Claude Code Mods: What They Are + Best Mods to Try
Claude Code mods are TypeScript functions in plugins that rewrite prompts, tool calls and UI. How they work, the best mods to try, and how to vet one first.
Sumus 博客
关于 AI 队友的使用指南、工作方法与产品新进展。
Claude Code MCP: Setup, Best Servers & New MCP Tasks
Claude Code MCP setup in one command: claude mcp add for HTTP, stdio and JSON servers, scopes, the best MCP servers, and the new MCP Tasks beta (v2.1.304+).
Claude on Google Cloud (Anthropic-Managed): Setup & Pricing
Claude on Google Cloud (Anthropic managed) went GA Oct 22, 2026. Setup steps, IAM roles, pricing, and Anthropic-managed vs Google-managed compared.
Claude Auto Mode Is Now Default: What It Does & How to Turn It Off
Claude auto mode becomes the default on Pro, Max and Team next week. What it does, how it differs from Claude Code auto mode, and how to switch back to Manual.
Claude Security Now Scans Continuously: What Changed (Oct 2026)
Claude Security now scans continuously, tracks fixes and spans 8-repo projects. What changed on Oct 21, 2026, how it works, pricing and the $25k credit offer.
Nemotron 3.5: NVIDIA's Open Agent Models (Super VL, Bolt) Explained
Nemotron 3.5 is NVIDIA's open agent model family: 120B Super VL with 1M context, Bolt for 24 GB PCs, Lightning. Specs, benchmarks, hardware, how to run.
Claude Usage Limits: Pro & Max Caps Explained (Oct 2026)
Claude usage limits reset every 5 hours plus a weekly cap. Pro, Max 5x and Max 20x limits, the Oct 2026 Max A/B test, and what to do when you hit them.
Recursive Flint: The 40B Local Coding Model, Explained
Recursive Flint is an open 40B MoE coding model with 2.5B active params: 74.2% SWE-bench Verified, 125+ tok/s on a MacBook. Specs, setup, local vs cloud.
DeepInfer: DeepSeek's Open-Source Inference Engine, Explained
DeepInfer is DeepSeek's Apache 2.0 inference engine behind V4.1 Flash: 26K decode tok/s per GPU, SM100 or Ascend 950, 128-GPU unit. Features, install, license.
MOSS-Image: OpenMOSS's 5B Text-to-Image Model for Readable Text
MOSS-Image is OpenMOSS's open 5B text-to-image model: 96.3% long-text coverage, #2 behind GPT Image 2 in blind tests, Apache-2.0. How to run it.
Inception Mercury 3 是什么：速度、价格、跑分与 API（2026）
Inception Mercury 3 是扩散式 LLM，输出速度 1,068 tokens/s，1M 上下文，首发价每百万 token $0.10/$0.35。参数、跑分、API 用法与限制一文看懂。
Claude Fable 5.5 发布时间：10 月 28 日仍未发布，最新消息汇总
10 月 27 日“Fable 5.5 发布日”传闻再次落空。截至 10 月 28 日，Anthropic 仍只列 Fable 5.1。泄露、预计 $10/$50 价格与 3 个上线信号。