DEV Community

AI Agent Harness Engineering Benchmarks Series' Articles

Back to tappers's Series
I Benchmarked 41 AI Agent Harness Patterns in Isolated VMs: How to Cut Tokens by 90% Without Losing Accuracy

I Benchmarked 41 AI Agent Harness Patterns in Isolated VMs: How to Cut Tokens by 90% Without Losing Accuracy

Comments
22 min read
How to Benchmark AI Agents Without Eval Cheating or API Key Leaks: Designing a Proxmox & systemd Sandbox

How to Benchmark AI Agents Without Eval Cheating or API Key Leaks: Designing a Proxmox & systemd Sandbox

Comments 1
14 min read
Stop Your Coding Agent From Stepping on the Same Rake Twice: Cold-Start Exploration vs. a 5-Item MEMORY.md

Stop Your Coding Agent From Stepping on the Same Rake Twice: Cold-Start Exploration vs. a 5-Item MEMORY.md

Comments
8 min read
Why Feeding Full Dependency Files to Your Coding Agent Hurts Accuracy: Full Source vs. Tree-sitter Interface Skeletons

Why Feeding Full Dependency Files to Your Coding Agent Hurts Accuracy: Full Source vs. Tree-sitter Interface Skeletons

Comments
7 min read
Stop Dumping Raw Pytest Logs Into Your Coding Agent: How a Structured JSON Test Filter Cuts Input Tokens by 94.1%

Stop Dumping Raw Pytest Logs Into Your Coding Agent: How a Structured JSON Test Filter Cuts Input Tokens by 94.1%

Comments
6 min read
Why Keeping Failed Code Attempts in Chat History Poisons Your Coding Agent: Error-Trace Pruning vs. Accumulated History

Why Keeping Failed Code Attempts in Chat History Poisons Your Coding Agent: Error-Trace Pruning vs. Accumulated History

Comments
6 min read
The Hidden "Schema Tax" of MCP Tools: Dumping 32 Tool Schemas vs. On-Demand Tool Search (-92.7% Input Tokens)

The Hidden "Schema Tax" of MCP Tools: Dumping 32 Tool Schemas vs. On-Demand Tool Search (-92.7% Input Tokens)

Comments 2
6 min read
How Save-Game Context Compaction Fixes Spec Drift in Multi-Step Coding Agents (-91.9% Input Tokens)

How Save-Game Context Compaction Fixes Spec Drift in Multi-Step Coding Agents (-91.9% Input Tokens)

Comments
6 min read
Benchmarking the Claude Code /advisor Pattern: Solo Executor Loops vs. Upfront Advisor Blueprint Routing

Benchmarking the Claude Code /advisor Pattern: Solo Executor Loops vs. Upfront Advisor Blueprint Routing

Comments
7 min read
Single Loops vs. Directed Graphs for AI Coding Agents: Cutting Input Tokens by 74.4% Without Losing Accuracy

Single Loops vs. Directed Graphs for AI Coding Agents: Cutting Input Tokens by 74.4% Without Losing Accuracy

Comments
6 min read
Why Role-Based Subagents Slow Down Coding Agents: Subagent Folding + Progressive Skill Disclosure (2.66x Faster)

Why Role-Based Subagents Slow Down Coding Agents: Subagent Folding + Progressive Skill Disclosure (2.66x Faster)

Comments
6 min read
Why AI Agents Write Weak Tests for Their Own Code: Single-Agent TDD vs. Two-Agent Blind TDD (0% vs. 100% Mutant Kill Rate)

Why AI Agents Write Weak Tests for Their Own Code: Single-Agent TDD vs. Two-Agent Blind TDD (0% vs. 100% Mutant Kill Rate)

Comments
5 min read