Shinsuke KAGAWA

Building agentic coding workflows and open-sourcing what I learn. I write about what breaks.

Japan Joined on Aug 21, 2025

Shinsuke KAGAWA

Sep 20

What Retrieval Still Hasn't Decided

#ai #rag #llm #machinelearning

11 min read

Want to connect with Shinsuke KAGAWA?

Create an account to connect with Shinsuke KAGAWA. You can also sign in below to proceed if you already have an account.

Create Account

Already have an account? Sign in

Shinsuke KAGAWA

Sep 19

I Found Three Broken Rules in My Own AGENTS.md

#ai #llm #softwareengineering #promptengineering

5 min read

Shinsuke KAGAWA

Sep 5

Switching from GPT-5.6 Sol to GPT-6 Astra: Start with Medium Effort

#ai #programming #tutorial #llm

6 min read

Shinsuke KAGAWA

Aug 20

Was the Monorepo Really Best for Coding Agents?

#ai #programming #monorepo #architecture

11 min read

Exploration multipliers spark overengineering

Shinsuke KAGAWA

Aug 5

Reasoning Effort Is Not a Quality Setting

#ai #programming #tutorial #reviews

12 min read

Good models fail on rigid scaffolds

Shinsuke KAGAWA

Aug 2

When Better Models Make Old Agent Workflows Worse

#ai #programming #promptengineering #llm

8 min read

Shinsuke KAGAWA

May 11

A second pair of eyes for Claude Code: building Galley, a local runner that checks the work before the PR opens

#ai #opensource #programming #productivity

10 min read

Shinsuke KAGAWA

Apr 2

I Built a Skill Reviewer. Then I Ran It on Itself.

#ai #llm #promptengineering #softwareengineering

5 min read

Shinsuke KAGAWA

Mar 18

Same Framework, Different Engine: Porting AI Coding Workflows from Claude Code to Codex CLI

#ai #productivity #programming #tutorial

7 min read

Shinsuke KAGAWA

Feb 12

Letting LLMs Jump — and Then Verifying Ruthlessly

#ai #llm #productivity #programming

5 min read

Shinsuke KAGAWA

Feb 4

Design Integration Checkpoints Before Letting LLMs Code

#ai #programming #productivity #architecture

3 min read

Shinsuke KAGAWA

Jan 26

Planning Is the Real Superpower of Agentic Coding

#ai #llm #softwareengineering #architecture

9 min read

Shinsuke KAGAWA

Jan 22

Stop Guessing If Your Prompt Is Better

#ai #promptengineering #productivity #tutorial

2 min read

Shinsuke KAGAWA

Jan 19

Stop Putting Everything in AGENTS.md

#ai #llm #softwareengineering #productivity

7 min read

Shinsuke KAGAWA

Jan 12

Why LLMs Are Bad at "First Try" and Great at Verification

#ai #llm #softwareengineering #promptengineering

6 min read

Shinsuke KAGAWA

Jan 6

Building a Local RAG for Agentic Coding: From Fixed Chunks to Semantic Search with Keyword Boost

#ai #rag #architecture #mcp

9 min read

Shinsuke KAGAWA

Dec 26 '25

How I Made Legacy Code AI-Friendly with Auto-Generated Docs

#ai #productivity #automation #tutorial

2 min read

Shinsuke KAGAWA

Dec 11 '25

Taming Opus 4.5's Efficiency: Using TodoWrite to Keep Claude Code on Track

#ai #productivity #programming #llm

2 min read

Shinsuke KAGAWA

Dec 8 '25

Stopping Cursor from Skipping Steps: A Structural Approach

#ai #llm #softwareengineering #productivity

19 min read

Shinsuke KAGAWA

Nov 26 '25

Rethinking Team Development in the Age of LLMs

#ai #llm #softwaredevelopment #productivity

12 min read

Shinsuke KAGAWA

Sep 8 '25

From 49 to 95: How Prompt Engineering Boosted Gemini MCP Image Generation

#ai #promptengineering #mcp #opensource

5 min read

Shinsuke KAGAWA

Sep 4 '25

Bringing Claude Code’s Sub-agents to Any MCP-Compatible Tool

#webdev #mcp #ai #tutorial

6 min read

Shinsuke KAGAWA

Aug 29 '25

How I Built an Image Generation MCP Server with Gemini 2.5 Flash Image (aka nano-banana)

#webdev #ai #typescript #tutorial

5 min read

Shinsuke KAGAWA

Aug 23 '25

Zero Context Exhaustion: Building Production-Ready AI Coding Teams with Claude Code Sub-agents

#ai #claudecode #programming #productivity

9 min read

DEV Community

Shinsuke KAGAWA

Badges

One Year Club

2 Week Community Wellness Streak

1 Week Community Wellness Streak

Writing Debut

GitHub Repositories

What Retrieval Still Hasn't Decided

Want to connect with Shinsuke KAGAWA?

I Found Three Broken Rules in My Own AGENTS.md

Switching from GPT-5.6 Sol to GPT-6 Astra: Start with Medium Effort

Was the Monorepo Really Best for Coding Agents?

Reasoning Effort Is Not a Quality Setting

When Better Models Make Old Agent Workflows Worse

A second pair of eyes for Claude Code: building Galley, a local runner that checks the work before the PR opens

I Built a Skill Reviewer. Then I Ran It on Itself.

Same Framework, Different Engine: Porting AI Coding Workflows from Claude Code to Codex CLI

Letting LLMs Jump — and Then Verifying Ruthlessly

Design Integration Checkpoints Before Letting LLMs Code

Planning Is the Real Superpower of Agentic Coding

Stop Guessing If Your Prompt Is Better

Stop Putting Everything in AGENTS.md

Why LLMs Are Bad at "First Try" and Great at Verification

Building a Local RAG for Agentic Coding: From Fixed Chunks to Semantic Search with Keyword Boost

How I Made Legacy Code AI-Friendly with Auto-Generated Docs

Taming Opus 4.5's Efficiency: Using TodoWrite to Keep Claude Code on Track

Stopping Cursor from Skipping Steps: A Structural Approach

Rethinking Team Development in the Age of LLMs

From 49 to 95: How Prompt Engineering Boosted Gemini MCP Image Generation

Bringing Claude Code’s Sub-agents to Any MCP-Compatible Tool

How I Built an Image Generation MCP Server with Gemini 2.5 Flash Image (aka nano-banana)

Zero Context Exhaustion: Building Production-Ready AI Coding Teams with Claude Code Sub-agents