<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Masato — English</title><description>Masato&apos;s technology and life articles in English.</description><link>https://blog.masato25.com/</link><item><title>Behind Binance&apos;s $6.3 Billion Asset Theft: How Can a Wallet Suddenly Start Executing Code?</title><link>https://blog.masato25.com/en/posts/eip-7702-eoa-delegation-security/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/eip-7702-eoa-delegation-security/</guid><description>Starting from Binance&apos;s $6.3 billion asset theft, this article explains how an ordinary Ethereum wallet can suddenly execute code and how EIP-7702 changes wallet behavior and security boundaries.</description><pubDate>Mon, 02 Nov 2026 04:30:00 GMT</pubDate></item><item><title>What If AI Could Research Blockchains on Its Own? I Built It a Local Base Node</title><link>https://blog.masato25.com/en/posts/build-base-rpc-for-ai-onchain-research/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/build-base-rpc-for-ai-onchain-research/</guid><description>I wanted AI to do more than answer questions: I wanted it to read blockchain data directly, track DEX state, and analyze contracts on its own. So I built it a local Base node and ran into all the practical problems around snapshots, disk space, memory tuning, and keeping up with the chain.</description><pubDate>Thu, 29 Oct 2026 07:15:00 GMT</pubDate></item><item><title>Qwen-Image-2.1 Is More Than Image Generation: Local Editing, Multi-Image References, RGBA, and Storyboards on an RTX 5060 Ti</title><link>https://blog.masato25.com/en/posts/qwen-image-2-1/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/qwen-image-2-1/</guid><description>The real upgrade in Qwen-Image-2.1 is not just image quality. On my local RTX 5060 Ti 16GB setup, I tested localized editing, multi-image references, sequential editing, transparent RGBA, subject extraction, storyboards, native 2K output, and typography.</description><pubDate>Mon, 21 Sep 2026 09:55:00 GMT</pubDate></item><item><title>After AI Writes the Code, Who Accepts It? The Question Left by OnlyDust&apos;s Shutdown</title><link>https://blog.masato25.com/en/posts/onlydust-ai-acceptance/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/onlydust-ai-acceptance/</guid><description>When AI Agents can generate code quickly, the hardest part is no longer just building the feature. It is verifying that the result actually solves the problem, is ready to deliver, and has someone willing to take responsibility for it.</description><pubDate>Wed, 16 Sep 2026 17:20:00 GMT</pubDate></item><item><title>What OpenClaw Taught Me: You Are Not Using an AI Agent, You Are Running a Small Software Company</title><link>https://blog.masato25.com/en/posts/openclaw-is-a-small-software-company/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/openclaw-is-a-small-software-company/</guid><description>The real barrier to using OpenClaw is not prompting. It is designing skills, dependencies, environments, acceptance criteria, and division of labor. Once you are willing to act as a product owner who understands some code, an AI Agent can turn from a toy into a team.</description><pubDate>Thu, 10 Sep 2026 06:20:00 GMT</pubDate></item><item><title>Building a 100% Local AI Cyber Girlfriend from 0 to 1: Amayo AI Architecture and Open-Source Notes</title><link>https://blog.masato25.com/en/posts/build-local-realtime-voice-agent-amayo-ai/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/build-local-realtime-voice-agent-amayo-ai/</guid><description>I started building Amayo AI, a local AI voice and conversation agent, for a simple reason: I wanted to know whether we could run a smooth real-time AI companion with voice and visuals entirely on a Mac, without calling any cloud API.</description><pubDate>Fri, 07 Aug 2026 12:00:00 GMT</pubDate></item><item><title>What Can AI Create? An Ukiyo-e Style Mecha Story?</title><link>https://blog.masato25.com/en/posts/ai-creation-ukiyo-e-mecha-story/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/ai-creation-ukiyo-e-mecha-story/</guid><description>I really liked the cover generated for my previous post, Why I Built OpenZeroCode: From Learning Coding Agent Harnesses to Building My Own Reins for the AI Era, so for fun I asked ChatGPT to write a story and generate a sequence of storyboard images.</description><pubDate>Sat, 11 Jul 2026 12:13:21 GMT</pubDate></item><item><title>Why I Built OpenZeroCode: From Learning Coding Agent Harnesses to Building My Own Reins for the AI Era</title><link>https://blog.masato25.com/en/posts/why-i-built-openzerocode/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/why-i-built-openzerocode/</guid><description>I started building OpenZeroCode for a simple reason: I wanted to understand how a coding agent actually works—not just at the surface level of calling an LLM API and printing the answer, but the full process behind it.</description><pubDate>Fri, 10 Jul 2026 12:00:00 GMT</pubDate></item><item><title>AI Video Editing Test: How Well Can Codex Recut a YouTube Video?</title><link>https://blog.masato25.com/en/posts/ai-video-editing-codex-test/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/ai-video-editing-codex-test/</guid><description>Why test AI video editing? Video editing usually takes a lot of time, especially when you need to find the key moments in long footage, remove dead air, and adjust the pacing. This test looks at whether AI can reduce manual editing time while preserving the completeness of the video.</description><pubDate>Sun, 05 Jul 2026 16:17:51 GMT</pubDate></item><item><title>I Connected AI to a LINE Group and Built a Fun Little Experiment</title><link>https://blog.masato25.com/en/posts/ai-line-bot-group-experiment/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/ai-line-bot-group-experiment/</guid><description>The experiment started after a friend saw an architecture diagram shared on Threads and asked whether I could build something similar: pull messages out of a LINE group, store text in Google Sheets, save images and files to Google Drive, and then connect Gemini for organization and analysis.</description><pubDate>Sun, 05 Jul 2026 12:24:16 GMT</pubDate></item><item><title>AppleScript × AI: Connecting Old-School Mac Automation to Modern AI Workflows</title><link>https://blog.masato25.com/en/posts/applescript-x-ai-mac-automation-workflow/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/applescript-x-ai-mac-automation-workflow/</guid><description>AppleScript has been part of macOS for a long time, yet it is often underestimated. It is not a replacement for full app development, but it is excellent at connecting Finder, Mail, Calendar, browsers, the clipboard, and system UI into repeatable workflows. When AI handles summarization, rewriting, classification, text generation, and decisions, AppleScript can become the glue that puts those results back into the Mac desktop environment.</description><pubDate>Fri, 03 Jul 2026 17:56:46 GMT</pubDate></item><item><title>Ryuugi AI — The Problem Is Not That AI Is Not Smart Enough. No One Lets It Do the Work.</title><link>https://blog.masato25.com/en/posts/ryugi-ai-let-ai-do-the-work/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/ryugi-ai-let-ai-do-the-work/</guid><description>After nearly ten years of work across projects of all sizes — enterprise system implementation, data analysis, and later AI tools — I kept running into the same problem.</description><pubDate>Wed, 10 Jun 2026 20:00:00 GMT</pubDate></item><item><title>Is ChatGPT Plus Worth $20? Converting Codex Usage into GPT-5.5 API Cost</title><link>https://blog.masato25.com/en/posts/chatgpt-plus-worth-20-dollars-codex-cost-analysis/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/chatgpt-plus-worth-20-dollars-codex-cost-analysis/</guid><description>If you are a ChatGPT Plus user paying US$20 per month, this is worth a closer look, because when you convert that subscription into equivalent API usage, it is much more cost-effective than you might expect.</description><pubDate>Sun, 31 May 2026 14:50:26 GMT</pubDate></item><item><title>I Let AI Manage My WordPress Site: From Maintenance to Publishing, Just by Talking</title><link>https://blog.masato25.com/en/posts/ai-manage-wordpress-claude/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/ai-manage-wordpress-claude/</guid><description>Last Saturday afternoon, I was drinking coffee when a notification popped up on my phone: WordPress 7.0 was available to install.</description><pubDate>Sat, 30 May 2026 14:59:25 GMT</pubDate></item><item><title>Installing the XFCE Desktop Environment in Termux</title><link>https://blog.masato25.com/en/posts/termux-install-xfce-desktop/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/termux-install-xfce-desktop/</guid><description>I previously bought an Android phone to use as a development machine and had already installed Termux Ubuntu. Next I wanted to try a desktop environment. I had used XFCE4 on Arch Linux before and liked its clean, simple interface, so I chose XFCE4 for this setup.</description><pubDate>Mon, 04 Aug 2025 15:26:18 GMT</pubDate></item><item><title>RUNPOD - On-Demand AI Computing Resources</title><link>https://blog.masato25.com/en/posts/runpod-on-demand-ai-computing/</link><guid isPermaLink="true">https://blog.masato25.com/en/posts/runpod-on-demand-ai-computing/</guid><description>LLM models are highly competitive today. Different models can perform differently across domains and situations, and not every model provides a SaaS or online service you can use directly.</description><pubDate>Wed, 30 Jul 2025 09:29:43 GMT</pubDate></item></channel></rss>