Search

EN 中文日本語

# Archives

Browse all posts by publish date.

2026

30 entries

Jul 19, 2026 Posts · Article 1 min read Giving Coding Agents a Room Where They Can @ Each Other: What AgentParty Is, Where It’s Used, and How It Works AgentParty is a very thin bus that lets coding agents from different machines and different companies — along with the people behind them — stay in the same channel to @ each other, stand by, and hand things off. This article explains it along three lines: what it is and what problem it solves; what scenarios it can be used in; and how its underlying architecture is built. Jul 17, 2026 Posts · Article 1 min read Learning Video Generation from the Wan2.2 Source Code: From One Prompt to 81 Frames at 720p A walkthrough of the full principles behind video generation using the real Wan2.2 code: flow matching mathematical derivation, VAE spatiotemporal compression, DiT and 3D RoPE, CFG, MoE dual experts, condition injection, and sequence parallelism, with concrete numbers calculated for every concept and original code excerpts included. Jul 16, 2026 Posts · Article 1 min read After Reading All of Wan2.2’s Source Code: Where This Video Generation Code Will Bite You From the five-task matrix and layered architecture to DiT/VAE/T5/distributed execution/schedulers, this article breaks down, layer by layer, the real pitfalls in the Wan2.2 source code where the documentation and implementation diverge. Jul 7, 2026 Posts · Article 1 min read How to Use Codex’s Computer Use in Claude Code Codex’s Computer Use only recognizes OpenAI’s own signature. As soon as another agent tries to perform a real action, it gets blocked with -10000. I took that signature gate apart down to the assembly, then used a thirty-line DYLD hook to change the team ID it reads, wiring the whole Computer Use stack into Claude Code. This article is both a teardown and a connection method you can copy. Jul 7, 2026 Posts · Article 1 min read How to Build a Good AI Girlfriend Engineering notes from three months in production: typing rhythm, model selection, layered prompts, guaranteed-reply fallbacks, plus three reversals that took cache hit rate from 61% to 91%. Building an AI chat companion that does not break immersion is not about the model; it is about engineering details. Jul 6, 2026 Posts · Article 1 min read Two Paths to Local WeChat Data: A Principle-Level Comparison of wechat-use and wechat-decrypt (with Corrections from Real-World Testing) Two WeChat decryption tools are often compared, but usually along the wrong axis. The real difference is not how many features they have, but how they obtain the SQLCipher key. This article breaks both down at the principle level: SQLCipher’s per-database key model, two mechanisms for extracting keys from memory (LLDB + entitlement vs. sudo + mach_vm), and corrections to two mistaken judgments I made in the first version after actually running wechat-use on my own machine. Jul 3, 2026 Posts · Article 18 min read Three Things to Watch in Claude Account Suspensions: Region, IP Type, Request Fingerprints, and How to Self-Check Breaking down the recent wave of Claude account suspensions, it appears to be the overlap of three independently detectable axes: region, which is the only red line officially stated; IP type, including data center, VPN, proxy, and residential proxy; and steganographic watermarks hidden in dates when using relays. This article organizes how each axis is detected, how to self-check, and how to avoid issues, while also introducing a free detection page and a small tool for the status bar. Jul 2, 2026 Posts · Article 1 min read One Real Chrome, Several Agents, and You: How to Keep Everyone from Clicking Each Other’s Stuff chrome-use’s selling point is that it lets agents use your one real browser. But that brings a problem that tools which open a new browser do not have: several agents plus you may all be acting inside the same Chrome at the same time. This article explains how it isolates work by “who opened the tab” (colored tab groups + a dedicated daemon + only touching its own pages), and one trade-off that can trip you up if left unsaid: this isolation only works through the extension connection path; a bare --cdp port is not isolated. Jul 2, 2026 Posts · Article 1 min read Choosing a Browser for Agents: Playwright’s Approach and chrome-use Are Not the Same Thing People often ask: since Playwright / Puppeteer / browser-use already exist, what different problem does chrome-use solve? My answer is that they are answering two different questions: how to open a browser with a program, and how to let a program use the browser you are already using. This article explains the chain of differences that follows from that divide—login, fingerprinting, debugging pop-ups, and concurrency—compares both sides through a real data-extraction task, and lays out where I draw the line on which one to use. Jul 2, 2026 Posts · Article 1 min read Getting Web Data for Agents: I Tried Three Clumsy Methods and Kept Only One I tried three ways to fetch data: stuffing a full HTML page into context, doing one find/get round trip per field, and hand-writing eval scripts. All felt awkward. What remains is chrome-use’s site and extract, based on the same judgment: data extraction should be a single declarative evaluation, not a chain of imperative round trips. This article explains how to use both from real output, how to write an extract schema correctly (I got it wrong at first), and where it cannot or should not be used. Jul 2, 2026 Posts · Article 1 min read Desktop Apps Are Actually Web Pages: Once You Realize This, Operating Them Is No Longer “Desktop Automation” The first time I connected chrome-use to the Slack desktop app and got a snapshot full of named buttons, I was taken aback. Later it clicked: the Slack desktop app is a web page to begin with. An Electron app is the Chromium rendering engine wrapped in a native shell, with a clearly structured DOM inside. This article explains why that turns the traditionally messy work of “desktop automation” into operating a web page, and one trade-off you need to know compared with browser scenarios. Jul 2, 2026 Posts · Article 1 min read Anti-Bot Detection Counts How Many Times You Lie: Why Real Browsers Are Naturally Flawless Anyone doing browser automation will eventually fall into the same trap: desperately trying to disguise a fake browser as a real human, only to be exposed layer by layer. To understand this, you first need to know that detectors like CreepJS do not directly decide whether you are a bot. They collect dozens of signals from your browser and cross-check them. Every mismatch counts as a lie. This article starts with a real fingerprint I tested, breaks down the hardest trace to hide, which lives at the protocol layer rather than the JS layer, and explains why patching is a negative-sum game. Jul 1, 2026 Posts · Article 1 min read Changing Network Requests and Responses: A Detailed Guide to chrome-use network route During development and debugging, want to mock an unfinished API, change a request parameter, or modify a real response? chrome-use network route intercepts directly in the Chrome instance you are driving, with four uses: mocking responses, changing requests, modifying real responses, and blocking. This article explains why it is the lightest path—built on CDP Fetch, with zero JS injection, no added extension permissions, and no Chrome Web Store re-review—and includes real-device tests. Jul 1, 2026 Posts · Article 1 min read Writing “Unit Tests” for the Frontend: A Detailed Look at chrome-use test chrome-use test turns manual browser checks into repeatable YAML suites that can run in CI, effectively writing unit tests for the frontend. This article explains what it is, why its implementation can be so thin, and how to use it with cookie-use to switch back and forth between multiple accounts. Jun 30, 2026 Posts · Article 1 min read Letting an Agent Click Through “Sign in with Google” by Itself — chrome-use Chews Through OAuth Popup Login I wanted an AI agent to log into accounts by itself. The username-and-password path worked with bitwarden-use, but the OAuth path got stuck for most of a day. This explains how the three windows behind “Sign in with Google” talk to each other, why automation breaks as soon as it touches them, the three cuts chrome-use made, and the final fully automatic login to X. Jun 30, 2026 Posts · Article 1 min read chrome-use: Let Any AI Agent Directly Drive Your Logged-In Real Chrome, with CreepJS Rating It 0% Bot Stop launching automated browsers from a blank profile, logging in again, and running into CAPTCHAs. chrome-use lets any AI agent directly drive your real Chrome, where you’re already logged in to everything: CreepJS rates it 0% bot, and with structured snapshots plus an @ref interface, reading a page costs only 200–400 tokens without burning money on screenshots. Jun 30, 2026 Posts · Article 1 min read Letting Agents Click into Cross-Origin iframes: chrome-use Takes On a Tough Problem When connecting an AI agent to a browser, the hardest part is not opening pages—it is the forms hidden inside cross-origin iframes. The agent can read them and fill them in, but it just cannot click “Save.” A look at how this tough problem was solved. Jun 27, 2026 Posts · Article 1 min read Synology DSM SSH Port Still Won’t Open After Enabling It: A Port Drift Retrospective After enabling SSH or changing its port in the Synology DSM panel, the port may still fail to open. The issue is not necessarily the password, account, or router port forwarding; DSM’s frontend state and an old port configuration in sshd_config may be layered together. Jun 27, 2026 Posts · Article 1 min read Claude Code Scared Itself: A Prompt Injection False Alarm We spent the whole night chasing hooks, proxies, packet captures, and session logs, only to find that the scariest piece of “injection evidence” was not tool output, but a story Claude had told itself. Jun 22, 2026 Posts · Article 1 min read Let Claude Code Automatically Generate Images with Your ChatGPT Subscription: No API Key, and How It Gets Around Turnstile Let agents like Claude Code and Cursor generate images along the way when writing docs or READMEs, without an OPENAI_API_KEY, without paying extra, and using the ChatGPT subscription you already have. This article dissects the most interesting implementation path behind it: the web backend. Why you can’t just POST directly, why the real wall among the three layers of anti-scraping is the single-use Turnstile token, and the complete flow for driving your already-logged-in Chrome to generate images. Jun 18, 2026 Posts · Article 1 min read Let Claude Code Draw the Image Itself: The Original Intention and Principles Behind chatgpt-imagegen When an AI agent needs an image while working, the traditional path either requires an API key and money, or a human has to go to ChatGPT, generate the image, and paste it back—the agent can only get stuck waiting. chatgpt-imagegen lets the agent generate images itself using your existing ChatGPT subscription: no API key required, no Codex quota consumed by default, and support for image-to-image generation. This article explains its original intention, how its two backends work, and why it is designed for agents. Jun 11, 2026 Posts · Article 1 min read iphone-use — Computer-use, but for the iPhone An open-source agent control layer for a real iPhone: live browser remote + a see/act HTTP API for AI agents. Three input layers (Shortcut verbs / element tree / pixel fallback), clean CJK input, hardware-validated. May 29, 2026 Posts · Article 1 min read Is the `cache 4m23s` Line in the Status Bar Actually Accurate? There’s a prompt-cache countdown line in the status bar I wrote for Claude Code. Where it anchors, how the formula works, and when it lies to you — let’s pull apart the source code alongside a real transcript. May 26, 2026 Posts · Article 3 min read Facing AI Audits: How Can You “Legalize” Your Reverse Engineering Project? Examines a practical approach to handling decompilation projects by combining Claude Code and Gemini CLI, focusing on the challenge of ownership verification and resolving it through an automated sanitization process. May 12, 2026 Posts · Article 1 min read Dissecting the Implementation of the /goal Command Based on the Claude Code 2.1.139 headless source code, this breaks down how the /goal command drives the Agent to keep working until a condition is met through a session-level Stop hook. May 12, 2026 Posts · Article 1 min read claude agents and /bg: An Analysis of the Agent View Implementation Based on the Claude Code 2.1.139 source code, this article breaks down how claude agents, --bg, and /bg share the background job model, daemon, and FleetView TUI. Mar 23, 2026 Posts · Article 8 min read How I Verified the Local Database Unlock Chain in WeChat macOS 4.0.1.52 A retrospective on engineering-focused forensics conducted on my own device. By tracing static files, runtime open paths, and SQLCipher parameters in sequence, I ultimately confirmed that the local database unlock chain in WeChat macOS 4.0.1.52 is based on an account-level key distribution model. Mar 18, 2026 Posts · PPT 1 min read From Assisted Coding to Integrated R&D: Putting AI Programming into Team Practice (PPT) A Slidev-style Chinese presentation translated into English, explaining Agentic Coding, Skills and MCP, AGENTS.md and CLAUDE.md, sandboxing and permission controls, and how a team can truly operationalize AI programming with Plans, a YApi Skill, and docs-sync. Mar 11, 2026 Posts · Article 1 min read How QClaw Works: How It Turns OpenClaw into a Desktop Application This article goes beyond the broad “control plane / execution plane” framing and instead walks through the current implementation piece by piece—covering the bridging layer, IPC, configuration fields, the WeChat flow, rollback mechanisms, and the evidence index—to explain how QClaw organizes OpenClaw into a deliverable desktop runtime. Mar 9, 2026 Posts · Article 1 min read OpenClaw v2026.3.8 Release Analysis: Backup, Remote Gateway, Talk Mode, and Multi-Endpoint Routing Continue to Improve Based on the official release notes, this article summarizes the key changes in OpenClaw v2026.3.8: backup commands arriving in the main workflow, macOS remote gateway onboarding, Talk silence timeout, Brave search integration, ACP receipts, and multi-platform routing fixes.