-code-like-a-human-will-maintain-it Comments URL: https://news.ycombinator.com/item?id=48859701 Points: 331 # Comments: 283

Article URL: https://ianreppel.org/how-successful-companies-go-blind/ Comments URL: https://news.ycombinator.com/item?id=48859678 Points: 217 # Comments: 78

Article URL: https://avi.press/posts/2026-07-10-after-7-years-in-production-scarf-has-reluctantly-moved-away-from-haskell.html Comments URL: https://news.ycombinator.com/item?id=48859673 Points: 190 # Comments: 227

Smart Tiered Cache allows for precise upper tier selection for origins hosted on AWS, GCP, Azure, and Oracle Cloud with customer-provided cloud region hints.

以色列特拉維夫大學（Tel Aviv University）等研究人員揭露一種名為HalluSquatting的新型大型語言模型（LLM）攻擊手法。攻擊者可先找出LLM在查找儲存庫、AI代理技能（skill）等外部資源時容易編造出的錯誤名稱，搶先註冊這些名稱，藉此建立假資源，並植入對抗性提示內容；一旦AI程式開發助理拉取這些假資源，就可能執行攻擊者控制的指令。

OpenAI周四（7/9）宣布升級ChatGPT桌面程式，新版整合原本獨立的Codex程式開發工具，也內建了瀏覽器與電腦操作等功能，並將逐步淘汰去年10月推出的瀏覽器ChatGPT Atlas。根據The Verge的報導，ChatGPT Atlas預計會在8月9日停止運作。

在預覽最新GPT-5.6模型系列近二周後，OpenAI獲得美國政府許可，周五（7/10）宣布將GPT-5.6旗艦模型Sol及Terra、Luna模型全面推上線，包括合作的GitHub與Microsoft 365 Copilot助理。 GPT-5.6號稱能兼顧智能和運作效率，在程式撰寫、知識工作、生命科學及網路安全領域都達到領先（state-of-the-art）等級，在不同領域超越前代模型，或Google Gemini與Anthropic Claude Mythos。OpenAI還新增ultra模式，可運用多個子代理人加速複雜任務的進行。

社交工程手法ClickFix過去常被用於散布竊資軟體與遠端存取木馬，如今也被攻擊者用於金融詐欺。

The Linux Foundation has launched Akrites, a new industry-wide initiative aimed at defending the world's most critical open source software against a rapidly evolving generation of AI-enabled cyber threats. By Craig Risi

Article URL: https://acoup.blog/2026/01/30/collections-the-late-bronze-age-collapse-a-very-brief-introduction/ Comments URL: https://news.ycombinator.com/item?id=48858737 Points: 367 # Comments: 251

"I have been waiting a long time to finally get up there..."

The AC culture wars may be solved by advances in environmentally friendly technology.

AI安全業者Adversa AI於6月30日揭露名為GuardFall的攻擊手法。研究指出，多款開源AI程式開發代理與電腦操作型AI代理在執行Shell命令前，安全檢查主要仰賴原始命令字串比對或阻擋清單，攻擊者可利用Bash等命令列直譯器（Shell）的特性，在實際執行命令前，先展開變數、移除引號並重新解析命令，避開代理程式檢查，讓惡意命令以開發者帳號權限，在開發者主機或CI環境執行。

從4月開始，外號為Nightmare Eclipse的研究人員Chaotic Eclipse陸續披露多個針對Microsoft Defender與Windows內建功能的零時差漏洞，而且是以漏洞利用的概念驗證程式發表，震撼整個資安領域，這些漏洞後續大多都已獲得修補，最近於6月上半揭露的RoguePlanet漏洞CVE-2026-50656

"We are delighted to actively help shape the ramp-up of the Ariane 6."

Article URL: https://ec.europa.eu/commission/presscorner/home/en Comments URL: https://news.ycombinator.com/item?id=48858292 Points: 262 # Comments: 177

Article URL: https://www.gingerbill.org/article/2026/07/10/good-tools-are-invisible/ Comments URL: https://news.ycombinator.com/item?id=48858121 Points: 430 # Comments: 198

隨著AI加速導入各產業與公共服務，

OpenAI周四（7/9）推出ChatGPT Work，將ChatGPT從對話工具擴展為可執行多步驟任務的AI代理人。ChatGPT Work可跨應用程式與檔案蒐集資訊，自主規畫並執行工作流程，定位類似Anthropic的Claude Cowork，主要鎖定企業與知識工作者市場。

GitHub has made the redesigned GitHub Copilot CLI terminal interface generally available. It adds a tabbed layout for sessions, gists, issues, and pull requests; an in-session, form-driven setup for MCP servers, skills, and plugins that avoids hand-editing config files; and a cle…

Adding cloud regions changes latency and cost in ways simple math can't capture. This article presents a framework from multiple launches: decompose your latency budget before committing to infrastructure, choose deployment patterns by consistency and traffic profile, and optimiz…

In this podcast Shane Hastie, Lead Editor for Culture & Methods spoke to Gabriela Moreira about making formal methods accessible through the Quint specification language, how AI is dramatically lowering the barrier to entry for formal specification and model-based testing, and wh…

Article URL: http://yummymelon.com/devnull/in-emacs-everything-looks-like-a-service.html Comments URL: https://news.ycombinator.com/item?id=48857230 Points: 240 # Comments: 102

GitHub前執行長Thomas Dohmke的新創公司Entire，推出分散式Git託管網路，讓使用者可將GitHub上的公開或私有程式碼儲存庫，鏡像到Entire位於不同區域的節點。AI程式代理不必都向原本的GitHub儲存庫大量讀取資料，而可從較近的Entire節點下載程式碼與拉取更新，降低原始儲存庫在高並行存取下的負載。

隨著AI模型加入安全護欄，已減少提示注入攻擊的可能性。但研究人員發展出新提示注入手法，將攻擊行為拆成一組工作流（workflow），成功讓GitHub Copilot撰寫出惡意程式碼。 研究人員Abhishek Kumar與Carsten Maple將其手法稱為工作流層級越獄建構（workflow-level jailbreak construction）。這種攻擊法是將一個惡意攻擊指令，切成看似一般軟體開發流程的多個階段，這些階段一旦組合起來，就能對GitHub Copilot起作用，按研究人員意思撰寫攻擊程式碼。

CERT/CC協調中心警告，HP家用多功能事務機Deskjet 2800系列傳出韌體存在缺少授權的漏洞CVE-2026-13753，可讓攻擊者可能取得Wi-Fi憑證、管理配置及其他敏感資訊。目前HP尚未修補。

In a recent article, Datadog engineer Arnold Wakim shared what worked, what didn't, and the lessons they learned while evolving a critical production system using AI to overcome hard limits in its storage backend and significantly improve performance. By Sergio De Simone

Article URL: https://nevo-project.epfl.ch/ Comments URL: https://news.ycombinator.com/item?id=48856904 Points: 278 # Comments: 230

WordPress 7.0, released on May 20, 2026, includes new AI infrastructure, a redesigned admin interface, and updated design tools. Key features comprise an AI Client, Abilities API, and Command Palette, alongside increased PHP requirements. Community feedback is mixed, particularly…

漏洞攻防態勢面臨結構性轉折，資安界頻頻引用「Zero Day Clock」網站最新數據，反映漏洞揭露到被利用的時間縮短至僅2小時，而近期Citrix、Adobe的漏洞，在揭露24小時、甚至2小時內，便被資安業者觀察到出現利用的攻擊活動，亦證實上述攻擊態勢。 另有FatFs漏洞、中國騰達路由器後門的消息亦值得特別留意，皆存在研究人員或機構通報未得到回應，導致無修補程式可用的情形，突顯物聯網設備資安風險在快攻態勢下更形嚴峻。 此外，數發部公布「人工智慧風險分類框架」，以及資安署聚焦臺灣「軟體業」啟動產品漏洞獵捕活動，亦呈現臺灣在AI風險管理與產品資安強化的重…

華爾街日報報導，OpenAI應用部門執行長Fidji Simo將下臺，公司或將重新尋找繼任者。 報導指出，Simo在內部公開信中說明，其病假超出原先預期，情況也更嚴重，因此她將辭去現任應用事業群執行長一職，轉為兼任顧問。她是於4月首度披露自己罹患神經免疫方面的疾病而必須請長假。

Meta周四（7/9）發表Muse Spark 1.1，升級旗下由超級實驗室（Meta Superintelligence Labs，MSL）開發的首款模型Muse Spark，提升模型在代理人任務中的規畫、協調與執行能力，並強化工具使用、程式開發、電腦操作及多模態理解。目前Muse Spark 1.1已登上Meta AI App及meta.ai的Thinking模式，Meta亦首度釋出Meta Model API預覽版供開發者存取。