Architecture decisions from running a real MSA in production
The build & ops log of an AI-agent algorithm-study platform built by a solo developer. From 12-agent orchestration, k3s ops, CI/CD, and ADRs to production debugging — the problems and decisions behind a service with real users.
Updated 2026-06-10 · 216 ADRs
Recent Posts
- Retrospective
The Art of Deletion
- AI Agent
Which Model, When, Where — A Strategy for Targeted Premium Deployment
- Platform
I Fixed an Outage That Never Happened — When a Static Manifest Judges Runtime
- Retrospective
I Chose to Spend More Tokens — Building an ADR View for Humans
- AI Agent
Memory Fades, Documents Pile Up — A Sliding Window for Agent Context
- AI Agent
Toward a Model-Agnostic Harness — Adopting a Critic
- CI/CD
AlgoSu CI Refactoring
- Platform
Baekjoon Is Gone?
- Backend
Why Does My Session Keep Dropping? — Debugging 4 Layers of Desynchronization
- Retrospective
Looking Back on 67 Sprints
- CI/CD
Safety Net for AI Code — Building 15 CI/CD Jobs in Practice
- AI Agent
How to Control 12 AI Agents
- Architecture
MSA Design: Humans Decide, AI Executes
- AI Agent
AI Agent Orchestration in Practice