1. AlphaGenome: AI for better understanding the genome (deepmind.google)
Introducing a new, unifying DNA sequence model that advances regulatory variant-effect prediction and promises to shed new light on genome function — now available via API.
28 篇热帖
Introducing a new, unifying DNA sequence model that advances regulatory variant-effect prediction and promises to shed new light on genome function — now available via API.
Richard P. Feynman在1966年写给前学生Koichi Mano的信中,强调了选择研究问题的重要性。Koichi在研究电磁波在湍流大气中的传播,称其为“谦虚和实际”的问题,但Feynman认为他对什么是有价值的问题有误解。Feynman指出,有价值的问题是那些能够真正解决或帮助解决的问题,无论大小,只要能对科学或社会有所贡献。他建议Koichi从更简单、更谦虚的问题入手,逐步获得成功和满足感。
Feynman通过自身经历说明,他曾研究过许多看似平凡的问题,如摩擦系数、晶体弹性、电镀附着力、中子扩散、电磁波反射、冲击波、中子计数器、电子捕获、折纸玩具、轻核能级和湍流理论等。这些研究虽然部分失败,但他从中获得乐趣并推动了科学进展。他强调,没有问题太小或太微不足道,只要能真正解决。
信中,Feynman还鼓励Koichi不要觉得自己是无名的。他指出,Koichi在家庭和同事中都有价值,应该公平评估自己,避免受错误理想的影响。这封信的核心是倡导务实的研究态度和积极自我认知。
A tetrahedron is the simplest Platonic solid. Mathematicians have now made one that’s stable only on one side, confirming a decades-old conjecture.
We're almost ready to issue certificates for IP address SANs from Let's Encrypt's production environment. They'll only be available under the shortlived profile (which has a 6-day validity period), and that profile will …
A blog
根据Ars Technica的测试,SteamOS在多数游戏中的运行帧率高于Windows 11。测试的五款游戏中,除《Borderlands 3》两者表现基本持平外,其余四款游戏在SteamOS上均有明显帧率提升。
使用设备原装Windows驱动时,性能差距尤为显著。例如,在《Returnal》高画质设置下,SteamOS平均帧率可达33 FPS,而Windows下仅为18 FPS。即使更新了华硕官方驱动,Windows性能有所改善,但除《Homeworld 3》的低画质测试外,其余游戏的帧率仍比SteamOS低8%至36%。
尽管SteamOS需要通过Proton翻译层运行Windows游戏,但Valve多年来对Proton效率的持续优化及对Linux Mesa图形驱动的改进显著提升了图形性能。同时,SteamOS相比通用操作系统Windows具有更少的系统后台开销。微软已意识到这一问题,并宣布未来针对掌机的“Xbox Experience”模式将通过减少后台活动来提升游戏帧率。
Classic Macintosh emulator - About
Official QEMU mirror. Please see https://www.qemu.org/contribute/ for how to submit changes to QEMU. Pull Requests are disabled. Please only use release tarballs from the QEMU website. - docs: define policy forbidding use of AI code generators · qemu/qemu@3d40db0
An innovative microgrid and solar power system in the town of Adjuntas, Puerto Rico maintained electricity during April's island-wide blackout. Puerto Rico's solar microgrids are increasingly becoming a solution to the island's endless blackouts and aging grid. Meanwhile, the U.S. is redirecting funding away from solar toward grid fixes.
Build and share AI-powered apps directly in Claude. Update: Now available on Team and Enterprise plans. (July 31, 2025) Update: Upload PDFs, images, and more into your artifact.
In 2024 and 2025, I served for six months as an international volunteer on a first-person view attack drone team in the Armed Forces of Ukraine. My team
Long sentences and recidivism kept prison populations high for decades, but prisons are now starting to empty.
埃塞俄比亚自学成才的开发者 Bereket Engida 创立的开源认证框架 Better Auth 成功获得约 500万美元种子轮融资,投资方包括 Peak XV(前红杉印度)、Y Combinator、P1 Ventures 和 Chapter One。这是 Peak XV 首次直接投资非洲创始人。
Better Auth 是一个基于 TypeScript 的开源认证库,旨在简化开发者管理用户认证的流程。与 Auth0、Firebase 等托管服务不同,其核心优势包括:
Bereket Engida 在埃塞俄比亚自学编程,曾从事远程开发工作。他在实践中发现现有认证工具存在功能局限、定制困难及规模化成本高昂等问题,因此着手构建更灵活的解决方案。Better Auth 的初始版本完全由他在埃塞俄比亚开发完成。
产品于2024年9月发布后迅速获得开发者关注:
Better Auth 通过开源模式推动认证工具的灵活性与数据主权,其成功也反映了非洲科技创始人在全球基础设施领域的发展潜力。
How we made our AI code reviewer stop being so noisy
New in LM Studio 0.3.17: Model Context Protocol (MCP) Host support. Connect MCP servers to the app and use them with local models.
Join The Offline Club's global movement towards a more humane world. We host offline events and retreats for you to unplug and connect, community-style.
IP addresses break, dial keys instead. Modular networking stack in Rust. - n0-computer/iroh
Journavx was approved this year. Why did it take so long to develop?
这篇文章重新审视了Donald Knuth关于“过早优化”的著名论文,旨在澄清其原意并提供现代视角。
核心误解澄清:
人们通常引用Knuth的“过早优化是万恶之源”来反对进行微小的代码优化。然而,原文的上下文是在讨论用goto语句实现高效的多集合(multiset)数据结构,其核心论点是:在已通过基准测试确认代码关键部分后,针对该部分的合理优化(如循环展开)是值得且必要的。
关键代码示例与性能分析: 作者通过构建一个计数多集合(counting multiset)来说明不同实现方式的性能差异:
insert_1):使用goto的线性搜索。对于小数据集,其性能优于基于std::map的O(log n)方案(在约340个元素以下时更快)。insert_2):通过“乐观插入”(先插入元素再搜索)减少循环条件判断,性能提升约13.5%。insert_2a):在insert_2基础上进行循环展开,性能再提升约7%。尽管提升幅度看似不大,但Knuth认为在关键代码中这并非“过早优化”。insert_0):将循环改为递减至零,利用特定CPU指令减少一次比较操作。在小数据集上表现最佳,但现代编译器(如Clang、GCC)有时无法完美优化此模式,需手动汇编实现。现代性能考量与基准测试: 作者通过基准测试发现:
std::find(内部进行了循环展开)在各种规模下综合表现优异。而使用哈希表(如std::unordered_map或更快的平面哈希表)在数据量极小时即可超越所有线性搜索方案,是更合理的数据结构选择。核心结论:
std::find);对于需要集合数据结构的场景,应尽早考虑使用哈希表等高效数据结构,而非过度优化线性搜索。Libxml2, an XML parser and toolkit, is an almost perfect example of the successes and failures [...]
Up until today, all capable generative image editing models were only available as proprietary tools.
AI Dataset Generator – Create realistic datasets for demos, learning, and dashboards - metabase/dataset-generator
An admittedly partial - but deeply troubling - case against America's most brazen medical grift, involving Chinese knockoffs, subscription traps, and prescription gas station sex pills.
Apptainer is an open source container platform designed to run complex applications on high-performance computing (HPC) clusters in a simple, portable, and reproducible way.