uptonking

By the end of the year we should have: Next.js 17 React Router 9 Nuxt 5 Astro 8 SvelteKit 3 Remix 4 TanStack Start 2 SolidStart 2 Never in the history of JavaScript frameworks has the frontier been so multipolar. The benefits to full-stack developers, routing solutions, reactivity systems, and meta-framework innovation / interoperability are ramping up. build accordingly

Today, we’re announcing Bonsai 27B: the first 27B-class model to run on a phone. Bonsai 27B is the new multimodal flagship of the Bonsai family. Based on Qwen3.6 27B, it brings a new capability tier to local AI: multi-step reasoning, structured tool use, long-context workflows, and coherent agentic loops. Until now, models in this class have been impractical to deploy locally. A 27B model occupies roughly 54 GB in 16-bit precision, and even a strong 4-bit build is around 18GB - too large for a phone and for most laptops. Bonsai 27B changes that. It comes in two variants: • Ternary Bonsai 27B: 5.9 GB, 1.71 effective bits per weight, optimized for laptop-class quality. • 1-bit Bonsai 27B: 3.9 GB, 1.125 effective bits per weight, optimized for phone-class footprint. Everything is open-sourced today under the Apache 2.0 license.

Today we’re releasing Reflect Open, a complete rewrite of Reflect for the age of AI. It’s local-first and open source, and every note is stored as a Markdown file on your computer.

推荐了解关于 computer use 的现状。Ethan Mollick 刚在 X 上说 Codex 的 computer use "在 PC 上变得非常好"——看着光标在幽灵的控制下移动，"让你在感官上意识到一个脱离了身体的智能握着鼠标和键盘能做多少事。"与此同时 Pi 刚发布了 computer-use v0.4.3，加了幽灵光标和并行子 agent。而 Claude 还没有自己的 computer use 故事。 Computer Use 的三方竞赛 Codex：第一个"感觉像真的" Ethan Mollick 在 7 月 13 日发表了这个评价——Codex 的 computer use 在 PC 上已经到了一个阈值。他最新的实验是让 GPT-5.6 Sol 在 Codex 里控制他的电脑，玩 Slay the Spire 2 的每日挑战——5 小时的自主游戏，做出复杂的游戏决策，赢了。 这条推文拿了 2,266 赞。但 Mollick 的重点不是"AI 很强"——是 "AI 公司很不会解释他们的产品到底能做什么"。 "这不是一个'你太早了还没看到'的帖子——这是一个'AI 公司在清晰地解释自己的系统能做什么上做得很差'的帖子。" Pi：三步追赶 Pi 的 computer use 更新到 v0.4.3（7 月 12 日）。三个变化： • 幽灵光标——在屏幕上显示 CUA 正在操作的位置。之前 computer use 最大的痛点——"它是不动还是在想？"——被修好了。 • 并行子 agent——多个子 agent 可以在不同应用上安全地并行操作。 • 批量优化——延迟改善近 50%，Microsoft 应用支持更好。 Pi 的定位是开源、可定制、带 harness 的 computer use——和 Codex 的端到端封闭体验形成对比。 Claude：缺席者 截至 2026 年 7 月，Claude 还没有 native computer use。Claude Code 有终端——terminal computer use——但没有桌面级的、像 Codex 和 Pi 那样的"控制你的鼠标和键盘"。Anthropic 的 UI 故事仍然以聊天为界。 哪一条路会赢 三条路在浮现： | 平台 | 方式 | 风格 | |------|------|------| | Codex | 端到端——"你的电脑 + 我们的模型" | 封闭但体验好 | | Pi | 开源 harness + 你自己的模型/API | 开放但需要配置 | | Claude | 终端级——还在聊天框里 | 安全但限制多 | Computer use 是 coding agent 的下一层——不只是写代码，而是使用软件。谁先解决 reliability + safety + latency 的三角，谁就定义了下一个 agent 界面。 来源：@emollick, Jul 13, 2026 · @injaneity pi-computer-use v0.4.3, Jul 12 · @thsottiaux CLIProxyAPI, Jul 12 #ComputerUse #Agent #Codex #Pi

Pi社区开源了codex没有开源的computer-use实现。 深入分析了一下仓库，眼前一亮！

New TIL: Using uvx in GitHub Actions in a cache-friendly way I finally found a recipe that I like for running `uvx tool-name` in GitHub Actions without downloading a fresh copy of the package every time til.simonwillison.net/github-actions…

introducing namethatui.com a dictionary for ui things you can see but can't name made it because i'm primarily a designer, and my biggest resistance was always knowing what things are called when prompting my agents it learns as people use it: every search teaches the site new words, and the built-in pocket dictionary grows with it give it a try and let me know what you think can't find something? dm me and i'll add it i want this to be the lowest resistance resource you have you can just build things

the problem with "spec driven development" is that most software can't be spec'd up front software is a creative act, where you figure out what you're building as you build it you need to get your hands dirty in the details, and react to incremental versions it's telling that all the examples of spec driven development are sorting a list or porting thoroughly tested code (like a js runtime or browser engine), which are the exception, not the rule. the vast majority of software doesn't have a spec – or if it does, the spec was created *after*

An in-app browser is the right primitive. We went one step further in @OpenDesign over a month ago. Open any site in the built-in browser, screenshot it, and rebuild it into real, editable files with any agent you like. Plus 80+ curated design sites built in. Demo 👇

I don't like the functional options pattern in Go, and I'm pretty disappointed that it ended up in the standard library with the new json/v2 package. An honest, plain Options struct is much better.

Want to run a 70B+ model but don't have an 80GB GPU? Mesh LLM distributes inference across the devices you actually have. The architecture is incredibly smart: → If a node can fit the model, it runs locally → If not, it routes to a peer that can → If the model is too big for any single box, it uses "Skippy stage splits" to shard the model by layer across your hardware Open source in Rust. Worth checking out

Announcing LlamaCoder v4 – generate apps in 1 prompt! • Rebuilt it around GLM 5.2 & its strengths • Migrated to using Base UI w/ @shadcn • Improved parsing, planning, & design of apps • New WebAssembly-based renderer 100% free, open source, and powered by @togethercompute.

Working on #MobX 7, a major clean-up removing things that are no longer needed. E.g. non-proxy support. I suspect that APIs like spy, observe, intercept, dehancer also receive very little usage. Does anybody use those? Feel free to chime in! #issuecomment-4955238021" target="_blank" rel="nofollow noopener">github.com/mobxjs/mobx/pu…

老子辛辛苦苦的蹬到83%，你告诉我延长到19号。又不Reset Weekly Plan。真特么的没见过A\这么傻逼的公司。

When TanStack Table V9 launches as stable (really soon™), it will be offering a record number of adapters. Preact, Ember, and Alpine are new! And we're also still exploring potential Marko and Qwik V2 adapters. Writing our libraries in 95%+ vanilla JS makes this maintainable.

看了 agent-device 非常不错，它是 Callstack 开源的移动端自动化 CLI。 它给 coding agent 操作真实 iOS/Android app 的能力：打开 app、读 accessibility snapshot、按语义 ref 点击/输入、截图、录屏、采日志、抓性能、把探索过程保存成可 replay 的 .ad 脚本。官方定位是 “Mobile app verification for AI agents”，不是传统测试框架替代品。 除了 Native App，使用 Expo 和 Flutter 构建的应用也能操作，真机和模拟器上运行的也可以操作。 GitHub 地址: github.com/callstack/agen…

正确的codex 级别切换效果…

原来 GNOME对应提OSX风格，KDE 对应的是WINDOWS，我就说Fedora44 怎么长得那么像黑苹果

what are the best open source repos (in terms of quality) that have been written exclusively, or almost exclusively, by ai (bun excluded)

你现在用的 Windows，后台可能正运行着 200 多个系统服务 其中相当一部分，要么你根本用不上，要么一直在后台收集数据、推送广告、追踪位置 有人做了一个完全免费、开源的工具，专门把 Windows 这几年那些反用户设计尽可能一键关掉 Cortana、Copilot、开始菜单广告、遥测数据、位置追踪、莫名其妙的推荐应用…… 统统可以关闭 一个 8MB 左右的 .exe 文件，不需要安装，点几下就能完成 它叫 optimizerDuck 这是我目前见过最接近： 把 Windows 调整成它本来就应该有的样子的工具 它支持： → 35+ 项系统优化，覆盖性能、隐私、GPU、电源、系统服务和操作体验 → 200+ 个 Windows 服务可自由开关，并标注低、中、高风险等级 → 针对 AMD、NVIDIA、Intel 显卡提供专门优化 → 一键清理系统预装软件，而且删除前可以先预览 → 每次操作都会生成回滚文件，改错了还能恢复 → 修改系统前会主动提醒你创建还原点 没有安装器 没有 没有广告 没有所谓高级会员版。 GPL v3 开源协议。支持完全离线运行 兼容 Windows 10 / 11 x64