Avatar 🍥

KnightLi Blog

记录并分享日常

  1. Home
  2. About
  3. Archives
  4. Search
  5. Links
    1. Dark Mode

Archives

2026 383
2025 23
2024 5
2023 9
2022 33
2021 5
2020 8

Categories

AI Tools Technical Docs Hardware AI Industry Development Tools Operations Security Updates Business Analysis

Tags

Ai-Tools AI Agent AI Coding Developer Tools Codex Claude Code Local LLM Linux Openai MCP Ubuntu ChatGPT Claude Ollama Gemini Anthropic Llama.cpp NAS Open Source GPU AI Art Gemma 4 Google Prompts Windows DeepSeek GGUF LLM Python AI Models
Security Updates

ssh-keysign-pwn (CVE-2026-46333): Linux Local Information Disclosure, SSH Host Keys, and /etc/shadow Risk

A practical review of ssh-keysign-pwn (CVE-2026-46333): impact, root cause, patch status, temporary mitigations, and operations guidance for a Linux kernel ptrace access-check race that may expose SSH host private keys and /etc/shadow.

2026-05-17
8 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Gemini Intelligence on Android: Google Is Turning the Phone into a Proactive AI System

A summary of Google's May 2026 Gemini Intelligence on Android announcement: multi-step automation, smarter Chrome browsing, Autofill, Rambler, natural-language widgets, and Android's shift toward a proactive AI system.

2026-05-17
8 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Codex Now Supports Remote Access from ChatGPT Mobile, with Access Tokens for Enterprise Workspaces

A look at OpenAI's May 14, 2026 Codex update: remote access to long-running Codex tasks from ChatGPT mobile, and Codex access tokens for Enterprise workspaces.

2026-05-17
6 minute read
中文简体 中文繁體 日本語 Español
AI Industry

Anthropic’s 2028 AI Leadership Report: The US, China, Compute, and Two Future Scenarios

A summary of Anthropic’s May 2026 essay “2028: Two scenarios for global AI leadership”: how it frames US-China AI competition, compute advantage, export controls, distillation attacks, and two possible 2028 futures.

2026-05-17
7 minute read
中文简体 中文繁體 日本語 Español
Technical Docs

LLM Architecture Evolution from 2023 to 2026: Tokenizers, Positional Encoding, Attention, MoE, Normalization, and Activation Functions

A beginner-friendly review of how LLM architecture evolved from 2023 to 2026: what tokenizers, positional encoding, attention, MoE, normalization, and activation functions solve, and why most changes focus on efficiency, long context, and inference cost.

2026-05-17
8 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Why Did Codex Usage Limits Suddenly Reset? History and Sources

A summary of why Codex usage limits sometimes reset without warning, the historical context, how users can interpret these resets, and sources such as Tibo's posts, OpenAI Status, GitHub issues, and community discussions.

2026-05-17
6 minute read
中文简体 中文繁體 日本語 Español
AI Tools

easy-vibe: A Learning Map for Vibe Coding Beginners

datawhalechina/easy-vibe is an open source learning project for Vibe Coding beginners. Through tutorials, exercises, and an advanced path, it connects AI coding, RAG, terminal tools, Claude Code, MCP, Skills, and Agent Teams into an easier starting route.

2026-05-16
6 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Anthropic financial-services: Reusable Templates for Financial Agents

anthropics/financial-services is a reference project from Anthropic for the financial services industry. It provides examples of Agents, Plugins, Skills, and MCP connectors for workflows such as investment banking, research, private equity, wealth management, fund operations, and KYC.

2026-05-16
6 minute read
中文简体 中文繁體 日本語 Español
AI Tools

DeepSeek-TUI: Turning DeepSeek V4 into a Terminal Coding Agent

DeepSeek-TUI is a terminal coding agent project for DeepSeek V4. It provides a TUI, tool calling, Auto mode, sub-agents, sandboxing, and a persistent task queue for developers who want to use DeepSeek for coding tasks from the command line.

2026-05-16
6 minute read
中文简体 中文繁體 日本語 Español
AI Industry

Why AI Data Centers Are Driving HDD Demand Again

AI training and inference do not just consume GPUs. They also keep producing checkpoints, training data, logs, and audit records. Massive cold data is making hard drives a key layer of data center storage again.

2026-05-16
5 minute read
中文简体 中文繁體 日本語 Español
AI Industry

How Did AI Agents Evolve? A Complete 2022-2026 Five-Generation Timeline

A timeline-based overview of AI Agent evolution from 2022 to 2026: from the ChatGPT chat box to tool calling, engineered workflows, Computer Use, MCP, Skills, and persistent digital workers.

2026-05-16
10 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Codex Mobile Remote Access: Use the ChatGPT App to Follow Coding Tasks on Your Mac

A practical overview of Codex mobile remote access: requirements, setup, remote controls, limitations, troubleshooting, and when it is useful for following Codex coding tasks from the ChatGPT mobile app.

2026-05-16
5 minute read
中文简体 中文繁體 日本語 Español
AI Tools

What Is ChatGPT File Library? File Storage, Limits, and Privacy Boundaries

A practical guide to ChatGPT File Library: what it stores, storage limits by plan, file limits, deletion and download behavior, Temporary Chat exceptions, and privacy considerations.

2026-05-16
5 minute read
中文简体 中文繁體 日本語 Español
Technical Docs

Running Android Apps on Ubuntu 26.04 LTS: Waydroid Setup and Practical Notes

A practical guide to running Android apps on Ubuntu 26.04 LTS with Waydroid, including setup commands, APK installation, multi-window mode, Google Play trade-offs, common issues, and suitable use cases.

2026-05-16
5 minute read
中文简体 中文繁體 日本語 Español
AI Industry

The U.S. Clears Nvidia H200 Sales: 10 Chinese Companies Approved, but Delivery Is Still Uncertain

A summary of the U.S. Commerce Department's approval for about 10 Chinese companies to buy Nvidia H200 chips: approved buyers, purchase limits, Lenovo's confirmation, pending delivery, and remaining policy variables on both sides.

2026-05-16
6 minute read
中文简体 中文繁體 日本語 Español
Hardware

How to Protect Hardware from PCB Cloning: From Marking Removal and Potting to Security Chips

A practical overview of common PCB anti-copy strategies: chip marking removal, potting, multilayer PCBs, blind and buried vias, security chips, uncommon parts, parasitic parameters, and decoy circuits.

2026-05-16
7 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Do Not Push API Keys to GitHub: A Secret-Leak Prevention Guide for AI Coding

A practical guide to API key leaks in the age of AI coding: why .env files, config files, and frontend code often expose secrets on GitHub, and what to do after a leak.

2026-05-16
6 minute read
中文简体 中文繁體 日本語 Español
AI Industry

Gemini 3.5 Pro Leaks: Google Wants Spark Agent to Win Back the AI Coding Entry Point

A concise look at the latest leaks around Gemini 3.5 Pro, Gemini Spark, and Google's AI coding products: model capability is still catching up, Agent entry points are getting more aggressive, and coding tools are becoming a key battlefield for major model companies.

2026-05-15
7 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Claude Code + Ollama Local Deployment Guide: Build a Free AI Coding Assistant with CC Switch

A practical walkthrough for connecting Claude Code to local Ollama models through CC Switch: it keeps Claude Code's agent workflow while moving inference to a local model, with clear limits around long context, large repositories, and multimodal compatibility.

2026-05-15
6 minute read
中文简体 中文繁體 日本語 Español
Security Updates

How to Check CVE-2026-42945: Nginx Rift Trigger Conditions, Version Checks, and Upgrade Advice

A practical summary of Nginx Rift / CVE-2026-42945: it affects ngx_http_rewrite_module and can be triggered by unauthenticated requests under specific rewrite configurations, potentially restarting workers; code execution is possible when ASLR is disabled.

2026-05-15
9 minute read
中文简体 中文繁體 日本語 Español
AI Tools

OpenHuman Quick Read: The Desktop Route for an Open-Source Personal AI Agent

Based on the tinyhumansai/openhuman README and official site, this article summarizes OpenHuman's positioning, installation, memory system, third-party integrations, TokenJuice compression, privacy design, and target users.

2026-05-15
7 minute read
中文简体 中文繁體 日本語 Español
Development Tools

Ghostty Docs Quick Read: Installation, Configuration, and Daily Usage Notes

Based on the official Ghostty documentation, this article summarizes its positioning, installation paths, configuration files, keybindings, themes, fonts, and Shell integration to help you decide whether it is worth replacing your current terminal.

2026-05-15
6 minute read
中文简体 中文繁體 日本語 Español
Security Updates

Dirty Frag, Copy Fail, and Fragnesia: Comparing Three Recent Linux Local Privilege Escalation Flaws

A comparison of Dirty Frag CVE-2026-43284, Copy Fail CVE-2026-31431, and Fragnesia CVE-2026-46300. All three point to page-cache writes and local privilege escalation, but their entry points, modules, mitigations, and operational priorities differ.

2026-05-15
7 minute read
中文简体 中文繁體 日本語 Español
Security Updates

Fragnesia (CVE-2026-46300): Impact and Mitigation for a Linux Kernel Local Privilege Escalation Flaw

A concise look at Fragnesia (CVE-2026-46300), a Linux kernel local privilege escalation flaw related to the Dirty Frag attack surface. The issue sits around XFRM ESP-in-TCP and shared page-fragment handling, with the risk of modifying read-only files through the page cache and gaining a root shell.

2026-05-15
5 minute read
中文简体 中文繁體 日本語 Español
AI Industry

Which industries will LLMs disrupt first? AI impact through the lens of workforce disruption

A workforce-disruption view of the industries and roles most affected by current large language models: customer support, administration, marketing, software, finance, law, education, media, consulting, medical documentation, and R&D support.

2026-05-15
7 minute read
中文简体 中文繁體 日本語 Español
AI Tools

web-video-presentation: an Agent Skill for turning articles into screen-recordable web videos

A summary of web-video-presentation from ConardLi/garden-skills: turn articles or scripts into click-driven 16:9 web presentations using Vite, React, TypeScript, theme tokens, chapter-by-chapter development, and hard checkpoints.

2026-05-15
5 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Prompt-Vault: a prompt specification library for testing AI coding ability

A summary of w512/Prompt-Vault: Bubble Sort visualization, todo list, sorting visualization, Kanban board, and Tauri Markdown editor prompts organized by difficulty for testing AI coding agents.

2026-05-15
4 minute read
中文简体 中文繁體 日本語 Español
AI Tools

What is Token Efficiency? DeepSeek V4, big-model planning, and small-model execution

A practical view of Token Efficiency in AI coding: DeepSeek V4 Pro / Flash positioning, big models for planning and consultation, small models for execution, plus context budgets, DAG orchestration, task replicas, evaluation, and atomic business workflows.

2026-05-15
6 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Superpowers: a skills framework that pulls coding agents back into engineering process

A summary of obra/superpowers: positioning, installation targets, base workflow, skills library, and boundaries. It combines brainstorming, planning, TDD, code review, worktrees, and subagents into a coding-agent methodology.

2026-05-15
4 minute read
中文简体 中文繁體 日本語 Español
Hardware

Honeywell PTM7950 confusion: do not judge only by thickness, origin, or black spots

A practical look at PTM7950 and PTM7950SP market confusion: 0.2 mm vs 0.25 mm, origin, black spots, color, COA, and authorization cannot alone prove authenticity or quality. Buyers should focus on batch traceability, testing, quality control, and after-sales support.

2026-05-15
5 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Reject Vibe Coding: Matt Pocock's skills repo adds engineering constraints to AI coding

A summary of Matt Pocock's skills repository: use grill-me, grill-with-docs, TDD, diagnose, and architecture review to bring AI coding back to requirement clarity, domain language, test feedback, and long-term maintainability.

2026-05-15
4 minute read
中文简体 中文繁體 日本語 Español
Technical Docs

GPT-5.5 Prompt Migration Guide: Why old prompts should be trimmed before rewritten

A practical summary of OpenAI's GPT-5.5 prompting guide: shorter outcome-first prompts, reasoning effort, preambles and phase, retrieval budgets, validation rules, and what to remove first when migrating old prompts.

2026-05-15
15 minute read
中文简体 中文繁體 日本語 Español
AI Tools

What is cc-haha? A project that turns Claude Code into a desktop workbench

A look at NanmiCoder/cc-haha: its positioning, desktop workbench, Computer Use, multi-model setup, H5 remote access, installation flow, and risk boundaries.

2026-05-14
9 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Codex /goal vs Claude Code /goal: running long tasks until they are done

A comparison of the /goal command in Codex CLI and Claude Code: both target long-running tasks and completion conditions, but they differ in availability, setup, evaluation, and best-fit workflows.

2026-05-14
7 minute read
中文简体 中文繁體 日本語 Español
AI Industry

What Jensen Huang Was Really Saying in His CMU Speech

A concise reading of Jensen Huang's CMU speech: young people may need to relearn hardship, traditional career paths are changing, and the hard problems of the AI era will be harder than they first appear.

2026-05-14
4 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Connecting Claude to Fusion 360: An Example of Editing STEP Models With AI

A practical walkthrough of connecting Claude to Fusion 360: enabling the API/MCP service, connecting the port, letting AI analyze a gear structure, and converting a screw-mounted planetary gear into a bearing-based design.

2026-05-14
5 minute read
中文简体 中文繁體 日本語 Español
AI Tools

How Can Codex Use Chinese LLMs? Managing OpenAI-Compatible APIs with CCX

CCX is an AI API proxy and protocol-conversion gateway for Claude Messages, OpenAI Chat, OpenAI Images, Codex Responses, and Gemini. This article explains its positioning, deployment, endpoints, channel orchestration, environment variables, and operational cautions.

2026-05-13
8 minute read
中文简体 中文繁體 日本語 Español
AI Tools

How Can Codex Use Chinese LLMs? OpenAI-Compatible APIs and the CodexBridge Approach

CodexBridge wraps Codex CLI/SDK as an OpenAI-compatible chat API, allowing OpenWebUI, Cherry Studio, curl, and other clients to call local Codex through /v1/chat/completions. This article explains its use cases, deployment, sessions, multimodal input, structured output, and common configuration.

2026-05-13
6 minute read
中文简体 中文繁體 日本語 Español
Technical Docs

Computer Terms in Plain Language: What TTS, STT, API, RAG, and Agent Really Mean

Many computer terms sound impressive, but they often describe very simple things. This article explains common terms such as TTS, STT, API, SDK, CRUD, Cache, Queue, Embedding, RAG, and Agent in plain language.

2026-05-12
9 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Can Sulphur 2 Run on 8GB VRAM? Notes on Local Deployment of an LTX 2.3 Video Model

Sulphur 2 is a video generation model from SulphurAI based on LTX 2.3. It supports text-to-video, image-to-video, and multiple LTX 2.3 workflows. This article covers local entry points, 8GB VRAM feasibility, tool choices, and common failure causes.

2026-05-12
9 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Running DeepSeek 4 Locally: Antirez's ds4 Experiment on Apple Silicon Mac

ds4 is a local DeepSeek V4 Flash inference engine written by Antirez for Apple Silicon, with CLI, HTTP server, and basic agent capabilities.

2026-05-11
3 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Why DeepSeek Became the Cost-Saving Key in This Round of AI Coding Tools

A look at the cost logic behind AI coding tools: why Claude Code, OpenClaw, Superpowers, and similar agent tools consume so many tokens, and why DeepSeek V4's long context and low cache pricing make it a key cost saver.

2026-05-11
7 minute read
中文简体 中文繁體 日本語 Español
AI Industry

ProgramBench Raw Leaderboard Data: Model Scores, Costs, and 200 Task Records

A structured copy of ProgramBench's public leaderboard, extended results, and 200 task records, preserving model scores, costs, call counts, test counts, and best scores.

2026-05-10
18 minute read
中文简体 中文繁體 日本語 Español
AI Industry

ProgramBench 0% Explained: The Scary Part Is Not Failure, but a Clear Roadmap

A concise explanation of ProgramBench, its 0% result, and what it really means for AI Coding: today's models cannot yet rebuild complete software from scratch, but full software engineering has now become a benchmarkable target.

2026-05-10
10 minute read
中文简体 中文繁體 日本語 Español
AI Tools

How to Choose Between GPT-5.5, GPT-5.4, and GPT-5.3-Codex

Based on official OpenAI documentation, this article compares GPT-5.5, GPT-5.4, and GPT-5.3-Codex in terms of use cases, credit consumption, Codex usage, and practical differences across common scenarios such as site rewriting, translation, Q&A, coding, and automation.

2026-05-10
9 minute read
中文简体 中文繁體 日本語 Español
AI Tools

How to Choose AI Coding Plans: Convenience for Light Users, Flexibility for Heavy Users

A practical guide to choosing AI coding tools and model plans: light users should prioritize convenience, mid-level users should focus on value, and heavy users should decouple models from tools to avoid being locked into a single ecosystem.

2026-05-10
7 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Chrome Silently Downloads 4GB Gemini Nano: How to Check, Disable, and Delete It

A concise look at the controversy around Chrome silently downloading the roughly 4GB Gemini Nano local AI model, including file locations, affected platforms, Google's response, and how users can check and disable it.

2026-05-09
5 minute read
中文简体 中文繁體 日本語 Español
AI Tools

A Practical llama.cpp Multi-GPU Benchmarking Approach: Is 2x V100 16GB Faster Than One 32GB Card?

A practical look at llama.cpp multi-GPU offload performance: dual GPUs are not always faster when one card can fit the model, but they can help a lot when a single 16GB card would fall back to CPU offload. Also covers V100 PCIe and NVLink differences.

2026-05-09
8 minute read
中文简体 中文繁體 日本語 Español
AI Tools

Claude Code Limits Doubled: Anthropic Uses SpaceX Compute Expansion to Ease Usage Constraints

A summary of Anthropic's May 2026 increase to Claude Code and Claude API limits, and what its SpaceX compute partnership means for Claude Pro, Max, Team, and enterprise users.

2026-05-09
5 minute read
中文简体 中文繁體 日本語 Español
AI Tools

OpenAI's New Realtime Voice Models: GPT-Realtime-2, Live Translation, and Streaming Transcription

A concise look at OpenAI's May 2026 Realtime API voice models, including GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper: capabilities, use cases, pricing, and developer impact.

2026-05-09
5 minute read
中文简体 中文繁體 日本語 Español
1 2 3 4 5 6 7 10
© 2022 - 2026 KnightLi Blog
记录并分享
Built with Hugo
Theme Stack designed by Jimmy