How an AI Agent Works: The Loop Explained Step by Step
An AI agent doesn't respond in one shot: it works in a loop that repeats until the task is done. Here's how that loop works, jargon-free and step by step.
Library
All AI Coding Patterns
Technical articles: agents, architecture, tools and design decisions.
All articles
MCP vs API: differences and when to use each one
MCP doesn't replace your REST API: it wraps it so an agent can discover its tools on its own. Comparison and when to use MCP, API, RAG, or function calling.
What Is RAG in AI: A Simple Definition and What It's For
RAG means searching your documents and handing them to the model before it answers. What it is, what problem it solves, how it works, and when it isn't worth the trouble.
Why AI Agrees With You: The Probable, Not the Correct
Why AI agrees with you even when you're wrong: how it predicts the next token, why your question changes the answer, and how to ask better.
Checklist before launching an AI project (7 checks)
A no-hype checklist to decide on an AI project: use case, real cost, risk, mitigations, metric, stopping criteria and owner.
Feedback loop in AI agents: how to close the improvement cycle
A feedback loop is what makes an AI agent improve instead of just repeating. Deterministic vs. judgment-based (LLM-as-judge) and how to automate the cycle.
How to Choose an AI Consultant (and Spot a Guru)
A guide to choosing an AI consultant or vendor without falling for magic promises: the power of the NO, red flags, and the questions you should ask.
An AI adoption plan for your company: the step-by-step roadmap
How to adopt AI in your company without burning through the budget: six ordered steps, from building judgment to scaling. A realistic roadmap, with no guru promises.
AI proof of concept: how to frame one so you don't waste your time
How to set up an AI proof of concept (PoC) that actually decides something: a clear hypothesis, real data, a measurable success criterion and a fixed deadline. No hype.
AI Agent in an Infinite Loop: How to Prevent It
Why an AI agent gets stuck in an infinite loop and the defenses that teach the loop when to stop: iteration cap, stopping condition, no-progress detection, and feedback signals.
The ROI of AI: How to Measure If It's Worth It
How to measure AI ROI without the hype: real total cost, a baseline before you start, and when it is NOT worth investing.
Long-running AI agents: what you need before you let them loose
The engineering checklist for letting an AI agent run unsupervised for hours: acceptance criteria, budgets, sandboxing, and rollback.