The ideal shore
The ideal shore
CategoriesTagsContactFriendsAbout
The ideal shore

Focuses on system deconstruction and toolchain composition. Exploring Go, React, Linux, and local AI agent architectures.

Navigation

  • Home
  • Categories
  • Tags
  • Contact
  • Friends
  • About

License & Stats

Articles are licensed under CC BY-NC-SA 4.0 unless noted.

Sitemap
PV: .../UV: ...

© 2026 KevinYouu. All rights reserved.

Designed withbyKevinYouu

K

KevinYouu

Senior Full-Stack Engineer

Building systems, not apps. Focused on Go, React, and Linux.

GitHub EmailTwitter Telegram

Runtime Config

config.yaml
# Focus on tools rather than apps
engine_focus:"Toolchain Over Products"
# Local-first, reject over-dependence
control_level:"Local-First & Independent"
# UI as a system interface, data-driven
ui_philosophy:"Data-Driven Interface"
# Lightweight VPS, Docker, SSH-driven
infra_mindset:"Reproducible Sandbox"
visitor@kevinyouu: ~ (zsh)
visitor@kevinyouu:~$

Tech Stack

Backend

Go
Echo
Node.js
pgx
PostgreSQL
SQLite

Frontend

React
Next.js
TypeScript
Tailwind CSS

Infrastructure & Operations

Docker
Linux
Nginx
CI/CD
Git
Alacritty

AI Engineering

Claude Code
codex
llama.cpp
BingoCode

Engineering Principles

Decouple

System Deconstruction

Deconstruct complex tasks into composable CLIs, pipelines, and local microservices.

Agent

Agent Workflows

Build sandboxed workflows using local models (Ollama, Qwen) and structured caching.

Local-First

Local First

Prioritize local data ownership and storage, rejecting over-dependence on proprietary cloud backends.

DevOps

Minimal DevOps

Prefers single-instance Docker and SSH/CLI remote management, rejecting bloated cluster abstractions.

Repro

Reproducible Env

Insists on reproducible runtimes and automated build pipelines, using scripts and CI/CD to eliminate manual steps.

STATUS: COMPOSABLE

bingo-code

FLAGSHIP AGENT RUNTIME

AI Coding Agent CLI

Code Scale
200K+ LOC
STATUS
ACTIVE R&D

An AI Coding Agent CLI built with Go + TypeScript + React/Ink. The goal is not to bind to a single model provider, but to build an Agent Runtime that can freely integrate different model suppliers for real software engineering work.

Go
TypeScript
React
Ink
Core Runtime Loop
Input / User Request→Context Compact & Cache→LLM Reasoning & Plan→Tool Execution→Sandbox & Rollback

Core Runtime & Models

Multi-model / Multi-providerAgent LoopContext CompactPrompt CacheAutomatic Retry

Agent Capabilities

Tool CallingSkillsSub-agentsCode Review AgentPlan / YOLO Mode

Developer Integrations

Git IntegrationWeb / CDP toolsImage Input

Sandbox & Experience

Sandbox & Permission ControlSQLite History & RollbackTUIToken / Cost Statistics

Things I'm Interested In

EXPLORATION

AI & Agent Runtime

AI Coding AgentsLLM Runtime & Tool CallingContext EngineeringPrompt CacheAgent Architecture

Developer Infra & Tools

Developer ToolsCLI / TUICode Intelligence

Evaluation & Open Source

Model EvaluationOpen Source
RADAR: ACTIVE TRACKING

Environment

SPEC SHEET
macOS
Primary Station (Apple Silicon)
Host OS
Linux / Debian
Build & Production Nodes
Server Nodes
Docker
Sandbox & Reproducible Stacks
Containers
Alacritty / Terminal
Immersive Terminal Workflows
Shell & TUI
Git
Version & Branch Tracking
Version Control
ENVIRONMENTREADY / DEV-OPS

Engineering Preferences

ENGINEERING PHILOSOPHY & DESIGN PATTERNS

7 PRINCIPLES
01Simple, explicit, and maintainable architecture
02Reduce unnecessary abstraction
03Evaluate performance with real data, not feelings
04Automated tests and rollback-ready design
05Local-first developer tools
06Provider-agnostic AI infrastructure
07Tools that actually solve problems, not tech for tech's sake
Agent System Pipeline
CLOSED-LOOP ARCHITECTURE
STEP 1Simple Protocol
›
STEP 2Small Context
›
STEP 3High Cache Hit Rate
›
STEP 4Precise Tool Usage
›
STEP 5Fast Feedback Loop
›
STEP 6Reliable Result

GitHub Contributions

2,522 contributions in the last year
Less
More