A living technical series

Agent
Harness

Engineering reliable coding agents through structured context, executable tools, cumulative memory, external evaluation, and disciplined feedback loops.

Read the series
The premise The system around the model determines the ceiling.
Published 01 case study

Series index

Field-tested notes

Case studies and design patterns for agents that must improve continuously without losing reliability, traceability, or control.

Harness 01
Case StudyValidationMemoryMulti-Agent

How a Coding Agent Reached the Global Top 10 on AgentTank in Under 24 Hours

How structured priors, shared failure memory, replay-driven iteration, and rigid validation gates took an AgenTank coding agent from unranked to a global peak of #7.

July 12, 2026 10 min readEnglish / 中文