AI Tracker下載 App

程式開發新聞，
為開發者整理。

各家 AI 公司最新的程式開發動態：發表、調價和停用，並附上對你的影響。

模型· 1 則討論

Artificial Analysis 發布 Coding Agent Index v2.0；Claude Code 搭配 Opus 5.5 以 61.7 領先

Artificial Analysis 推出 Coding Agent Index v2.0，這項指標會將模型與其 agent harness 一併評分，涵蓋軟體工程、終端機使用和程式碼庫問答等 290 項任務，權重均等。分數由 Artificial Analysis 自行評定，未經獨立驗證。GitHub Copilot· 程式開發

Copilot 使用量指標 API 現在會顯示自訂技能、代理和 MCP 伺服器的名稱

GitHub 表示，Copilot 使用量指標現在會顯示組織建立的技能、自訂代理、Model Context Protocol (MCP) 伺服器、斜線指令和外掛程式的設定名稱，而不再將它們歸類為 other 或 custom。這項更新延續了 9 月推出的代理式 CLI 自訂指標。Anthropic· 程式開發

Claude Code 2.1.309 修正沙箱與權限檢查繞過問題，並調整 `sed -i` 和 `rm` 的處理方式

根據發布說明，Claude Code v2.1.309 於 2026 年 10 月 22 日發布：Anthropic· 程式開發· 1 則討論

Anthropic 建議將子代理整合進主代理迴圈，並推出遷移工具包

在一篇日期為 2026 年 10 月 22 日的文章中，Anthropic 的 Applied AI 團隊表示，在 Claude Platform 上的數個企業部署案例裡，使用 skills 和工具搜尋的單一代理，在品質上往往勝過單一提示設計和子代理設計，成本與延遲也經常較低。該團隊表示，有位客戶從多代理框架改為單一迴圈後，執行速度約提升 2 倍。這些是 Anthropic 自行觀察的結果，並非已公開的測量數據。Anthropic· 模型· 2 則討論

Anthropic 表示，Claude 為 123 個開源函式庫撰寫了形式化證明，證明它們不會崩潰

Anthropic 表示，該公司與 Theorem 合作，使用 Claude 在組合語言層級（RISC-V）證明 100 多個真實世界開源函式庫的安全性質；Anthropic 稱這是 AI 首次大規模用於在該層級形式化驗證正式環境軟體。Anthropic· 程式開發· 1 則討論

Anthropic 將 Agent Skills 標準捐贈給 Agentic AI Foundation

Anthropic 於 10 月 22 日宣布，將 Agent Skills 捐贈給 Agentic AI Foundation (AAIF)。Anthropic 先前也將 MCP 捐贈給了這個基金會。Anthropic 表示，自 2025 年 12 月以來，已有超過 70 個代理平台整合此標準，包括 ChatGPT、Cursor、GitHub Copilot 和 VS Code。GitHub Copilot· 程式開發

GitHub Copilot CLI 1.0.103 淘汰舊版 `--remote` 旗標，並擴大沙箱支援

根據 2026 年 10 月 22 日發布說明，GitHub Copilot CLI 1.0.103 更新如下：OpenCode Go· 程式開發

OpenCode v1.19.4 upgrades its bundled Bun runtime for security fixes

OpenCode v1.19.4 upgrades the Bun runtime bundled with the tool to version 1.4.4. According to the release notes, that version includes security fixes for Brotli, TLS, HTTP/1.1, HTTP/2, HTTP/3, WebSocket and fetch, and updates the root certificates to NSS 3.131.1. The release notes list no other changes.程式開發

Amp lets you self-host orbs, free for individuals and teams

Thorsten Ball announced on X that orbs, a feature of the Amp coding agent, can now be self-hosted. He says self-hosting is free for individuals and teams. Details are in Amp's news post, "Self-hosted orbs".OpenAI· 程式開發

OpenAI publishes an official command-line tool for its API

OpenAI published openai-cli on GitHub, which it calls the official CLI for the OpenAI REST API. It is written in Go (1.26 or later to build from source), licensed under Apache 2.0, and installs with brew install openai/tools/openai or go install. Pull requests are limited to repository collaborators.Anthropic· 程式開發

Anthropic releases Claude API version 2026-10-21, with breaking changes

Anthropic's ClaudeDevs account announced API version 2026-10-21, which it calls the first official Claude API version release since 2023. Anthropic says it plans to release versions more often, so developers can handle breaking changes before a new model requires them. The older version, 2023-06-01, is not deprecated.Anthropic· 程式開發

Claude Security now scans continuously and tracks fixes

Anthropic announced on October 21 that Claude Security, in public beta on Claude Enterprise, now keeps its findings current as code changes. After repositories are connected once, Claude runs a deep first scan and then re-scans as changes land.OpenCode Go· 模型

OpenCode adds Meta's Muse Wonder model

OpenCode announced on X on October 21 that Meta's Muse Wonder model is now available in OpenCode. The post gives no further detail, such as which variant, pricing or how to select it. A minute earlier the account had posted the same announcement under the name "Meta Wonder".模型· 1 則討論

Recursive releases Flint, an open-weight local coding model

Recursive (@Recursive_SI) announced Flint on X on October 21, calling it its first model: a fast coding model meant to run on a user's own machine. The post, by Richard Socher, says it was built with the company's automated research system, the Eureka Machine, and from scratch in under four months.程式開發

Trail of Bits publishes patching-bugs and regression-hunt plugins for coding agents

Security firm Trail of Bits published two plugins that let engineers hand patch development and regression checks to coding agents. The patching-bugs plugin reproduces a reported bug, writes a fix and tests it. The regression-hunt plugin checks whether a change breaks behavior that used to work. Both grew out of Patch the Planet, Trail of Bits' joint initiative with OpenAI, and draw on its engineers' Codex sessions.OpenAI· 模型· 4 則討論

OpenAI says its models are coming to Google Cloud

OpenAI has published a post titled "OpenAI models are coming to Google Cloud". Its summary says OpenAI and Google Cloud are bringing frontier models and Codex workflows to Gemini's Model Garden, available through Google Cloud Marketplace. This is OpenAI's own account. Posts on X about the announcement say the models and Codex arrive on Google Cloud in the coming weeks.模型· 1 則討論

KRAFTON AI releases Raon-1 mini preview, an open 30B mixture-of-experts model

KRAFTON AI announced a preview of Raon-1 mini on October 20, a mixture-of-experts language model trained from scratch with about 30B parameters in total and 4B active. The weights are on Hugging Face as KRAFTON/Raon-1-mini-preview under Apache 2.0.程式開發· 1 則討論

Vercel introduces durable-harness for running coding-agent harnesses durably

Vercel announced durable-harness on X. In Vercel's words, it lets users run Pi, OpenCode, Codex, Claude Code, eve, fx or their own harness durably on Vercel, and agents automatically resume work after failures and run for hours, days or months.Anthropic· 程式開發

Claude Code 2.1.305 fixes a 2.1.304 tool-search regression and adds a Linux sandbox setting

Claude Code v2.1.305, released October 20, 2026, per its release notes:Anthropic· 程式開發· 1 則討論

Anthropic's Frontier Red Team: Claude found 493 attack-surface issues across 25 open-source projects

Anthropic published "Using Claude to reduce the attack surface of code" on Oct 20, 2026. Everything below is Anthropic's own account.
較早的新聞