ブログ
PieBox チームによる技術インサイト、製品アップデート、エンジニアリングストーリー。
Claude Code vs PieBox: Code Assistant vs Product Delivery Platform
Claude Code and PieBox represent two agent philosophies: deep single-model integration versus flexible multi-model orchestration. It's not about which is better — it's about which fits your context.
PieBox — An AI Coding Agent Built for DeepSeek
DeepSeek offers incredible cost-performance, but using it as a coding agent isn't just about switching the API endpoint. Here's why we chose it, what we optimized, and what we got right.
From the Codex Paradigm to Product Delivery: What PieBox Built
PieBox deeply studied the Codex paradigm and other open-source Coding Agents, but we rewrote the model layer, added deployment, and built an API Hub service marketplace. Here's what we built and why.
5 Hard Lessons from Building an AI Coding Agent
After a year of building an AI coding agent, we've accumulated more battle scars than features. Here are 5 lessons paid for in real engineering hours.
Skills: Designing a 'Professional Certification' System for AI Agents
A Skill isn't a prompt template. It's a mechanism that lets an agent switch its thinking mode and behavioral boundaries for specific scenarios — and know when to switch back.
Building an AI Coding Agent: 5 Key Design Decisions
From first principles, we share the core design decisions behind PieBox's AI coding agent — from tool orchestration to context management.