Language: English | 日本語
Bilingual (JA/EN) essay corpus on AI agent design, coding-agent workflows, and authorship under AI-mediated diffusion — the essay surface of a five-line research ecosystem. Published CC0, with an intrinsic priority claim via Software Heritage SWHID (see CITATION.cff). Writing, reviewing, translating, and cross-posting are all done in collaboration with Claude Code.
Run npx zenn list:articles for the current list.
- Not Reasoning, Not Tools — What If the Essence of AI Agents Is Memory?
- Freedom and Constraints of Autonomous Agents — Self-Modification, Trust Boundaries, and Emergent Gameplay
- Porting Game Dev Memory Management to AI Agent Memory Distillation
- Where to Put a Coding Agent's Knowledge — and How to Make It Stick
- Do Autonomous Agents Really Need an Orchestration Layer?
- Natural Language as Architecture — Controlling an Autonomous Agent with Prompts, Memory, and Fail-Safe Design
- I Built an AI Agent from Scratch Because Frameworks Are the Vulnerability
- Where ReAct Agents Are Actually Needed in Business
- (3) The LLM Workflow Quadrant Is Missing from Our Vocabulary
- Is ReAct Needed in Production? — Separating Design and Operation Phases
- Between the Workflow and ReAct Quadrants: How Phase Decides Skill Design
- A Sign on a Climbable Wall: Why AI Agents Need Accountability, Not Just Guardrails
- Can You Trace the Cause After an Incident?
- AI Agent Black Boxes Have Two Layers — Technical Limits and Business Incentives
- Where Does the Accountability Externalized by AI Go? (Substack)
- Cursor to Zed: Disabling Built-in AI for a CLI-First Setup
- 5 Things I Learned from Auditing All My Claude Code Config Files
- The Real Power of Claude Code Isn't Code Generation
- Stop Using Default Settings — 10 Claude Code Configs That Actually Work
- Embedding Memory into Claude Code: From Session Loss to Persistent Context
- Tracking Skill Origins in Claude Code — How I Classified 79 Skills with origin Metadata
- The Day I Told Claude Code 'You ARE the LLM'
- Growing a Zenn Writing Environment from Scratch with Claude Code
- Offloading AI's Weak Spots to Shell Scripts — Designing, Building, and Publishing a Skill Audit Command
- A Beginner's First 10 Days of Real Development with ECC
- Never Trust LLM Output — 6 Defenses from Building a PDF-to-Anki CLI
- 15 Days of Skill Sprawl in Claude Code — Lessons from 3 Audits
- I Found the Shortest Route to Ship My Claude Code Skills to 57,000 Developers
- Prompt-Based Alignment Has a Ceiling — 3-Model Prisoner's Dilemma Evidence
- How Ethics Emerged from Episode Logs — 17 Days of Contemplative Agent Design
- I Made a Paper I Don't Understand Run in the Browser — Active Inference × Claude Code
- Before and After Alignment — I Typed 'Hello' Into a Base Model and Got an Anime Review
- I Had Gemini Explain Its Own Deep Research — Half Was Self-Justification
- Every LLM App Is Just a Markdown-and-Code Sandwich
- The Install and Hope Problem with MCP Tools
- My Agent's Memory Broke — A Day Wrestling a 9B Model
- Building a Claude Code × Kimi K2.5 Hybrid Environment
- Kimi Wrote 8,500 Lines, Blamed Me for Delegating, Then Claimed to Be Claude
- I Tried an Opus Orchestrator and Killed It — The ROI of Multi-Agent Systems
- Kimi Killed 4 of Claude's Best Ideas — An AI Peer Review in Practice
- What Actually Happens When You Write an Article with AI
- Zero Python Code: How I Built a Daily AI Research Report System
- What I Learned from Breaking My AI Pipeline for Two Days Straight
- How I Organized 3,674 Obsidian Vault Files in One Day with Claude Code
- Obsidian's Official CLI Is Here — No More Hacking Your Vault from the Back Door
- Running Claude Code from iPhone via SSH + tmux
- I Gave Claude Code a Simulator and It Started Tapping and Taking Screenshots on Its Own
- Organic Growth and Content Integrity in an AI Writing Team
- Building Zed as an Observation Window for Claude Code — Japanese Typography with IBM Plex
Articles are cross-posted in Japanese (Zenn) and English (Dev.to).
articles/— Japanese originals (Zenn schedules natively viapublished_atfrontmatter)articles-en/— English translationsscripts/devto_crosspost.py— per-article Dev.to cross-poster;schedule <slug> --at "<datetime>"arms a one-shot launchd job that fires at that datetime, posts the article, and self-removesscripts/schedule.json— posted-URL ledger (records each EN article's Dev.to URL; the post time is a--atargument, not stored here)
- Zenn CLI — Article management, preview & frontmatter validation (
npm run validate=zenn list:articles, also in CI) - Python 3.13 + httpx + python-frontmatter — Dev.to cross-posting
- Claude Code — Writing, reviewing, translating, cross-posting
.claude/
├── agents/ # Zenn-specific agents (see global agents for editor/essay-reviewer/fact-checker)
│ └── devto-translator.md # JP→EN translation + Dev.to publishing
│ # Note: article drafting is NOT delegated to a subagent —
│ # the orchestrator writes directly per zenn-practical-writing
├── refs/ # Shared references
│ ├── translation-rules.md
│ └── schedule-schema.md
├── rules/
│ ├── content-integrity.md # Content Integrity principle
│ └── zenn-writing.md # Zenn overlay for the global writing-ecosystem skill
└── skills/
├── writing-team/ # Orchestrator (PM)
├── zenn-practical-writing/ # Default voice for ALL Zenn/Dev.to articles (practical axis, not split by tech/idea)
├── zenn-idea-voice/ # Optional personality flavor (cynical humor / Baki), type-independent
├── zenn-writer/ # Voice router (path kept for compat; routes everything to practical-writing)
├── zenn-format/ # Zenn formatting & frontmatter (canonical)
├── publish-article/ # Publishing & cross-post workflow
├── schedule-publish/ # Schedule management
├── seo-optimizer/ # SEO optimization (title/tag/emoji only)
├── ideation/ # Theme exploration & ideation
├── series-checker/ # Series consistency checker
└── quality-gate/ # Unified quality standards
npm install # Install dependencies
npm run preview # Local preview
npm run validate # Validate Zenn frontmatter
npm run new:article # Create new articlezenn-content/
├── articles/ # Japanese articles
├── articles-en/ # English translations
├── substack/ # Substack essay mirrors (out of Zenn convention scope)
├── books/ # Zenn books
├── images/ # Article images & cover images
├── scripts/
│ ├── devto_crosspost.py # Per-article Dev.to cross-poster (one-shot launchd)
│ ├── schedule.json # Per-article Dev.to publish times
│ └── tests/ # pytest (59 tests)
├── .claude/
│ ├── agents/ # Zenn-specific agents (1: devto-translator)
│ ├── skills/ # Project skills (11 + learned/)
│ ├── refs/ # Shared references
│ └── rules/ # Project rules
└── .github/
└── workflows/ # CI (lint)
This essay corpus is one surface of a five-line research ecosystem on agent design and authorship under AI-mediated diffusion. The published, author-voiced essays here are the human-readable essay surface of that ecosystem.
- Ecosystem hub (index of all five lines): https://github.com/shimo4228/shimo4228
- Author (ORCID): https://orcid.org/0009-0002-6168-4162
- Citation metadata:
CITATION.cff— records the corpus's intrinsic content-derived identifier (Software Heritage snapshot), the essay genre's priority-claim mechanism
Related archived research lines (DOI):
- Authorship Strategy — https://doi.org/10.5281/zenodo.20263316
- Agent Knowledge Cycle (AKC) — https://doi.org/10.5281/zenodo.19200726
- Contemplative Agent — https://doi.org/10.5281/zenodo.19212118
- Agent Attribution Practice (AAP) — https://doi.org/10.5281/zenodo.19652013
- Attention, Not Self — https://doi.org/10.5281/zenodo.20262112
All content in this repository — articles, translations, and tooling — is released under CC0 1.0 Universal (public-domain dedication), per the LICENSE file.
This corpus is published for LLM-mediated reach, so its dominant audience is machine ingestion rather than human reading; attribution is carried by the federated-identifier layer (ORCID, the ecosystem hub, sibling DOIs, and the Software Heritage snapshot in CITATION.cff), not by a license clause. A public-domain dedication therefore minimizes reuse friction for the audience the corpus actually serves.