← 所有主題指南← ALL TOPIC GUIDES
AI 編程與開發工具AI Coding & Developer Tools
全站 AI 編程文章總覽:coding agents 與開發工具實測、上下文工程、token 成本,以及把代理工作流帶進生產環境的做法。Every post here on AI coding: coding agent and developer tool reviews, context engineering, token economics, and running agent workflows in production.
主題簡介ABOUT THIS TOPIC
這個主題收集全站與 AI 編程相關的文章:Claude Code、Cursor 這類 coding agents 與開發工具的實測筆記,以及把它們放進日常開發流程的具體做法。
文章回答的問題都很實際:上下文該怎麼安排、系統提示要留多長?同一個 prompt 在不同模型上成本差多少?代理改完的程式碼要經過哪些檢查才能交給團隊?
想快速上手,先讀下面的必讀精選;完整時間線收錄此主題全部文章,新文在前。
This hub collects everything on the site about AI coding: hands-on notes on coding agents such as Claude Code and Cursor, and the working habits that make these tools useful in daily development.
The posts answer practical questions: how to arrange context and how long a system prompt should be, what the same prompt costs on different models, and what checks agent-written code should pass before it reaches the team.
Start with the must-read picks below; the full timeline lists every article in this hub, newest first.
必讀精選MUST READ
6Claude 5 世代上下文工程新規則:系統提示砍八成
Anthropic 於 7 月 24 日發表 Claude 5 世代模型的上下文工程指南:Claude Code 系統提示砍掉逾 80%,六組新舊做法對照,並更新 CLAUDE.md、Skills 與 /doctor 的建議。
閱讀 ↗把 Claude Code 放到備用 Mac:隔離 Agent 工作站的價值與安全界線
讓 Claude Code 常駐備用 Mac、從手機或主力機遠端交辦任務,是 HN 熱門的隔離工作站做法。本文整理指南的 12 步設定要點(乾淨帳號、SSH、防休眠、剪貼簿同步、Tailscale)、與 Container 的取捨,以及高權限配置的安全界線。
閱讀 ↗AI Token 計費陷阱:同一段文字,不同模型收費差 2.65 倍
Token 不是長度單位,而是供應商 tokenizer 的產物。同一段文字在不同模型可能產生 2.65 倍 token 差異,而輸入格式(如 HTML vs Markdown)更可造成 21 倍差距。本文拆解 tokenization 如何影響成本,並提供實務建議。
閱讀 ↗Cursor 花了一年才搞懂:雲端代理最難的不是模型,是環境
Cursor 回顧一年來打造雲端代理的工程教訓:最難的不是模型而是環境——開發環境本身就是產品、可靠性要靠任務編排解決,架構上把代理與基礎設施解耦,是本地代理上雲最值得借鑑的三條經驗。
閱讀 ↗AI 軟體工廠的關鍵不是代理,而是五道閘門
從 Sentry、Stripe、Spotify 的公開架構拆解:先建閘門再放代理,才能讓程式碼產出被團隊吸收。
閱讀 ↗OpenRouter MCP:讓 Coding Agent 用即時價格與評測選模型
OpenRouter 官方 MCP server 把即時模型目錄、Artificial Analysis 與 Design Arena 排名、價格、文件與測試推論交給 coding agent:十一個工具只有一個計費,OAuth 專屬 key 自帶七天效期與十美元上限。本文拆解接入方式、文中範例流程與安全邊界。
閱讀 ↗
Anthropic's New Context Engineering Rules for Claude 5
Anthropic cut over 80% of Claude Code's system prompt for Claude 5 models with no measurable eval loss, and rewrote its context engineering guidance.
READ ↗Running Claude Code on a Spare Mac: The Value and Security Boundaries of an Isolated Agent Workstation
Learn how to set up a dedicated Mac for Claude Code to safely run AI agents, with practical advice on permissions, credential isolation, and recovery.
READ ↗LLM Tokenization: Why the Same Text Costs 2.65x More on Different Models
Token counts vary wildly across models. Learn how tokenization affects cost, why format matters more than model choice, and how to measure effective price.
READ ↗Cursor's Year-Long Lesson: The Hardest Part of Cloud Agents Isn't the Model, It's the Environment
Cursor shares hard-won lessons from building cloud agents: environment fidelity over model strength, Temporal for reliability, and decoupling agent logic from machine state.
READ ↗The Gates Come Before the Agent Fleet
A software factory is five gated stages, and the agent is the cheap part — review capacity is the real constraint.
READ ↗OpenRouter MCP: Live Model Data for Smarter Agent Decisions
OpenRouter's MCP server gives coding agents live model data, benchmarks, pricing, docs, and test inference. Eleven tools, only one billed; the OAuth key has a seven-day expiry and a ten-dollar cap.
READ ↗
完整時間線
179 篇文章展開全部 179 篇文章
2026 / 170 篇
- 把面試排程交給 agent:Amazon Connect Talent 想解決的招募瓶頸
- Muse Spark 把「想久一點」變成可調參數:多代理推理與思考時間懲罰的取捨
- 模型不聽話時,該怎麼記錄、調查、公開?OpenAI 的 misalignment 通報框架
- 當掃描器看不見攻擊:Cloudflare 用 ML 拆解店面前的惡意 JavaScript
- 把商品標籤交給小模型:SageMaker serverless 客製化的取捨
- 把「探索」當成一種工程紀律:Christina Koch 與 James Manyika 對談裡,產品團隊該帶走的三件事
- DevFest 2026 回歸:開發者如何從 800 場實體活動中挑出對自己有用的 agentic AI 內容
- 把電子郵件拆成三十個小模型:Fyxer 如何讓 AI 助理值得信賴
- 把模型參數移出程式碼:用 OpenRouter Presets 管理 LLM 設定
- 把測試證據放進開發流程:Cognition 用 GPT‑6 Astra 讓 Devin 自己驗證成果
- 把賽前準備變成可查詢的流程:Google Search 三個備賽切入點
- AI 軟體工廠的關鍵不是代理,而是五道閘門
- 把 diff、終端機與瀏覽器收進同一個視窗:Copilot app 對審核流程的實際改變
- 把終端機交給任何模型:OpenRouter 的 Shell 工具如何改變代理的執行邊界
- 挑選 AI 語音代理平台:先拆解成本與部署層級,再談功能
- FDE 的真正價值:不是幫你部署,而是幫你學會自己營運
- 2026 年 AI Agent 結構化資料擷取工具挑選指南:先看層級,再看可驗證性
- 支援烏克蘭獨立新聞:OpenAI 與 WAN-IFRA 聯手推動 AI 轉型
- 自動化的早期足跡:從 ATE 資料集看 AI 工具的真實樣貌
- Claude Code 不只是終端機:artifacts、auto mode 與 /design 如何把它變成協作平台
- GitHub Copilot 的九月雙重調整:帳務先付後用,體驗預設全開
- MCP 2.0 無狀態化:一次請求取代兩段式握手,代理工具基礎設施補上最後一塊
- Meta Muse Spark 1.3:把「會問問題、知道極限」的代理行為當成主打功能
- 從三天到三小時:ATV Big Air Tour 如何用 ChatGPT Work 管理巡迴業務
- 用 Amazon Bedrock 偵測儀表板內容故障:AWS 團隊的實戰架構
- 2026 年 Claude Code 與 Codex 實測推薦:12 個 MCP Server 完整比較
- OpenAI 終止向 Cursor 供應模型:開發者與產品決策者的啟示
- Haiku R1/beta6 出爐:25 週年後的回歸之作
- Claude for Teachers:把教學標準和備課時間,還給老師
- 生成式 AI 落地:從商業需求出發,而非追逐熱潮
- AI Mode 旅遊規劃升級:追蹤票價、用點數訂房、直接完成飯店預訂
- AWS 收購 DuckLabs:DuckDB 母公司加入 Amazon,開源授權不變
- 用 Claude 直接查詢 Anthropic Economic Index:AI 對工作的影響,現在可以問了
- 從 11x 的 1,100 萬次請求看 AI 外展的關鍵:自動化研究簡報
- AI Token 計費陷阱:同一段文字,不同模型收費差 2.65 倍
- DuckDB v2.0 預覽:新解析器、新儲存格式與伺服器模式
- Firecrawl Developer Index:為 Coding Agents 而設的檢索層
- Go 1.27 登場:泛型方法、json v2 與後量子加密
- Stampli 用 ChatGPT Work 和 Codex 把產品上市時間從數月縮短到六週
- Cursor Origin 上線:當 code hosting 開始為 agent 重新設計
- Copilot canvases:把 agentic workflow 從聊天捲動搬上可審視的畫布
- 追獵十六年的 SQLite bug:Tailscale 的生產環境鑑識
- Grok 4.6 評測:61 分重返前沿,代理任務變主戰場,價格不變
- DeepSeek 開源 Agent Harness:一切皆外掛的開發者預覽
- 用 Cloudflare Access 一次點擊,保護所有內部 vibe-coded 應用
- 把試算表變成互動小工具:Google Sheets canvas 的實用切入點
- Docker Sandboxes:用 microVM 給 AI 編碼代理圈出安全區
- 從輔助到執行:企業如何讓 AI 真正動手做事
- Mojo 1.0 正式登場:給 AI 時代的穩定系統語言
- Claude Sonnet 5:把 Opus 級能力帶到 Sonnet 價格帶
- 2026 年值得安裝的 ChatGPT 插件:把對話變成工作台
- Edge 終結 Manifest V2:uBlock Origin 時代收尾
- Oracle 劃紅線:AI 生成的內容不得進入 OpenJDK
- Cloudflare AI Search:為你的 Agent 加上一個專屬搜尋引擎
- FFmpeg 9.0「Lei」發布:swscale 重寫與更安全的預設
- Chrome 兩個版本修復 1,072 個漏洞:AI 撐起整條資安工具鏈
- GitHub 堆疊 PR 公開預覽:把大改動拆成一疊小 PR
- Google Play 年齡訊號 API 全球推廣:家長一處設定,App 自動調適內容
- Keychron 開源滑鼠韌體 ZGM:把 QMK 的開源文化帶進遊戲滑鼠
- Anthropic 與 Cognizant 擴大合作:把 Claude 帶進企業的關鍵在於「領域知識」
- Ruff 0.16:預設規則從59條拉到413條
- Claude 5 世代上下文工程新規則:系統提示砍八成
- Claude Opus 5 實測:接近 Fable 5 的智慧,價格減半,但真正的亮點是「判斷力」
- Claude Opus 4.7 來了:更耐操、更會驗證,但安全閘門也變多了
- Block 開源 Buzz:團隊聊天、AI 代理與 Git 託管共用一條事件流
- Firefox 153 推出:Vulkan 影片解碼與 JPEG XL 終於上船
- 把 Claude Code 放到備用 Mac:隔離 Agent 工作站的價值與安全界線
- Kimi K3 登場:2.8T 參數、百萬 Token Context,開源模型走向長程 Agent
- Claude Code 60 秒自動續跑:一個不良功能的完整解剖
- LM Studio 推出 Bionic:開源模型專用的本機 AI Agent
- Grok Build 開源:Coding Agent 最值得讀的是 Harness,不是 UI
- Google Images 25 週年:從文字搜尋到視覺生成
- Claude 進軍創作工具:用連接器把 AI 帶進設計、3D 與音樂工作流
- Bun 1.4 改寫成 Rust:64 個 Claude 代理 11 天完成的大搬遷
- Claude Fable 5 簡評:不是更強的聊天模型,而是長任務的自主工作者
- Codex 5.6 簡評:模型分工、Sol Ultra 與更快見底的額度
- Anthropic 的 Safeguards 團隊如何為 Claude 建立多層防護
- Cognition 推出 Devin Security Swarm:代理群驗證漏洞並自動修補
- Sourcegraph Agentic Batch Changes 公測:代理自動完成大規模程式碼遷移
- Mistral 開源 Leanstral 1.5:miniF2F 滿分、每題 4 美元的 Lean 證明
- Fable 5 回歸:新分類器擋下 99% 越獄手法
- Claude Code 六月底連發:組織預設模型上線、MCP 邊界補強
- Cloudflare OAuth 全開放:自助式用戶端與零停機引擎升級
- 全端 AI 的真正意義:從 TPU 到 Gmail 的整合思維
- HP 的 OpenAI Frontier 策略:從試點到企業級 AI 的落地路徑
- OpenRouter MCP:讓 Coding Agent 用即時價格與評測選模型
- Runway Agent 2.0 不是另一個 AI 生成器,而是你的行銷營運夥伴
- AWS Lambda 推出 MicroVMs:AI 程式碼的隔離沙箱
- HTTP QUERY 成為 RFC 10008:讀取型查詢的新標準
- Patch the Planet:用 AI 幫開源維護者補洞,而不是增加負擔
- Cloudflare 臨時帳號讓 AI 代理免註冊直接部署
- Adobe 把 Firefly AI 助手推進 Premiere、Illustrator 與 InDesign
- Qwen Code 本週更新:Agent 團隊協作、持久化排程、目錄切換,以及更多實用細節
- Vercel 開源 eve:agent 框架終於有了一個標準形狀
- TensorZero 收攤:730 萬美元種子輪的開源 LLM 閘道一夜封存
- Meta Edits 加入 AI 助理並預告桌面版:把創作者留在自家生態
- Qwen Code 的 /fork、/skills 與跨專案記憶:把 Agent 的日常摩擦拆開修
- 從 Nextdoor 看 Codex:工程師不再只是寫程式,而是打造產品
- Lovable 年化營收突破 5 億美元,每週新增百萬專案
- Parallel AI 之外:四款 2026 年值得考慮的 Agentic 搜尋與資料擷取工具
- Miasma 蠕蟲再襲 Microsoft:73 個儲存庫停用,AI 編碼代理成靶
- Ollama 0.30 更新:GGUF 原生支援、NVIDIA 最高 20% 效能提升、Vulkan 預設啟用
- 微軟開源 ASSERT:把文字規格變成 AI 行為測試套件
- Cursor 花了一年才搞懂:雲端代理最難的不是模型,是環境
- Uber 對員工 AI 編碼工具設限:每月 1,500 美元上限的成本治理
- Visa 投資 Replit:讓 AI 開發平台原生內建金流
- Anthropic 開設米蘭辦公室:意大利企業與開發者的 AI 落地新據點
- Figma Make 直接編輯生產程式碼庫:設計工具跨進工程腹地
- Cognition 募資 10 億美元:估值八個月翻倍至 260 億
- 維珍航空如何用 Codex 在聖誕旺季前安全上線新 App
- Cursor 年化營收突破 30 億美元:企業客戶撐起成長,SpaceX 交易在即
- Runtime(YC P26)上線:把沙盒化 coding agents 開放給整個團隊
- NanoClaw 走紅之後:拒絕 2,000 萬美元收購,把 agent 關進容器
- Anthropic 收購 Stainless:SDK 生成工具收編為自家專用
- PwC 將 Claude 推進企業生產:從 10 週縮到 10 天的保險核保,背後是 Agentic 的落地策略
- 從「打得更快」到「想得更深」:Sea 如何用 Codex 重塑工程團隊
- 用 ChatGPT Work 把財務工作變成可審閱的資產
- ChatGPT 的普及曲線:2026 年第一季的訊號
- DuckDB 推出 Quack 協定:內嵌資料庫補上主從架構缺口
- TanStack npm 供應鏈攻擊解析:三個漏洞串出 84 個惡意版本
- Mini Shai-Hulud 蠕蟲襲捲 npm:連 Mistral SDK 與 SLSA 證明都淪陷
- OpenAI 如何安全部署 Codex:從沙箱到代理原生日誌
- Firecrawl Crawl Endpoint 實戰:從單頁 scrape 到全站 crawl 的效率取捨
- Redis 之父 antirez 開源 ds4:跑得動前沿模型的本地推理引擎
- 視覺代理比結構化 API 貴 45 倍:Reflex 的完整實測
- GPT-5.5 Instant 的系統卡透露了什麼:首次被列為高能力的 Instant 模型
- 用 Firecrawl 一個 API 呼叫搞定網頁抓取:格式、結構化資料與成本控制
- OpenAI 的「哥布林」之謎:獎勵機制如何悄悄塑造模型行為
- Ghostty 宣布撤離 GitHub:一本停機日誌終結 18 年依賴
- Anthropic 認了:Claude Code 變笨的三個原因與補救承諾
- GitHub Copilot 改用量計費:AI Credits 接棒 premium requests
- GPT-5.5 系統卡出爐:OpenAI 如何為複雜工作設計更強防護
- ChatGPT Images 2.0 的實用觀察:從精準控制到多語言排版
- Yelp 春季改版:AI 助理從問答走到訂位點餐一條龍
- 同一顆模型、兩倍差距:四款 CLI 編碼 Agent 腳手架實測
- Cursor 洽募 20 億美元:估值半年再翻倍至 500 億
- Codex 更新:從寫 Code 到幫你操作電腦
- Android CLI 與官方 Skills:Google 把代理開發帶進終端機
- OpenRouter 影片生成上線:一個 API 路由所有影片模型
- Chrome 推出 AI Skills:把常用提示詞存成一鍵執行的工作流
- 當部署者變成機器:Vercel 提出的「Agentic Infrastructure」是什麼?
- GPT-5.6 Luna 讓 Replit Free Mode 成真:模型價格效能如何打開軟體創作的大門
- Cursor 3 把 IDE 變成 Agent 工作台:從 fork 走向從零打造
- Google 將 AI 學習工具直接放進搜尋:互動圖表、練習測驗與筆記本一次看懂
- GLM-5.1 送進 Coding Plan:Z.ai 把 8 小時自主編程變成訂閱規格
- Cursor Composer 2 被抓包以 Kimi K2.5 為底:開源權重的署名難題
- ChatGPT 推出 File Library:檔案自動入庫,把對話工具變工作區
- OpenAI 推 Codex 學生方案:北美大學生免費領 100 美元額度
- Mistral 開源 Leanstral:專為 Lean 4 證明工程打造的 120B 稀疏模型
- Synopsys Converge 發表 AI 晶片設計工具:代理式驗證生產力最高五倍
- Google 全美開放 AI Mode Canvas:在搜尋裡寫文件、寫程式
- OpenAI 推出 Codex Security 研究 preview:找漏洞也幫你修的資安 agent
- Claude Code Remote Control 上線:手機遙控本機代理的新工作流
- Cursor agents 大更新:自己測試修改、自己錄下工作過程
- Wispr Flow 登陸 Android:懸浮氣泡讓語音挑戰鍵盤
- Firefox 148 的 AI 總開關:可封鎖現有與未來的所有 AI 功能
- Qodo 2.1 推出 Rule System:治好 coding agent 的團隊規範失憶症
- WordPress.com 內建 AI 助手上線:自然語言改版面、媒體庫生成圖片
- Spotify:頂尖工程師 12 月起沒再寫過一行程式碼
- MiniMax 開源 M2.5 與 Lightning:SWE-Bench 80.2%,十分之一價格逼近 Opus 4.6
- 16 個 Claude 代理寫出 C 編譯器:兩週、10 萬行、2 萬美元
- GPT-5.3-Codex 登場:OpenAI 最強 coding model 參與了自己的開發
- Mistral Vibe 2.0 登場:終端機編碼代理轉為付費,進攻企業市場
- Claude 對話即應用程式:Anthropic 推出 MCP Apps 互動連接器
- Ollama 推出 ollama launch:一行指令把 coding agent 接上本機 model
- Claude Code 新增 /security-review 與 GitHub Actions 整合:安全審查走進 agent
- GPT-5.2-Codex 登場:40 萬 token context、原生壓縮與更強工具呼叫
- Ultralytics 推出 YOLO26:去 NMS、更快的邊緣視覺模型
- Deepgram 募得 1.3 億美元、估值 13 億,收購 OfOne 進軍語音點餐
- Cursor CLI 更新:模型、規則與 MCP 管理走進終端機
2025 / 9 篇
FULL TIMELINE
183 ARTICLESOPEN ALL 183 POSTS
2026 / 174 POSTS
- Amazon Connect Talent: What AI-Led Interviews Change in Your Hiring Pipeline
- Muse Spark's Real Bet: Cheaper Pre-Training, Parallel Thinking at Inference
- A Misalignment Disclosure Process You Can Actually Copy
- Catching JavaScript That Waits for the Right Victim
- How Serverless Fine-Tuning Changes Product Tagging Economics
- What a 328-Day Mission Teaches Builders About Partnering With AI
- What DevFest 2026 Means for Building in the Agentic AI Era
- What Fyxer's 53% Draft Acceptance Rate Changes for How You Build Trustworthy AI Assistants
- Config-as-Code for LLM Calls: What OpenRouter Presets Change About Shipping
- Devin Now Shows Its Work: What Self-Testing Agents Change for Review
- Search as a Race-Prep Tool: What the Three Workflows Actually Require
- The Gates Come Before the Agent Fleet
- Copilot's Diff, Terminal, and Browser Panels: A Review Loop You Can Trust
- What a Hosted Sandboxed Shell Changes for How You Build Agentic Tools
- Choosing an AI Voice Agent Platform: What the Build vs. Buy Tradeoff Actually Costs
- FDEs That Build Your Team's Capability, Not Your Dependency
- Choosing a Structured Data Extraction Tool for AI Agents in 2026
- Supporting Independent Journalism in Ukraine with AI
- Automation’s Early Footprint: The ATE Dataset
- Claude Code Is More Than a Terminal: How Artifacts, Auto Mode, and /Design Made It a Platform
- GitHub Copilot's September Double Shift: Pay-First Billing and a Default-On Unified Experience
- MCP 2.0 Goes Stateless: One Request Replaces the Two-Step Handshake
- Meta's Muse Spark 1.3 Ships Agents That Ask Questions and Know Their Limits
- How a Two-Person Stunt Show Team Cut 3 Days of Work to 3 Hours with ChatGPT
- Detecting Silent Dashboard Failures with Amazon Bedrock
- Best MCP Servers for Claude Code and Codex in 2026 (Tried and Tested)
- OpenAI's Cursor contract wind-down after SpaceX acquisition
- Haiku R1/beta6 Arrives: Two Years of Work After Beta5
- Claude for Teachers: What It Means for Education Product Builders
- Generative AI for Business: A Practical Guide to Adoption
- Google AI Mode Adds Flight Price Tracking, Points/Miles Rates, and Hotel Booking
- AWS acquires DuckLabs: DuckDB stays open source under MIT
- Ask Claude About AI's Impact on Work: The Anthropic Economic Index Connector
- How 11x Processes 11 Million Requests to Automate Prospect Research for AI Outbound
- LLM Tokenization: Why the Same Text Costs 2.65x More on Different Models
- DuckDB v2.0 Preview: Server Mode, New Parser, New Storage
- Firecrawl Developer Index: A Specialized Retrieval Layer for Coding Agents
- Go 1.27 Lands: Generic Methods, json/v2, Post-Quantum TLS
- How Stampli Used ChatGPT Work and Codex to Cut Launch Time from Months to Six Weeks
- Cursor Origin: Code Hosting Redesigned for AI Agents
- Making Agentic Workflows Visible, Steerable, and Cost-Efficient with GitHub Copilot Canvases
- Hunting SQLite's 16-Year-Old WAL Reset Bug at Tailscale
- Grok 4.6: 61 on the Intelligence Index, Frontier Again
- DeepSeek Open-Sources Agent Harness in Developer Preview
- Secure All Your Internal Vibe-Coded Applications on Cloudflare Workers — in One Click
- Google Sheets Canvas: Turning Spreadsheets into Interactive Mini-Apps
- Docker Sandboxes: microVM Isolation for AI Coding Agents
- From Assistance to Execution: How Enterprises Are Putting AI to Work
- Mojo 1.0: A Stable Foundation for AI Systems Programming
- Claude Sonnet 5: Opus-Level Agentic Power at a Lower Price
- Best ChatGPT Plugins in 2026: Turn Conversations into a Workspace
- Edge Ends Manifest V2: What uBlock Origin Users Do Now
- Oracle Bans AI-Generated Code from OpenJDK Contributions
- Cloudflare AI Search: Give Your Agents a Search Engine for Your Data
- FFmpeg 9.0 'Lei': swscale Rewrite and Safer Defaults
- Chrome Fixed 1,072 Security Bugs, Largely with AI
- Stacked Pull Requests Arrive on GitHub in Public Preview
- Google Play Expands Age Signals API to Developers Worldwide
- Keychron's ZGM Brings Open-Source Firmware to Gaming Mice
- Anthropic and Cognizant Expand Partnership: Domain Knowledge Is Key to Enterprise AI
- Ruff 0.16 Ships 413 Default Rules, Up From 59
- Anthropic's New Context Engineering Rules for Claude 5
- Claude Opus 5: Near-Frontier Intelligence at Half the Price, with Judgment as the Real Edge
- Claude Opus 4.7: A Practical Guide for Product Builders
- Block Open-Sources Buzz: Chat, AI Agents, Git in One Place
- Firefox 153 Ships Vulkan Video Decoding and JPEG XL
- Running Claude Code on a Spare Mac: The Value and Security Boundaries of an Isolated Agent Workstation
- Kimi K3 Arrives: 2.8T Parameters, Million-Token Context, Open Models Go Long-Horizon
- Claude Code's 60-Second Auto-Continue, Dissected
- LM Studio Bionic: A Local-First AI Agent for Open Models
- Scrape Every Page: Firecrawl's Structured Output Approach to Pagination
- Grok Build Open Source: The Harness Is the Part Worth Reading, Not the UI
- Google Images at 25: From Text Links to AI Image Generation in Search
- Claude for Creative Work: New Connectors Bring AI into Design, 3D, and Music Tools
- Bun 1.4 Is Rust Now: 64 Claude Agents Ported It in 11 Days
- GPT-5.6 Prompting: Less Writing, Better Results
- Codex 5.6 Review: Model Tiers, Sol Ultra, and Faster Limit Burn
- Qwen Code Update: Auto Model Fallback and Nested Subagents for Reliable AI Agents
- How Anthropic Builds Multi-Layer Safeguards for Claude: A Blueprint for AI Product Teams
- Devin Security Swarm: Agents Verify Exploits and Ship Fixes
- Sourcegraph's Agentic Batch Changes Enters Public Beta
- Leanstral 1.5: Mistral Saturates miniF2F at $4 a Proof
- Fable 5 Is Back: The Jailbreak Other Models Replicated
- Browserbase Agents: From Per-Site Scripts to Reusable Browser Tasks
- Claude Code Late June: Org Model Defaults, MCP Hardening
- Cloudflare Opens Self-Managed OAuth to All Developers
- Full-Stack AI Explained: What It Means and How to Get Started
- HP's OpenAI Frontier Strategy: From Pilots to Enterprise AI Deployment
- OpenRouter MCP: Live Model Data for Smarter Agent Decisions
- Runway Agent 2.0: From AI Video Generator to Marketing Operations Partner
- AWS Lambda MicroVMs: Sandboxes for AI-Generated Code
- HTTP QUERY Becomes RFC 10008: What APIs Gain
- OpenRouter Unified Image API: Query Model Capabilities Before You Call
- Patch the Planet: How OpenAI and Trail of Bits Are Using AI to Help Open Source Maintainers
- Cloudflare Temporary Accounts: Agent Deploys Without Signup
- Adobe's Firefly AI Assistant Comes to Premiere, Illustrator
- Is Your Tool Truly Agent-Ready? Hugging Face Benchmarks the Full Workflow
- Qwen Code Weekly: Agent Teams, Durable Scheduling, /cd, and More
- TensorZero Archives Repo After $7.3M Seed, Winds Down
- Meta Edits Gets an AI Assistant and a Desktop Version
- Qwen Code's /fork, /skills, and Cross-Project Memory: Fixing an Agent's Daily Friction
- From Nextdoor to Codex: How Outcome Engineering Is Redefining the Role of the Engineer
- Lovable Passes $500M Annualized Revenue, 1M Projects a Week
- Beyond Parallel AI: Four Agentic Search and Data Extraction Tools to Consider in 2026
- Miasma Worm Hits Microsoft Repos, Targeting AI Coding Agents
- Microsoft ASSERT Turns Text Specs into AI Behavior Tests
- Cursor's Year-Long Lesson: The Hardest Part of Cloud Agents Isn't the Model, It's the Environment
- Uber Caps Employee AI Tool Spend at $1,500 a Month
- Visa Invests in Replit to Build Payments Into AI Coding
- Anthropic Opens Milan Office: What It Signals for AI Builders and Enterprises
- Figma Make Now Edits Your Production Codebase
- AI Coding Startup Cognition Raises $1B at $26B Valuation
- How Virgin Atlantic Used Codex to Ship a New App Before Christmas—Without the Bugs
- Cursor Passes $3B Annualized Revenue as SpaceX Deal Looms
- Runtime (YC P26): Sandboxed Coding Agents for Whole Teams
- NanoClaw Turned Down $20M to Keep Building Sandboxed Agents
- Anthropic Buys Stainless: SDK Generation Goes In-House
- PwC Puts Claude into Production: Underwriting Cut from 10 Weeks to 10 Days
- From Typing Faster to Thinking Better: How Sea Uses Codex to Reshape Engineering Teams
- Search Quality Shapes RL Outcomes: Why Your Agent's Backend Matters More Than You Think
- Turning Finance Work into Review-Ready Assets with ChatGPT Work
- ChatGPT Adoption in Q1 2026: Signals for Product Builders
- DuckDB Ships Quack, Its Own Client-Server Protocol
- Inside the TanStack npm Supply-Chain Compromise
- Shai-Hulud npm Worm Hits Mistral SDK; Provenance No Defense
- How OpenAI Deploys Codex Safely: Sandboxes, Rules, and Agent-Native Logs
- Mastering Firecrawl's Crawl Endpoint: From Single Page Scrape to Full Site Crawl
- antirez's ds4: A Local LLM Inference Engine Built for Metal
- Reflex Benchmark: Computer Use Costs 45x More Than APIs
- GPT-5.5 Instant System Card: What It Means for Product Builders
- Where the Goblins Came From: How Reward Signals Quietly Shape Model Behavior
- Ghostty Quits GitHub After a Month of Daily Outages
- Anthropic Postmortem: Why Claude Code Got Worse for a Month
- GitHub Copilot Moves to Usage-Based Billing With AI Credits
- GPT-5.5 System Card: What Product Builders Need to Know
- ChatGPT Images 2.0: What Builders Should Know About Text, Multilingual Layouts, and Realism
- Yelp's AI Assistant Can Now Book and Order in One Chat
- Deep Max: Why Exa Pushes Agentic Search Speeds to New Heights
- Same Model, 2x Gap: Benchmarking Four CLI Coding Agents
- Cursor in Talks to Raise $2B at a $50B Valuation
- Codex Update: From Writing Code to Operating Your Computer
- Android CLI: Google's Terminal-First Bet on Coding Agents
- Video Generation Is Live on OpenRouter: One API to Route Every Video Model
- Chrome Gets AI Skills: Prompts You Reuse on Any Page
- Agentic Infrastructure: What Vercel's Three-Layer Model Means for AI Builders
- How GPT-5.6 Luna Powers Replit's Free Mode: Model Economics in Action
- Cursor 3 Turns the IDE into an Agent Workspace Built From Scratch
- Google Adds AI Study Tools to Search: Interactive Visuals, Practice Quizzes, and Notebooks
- GLM-5.1 Arrives for Coding Plan Subscribers: Z.ai Makes 8-Hour Coding Runs a Product Spec
- Cursor Composer 2 Caught Building on Kimi K2.5 Weights
- ChatGPT File Library: OpenAI's Persistent Workspace Play
- OpenAI Gives Students $100 in Codex Credits
- Mistral Open-Sources Leanstral, a Lean 4 Proof Agent
- Synopsys Rewires Chip Design With AI Agents and Ansys
- Google Opens AI Mode Canvas to All US Users
- OpenAI Ships Codex Security in Research Preview: A Security Agent That Finds and Helps Fix
- Claude Code Remote Control: Phone Drives Your Local Agent
- Cursor's Big Agents Update: Coding Agents That Test Their Own Work and Record It
- Wispr Flow Lands on Android: Voice as the Primary Interface
- Firefox 148 Adds an AI Kill Switch for All Its AI Features
- Qodo 2.1 Adds a Rule System to Cure Coding Agent Amnesia
- WordPress.com Builds Its AI Assistant Into the Editor
- Spotify's Top Developers No Longer Write Code, Thanks to AI
- MiniMax M2.5: Open Coding Weights at a Tenth of Opus Cost
- 16 Claude Agents Built a C Compiler: 100K Lines, $20K
- GPT-5.3-Codex Arrives: OpenAI's Strongest Coding Model Helped Build Itself
- Mistral Vibe 2.0: Terminal Coding Agent Goes Paid
- Anthropic's MCP Apps Turn Claude Chat Into an App Runtime
- Ollama Ships ollama launch: One Command to Wire Up Local Coding Agents
- Claude Code Adds /security-review and GitHub Actions Integration
- GPT-5.2-Codex Arrives: 400k-Token Context, Native Compaction, Better Tool Calling
- Ultralytics YOLO26: NMS-Free Vision AI Built for the Edge
- Deepgram Raises $130M at $1.3B, Buys OfOne
- Cursor CLI Update: Model, Rules and MCP Control in Terminal
2025 / 9 POSTS
- Google open-sources Gemini CLI, a free terminal AI agent
- Altman: Meta's $100M offers failed to poach OpenAI talent
- Gemini 2.5 goes stable: Pro and Flash reach GA
- iPadOS 26 brings Mac-style windowing to the iPad
- Apple unveils Liquid Glass and year-based OS names
- Cursor maker Anysphere lands $900M at $9.9B valuation
- Mistral Code brings agentic coding to enterprises
- Windsurf says Anthropic cut its direct Claude access
- Google AI Edge Gallery runs open AI models offline on phones