📰 Dev.to · Shimo

The third article in the ReAct quadrant series. ReAct may be a tool for the design phase, not for production. Separate design and operation, and the legitimate

A follow-up to the 4-quadrant framework for business AI. The industry has no positive name for the LLM Workflow Quadrant — and that vocabulary gap is what makes

AI agent governance is stuck in the 'put up a sign' phase. Plato described the problem 2,400 years ago. The solution isn't better signs — it's accountability ar

I ran a distillation pipeline on an autonomous SNS agent for 17 days and hit knowledge saturation. Self-improvement needed human approval to break through. A de

Three weeks of operating a 9B autonomous agent revealed three facets of 'how much freedom to allow' — self-modification gates, trust boundaries, and security co

How 40 years of game development memory techniques — Importance Scoring, LOD, Object Pooling — improved a 9B model's knowledge distillation pipeline.

Rebuilding a local 9B agent's insight feature revealed the same architecture hiding inside Claude Code: natural language instructions and deterministic code, la

I built an autonomous SNS agent with a local 9B model and no framework. 13 Markdown prompts define behavior, Python enforces safety. This is what I learned abou

Introduction I tried to automate a Magento-based e-commerce site with Playwright. Magento...

OpenClaw imploded. In January 2026, a security audit uncovered 512 vulnerabilities — 8 of them...

In a corner of the AI research report that arrives at 5 AM every morning, there was a term: "Active...

I brought up token economics with Gemini, and the conversation expanded into a comparison of Deep...

Have You Ever Seen a "Raw" AI? The ChatGPT and Claude we use every day answer our...

Auto-generating structured explanations for 397 questions — and the quality was...

Introduction - What Is Everything Claude Code? Everything Claude Code...