DEV Community

# agents

Posts

đź‘‹ Sign in for the ability to sort posts by relevant, latest, or top.
Human-in-the-Loop Agent Systems in Java

Human-in-the-Loop Agent Systems in Java

Comments
6 min read
Why Your AI Agent Breaks After Week Two (And How to Fix It)

Why Your AI Agent Breaks After Week Two (And How to Fix It)

Comments
2 min read
ARC-AGI V3 Explained: The New AI Benchmark That Breaks Every Agent

ARC-AGI V3 Explained: The New AI Benchmark That Breaks Every Agent

Comments
3 min read
The $12 Cost Tracking Bug That Inverted My Score/$ Comparison

The $12 Cost Tracking Bug That Inverted My Score/$ Comparison

1
Comments
6 min read
Designing an MCP Registry for Agents, Not Just Humans: Metrics, Experiments, and a Call for Collaboration

Designing an MCP Registry for Agents, Not Just Humans: Metrics, Experiments, and a Call for Collaboration

1
Comments
2 min read
The Amnesia Tax: Why Stateless Agents Are Eating Your Margins

The Amnesia Tax: Why Stateless Agents Are Eating Your Margins

Comments
3 min read
Beyond the Hype: Building Practical AI Agents with Memory and Reasoning

Beyond the Hype: Building Practical AI Agents with Memory and Reasoning

Comments
5 min read
Beyond the Hype: Building Practical AI Agents with Memory and Reasoning

Beyond the Hype: Building Practical AI Agents with Memory and Reasoning

Comments 1
5 min read
Your Nginx Logs Are Your Sales Pipeline (In the Agentic Economy)

Your Nginx Logs Are Your Sales Pipeline (In the Agentic Economy)

1
Comments
5 min read
GPT-5.1 scored 26%. Gemini 3 Flash scored 74%. Same prompt, same tools.

GPT-5.1 scored 26%. Gemini 3 Flash scored 74%. Same prompt, same tools.

Comments
8 min read
The Context Window Is the New Memory Architecture

The Context Window Is the New Memory Architecture

Comments
4 min read
Taming MCPs, Skills, and Agent Chaos with ToolHive

The dotfile hell of AI tool sprawl

Taming MCPs, Skills, and Agent Chaos with ToolHive

11
Comments 5
2 min read
AI Evaluation Is Now a Capital Expense

AI Evaluation Is Now a Capital Expense

2
Comments 1
2 min read
OpenClaw's Credential Problem Is Structural, Not Incidental

OpenClaw's Credential Problem Is Structural, Not Incidental

Comments
3 min read
We Scanned 4,275 MCP Servers. Most of Them Shouldn't Be Trusted.

We Scanned 4,275 MCP Servers. Most of Them Shouldn't Be Trusted.

1
Comments
6 min read
đź‘‹ Sign in for the ability to sort posts by relevant, latest, or top.