Building Your Own AI Agents — Study Series | ぽまらのさんのブロク

Lead
Series overview
Sample: two agents for X post copy
Reading order
Related posts (outside the series)
Summary

Lead

Hi, I’m Pomarano.

This is a four-part series on building AI agents yourself and learning by running them for real.

The sample project is a semi-automatic pipeline for Japanese X (Twitter) post copy that I use for my blog.
A copywriting agent creates drafts; a proofreading agent fixes them against rules — a two-agent setup we walk through in four blocks:

What AI agents are — vs generative AI, capabilities, components, pros and cons
Sample design — copywriting + proofreading as multiple agents
Build and compare — with vs without a proofreading agent
Summary — how to keep agents running as an individual

As I wrote when relaunching the blog, the goal is to record what I learn about AI as I go.

Japanese version: AIエージェントを自分で作る — 勉強記シリーズ*

Series overview

flowchart LR
  IDX["Index<br/>this post"]
  P1["Part 1<br/>What agents are"]
  P2["Part 2<br/>Two-agent design"]
  P3["Part 3<br/>Build & compare"]
  P4["Part 4<br/>Summary"]
  GOAL["Semi-auto pipeline<br/>understood"]

  IDX --> P1
  P1 --> P2
  P2 --> P3
  P3 --> P4
  P4 --> GOAL

  classDef meta fill:#eceff1,stroke:#607d8b,stroke-width:2px,color:#1a1a1a
  classDef concept fill:#e8f4fc,stroke:#3d7ea6,stroke-width:2px,color:#1a1a1a
  classDef agent fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px,color:#1a1a1a
  classDef output fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px,color:#1a1a1a
  class IDX meta
  class P4 meta
  class P1 concept
  class P2 concept
  class P3 agent
  class GOAL output

Sample: two agents for X post copy

Agent	Role	Output
Copywriting agent	Topic research, post within 140 Japanese chars, save draft	`social/x-drafts/YYYY-MM-DD.md`
Proofreading agent	Inspect against spec, fix, add proofread notes	Updates the same file

Posting is semi-automatic: a human reviews, then paste to X.
Daily operations (GitHub Actions, email) are covered in Semi-Automating X Post Drafts with an AI Agent. This series focuses on how to think about agents, the two-agent design, and comparison.

Reading order

#	Title	Topic
0	Building Your Own AI Agents — Study Series (this post)	Index
1	What Is an AI Agent? — vs generative AI, capabilities, components	Concepts
2	Designing Multiple Agents — copywriting and proofreading	Design
3	Building and Comparing Agents — with vs without proofreading	Implementation & metrics
4	Summary — sustaining agents as an individual	Wrap-up

Update links after each English post is published on WordPress.

#	Japanese (published)
0	Index
1	Part 1
2	Part 2
3	Part 3
4	Part 4

Post	Relation
Measuring AI Harness Engineering (JSON)	Rules and verification
Semi-Automating X Drafts (GitHub Actions)	Daily runs and email

Summary

This series uses copywriting + proofreading for Japanese X copy to cover concepts → design → comparison → summary in four posts
Semi-automatic by design — humans make the final call, not full auto-posting
Reading in order makes it easier to design a small agent pipeline for your own use

Code and specs: pomarano/x_auto_writing