# Hermes Learning Loop

> The way Hermes looks back after a turn and saves what it learned.

**Field:** Artificial Intelligence  
**Category:** Hermes Agent  
**Also known as:** Hermes background review, Hermes self-improvement

## Think of it like

It is like a diary you write each night.

## Definition

The learning loop is how Hermes Agent gets better from use. After a turn, a background review looks back at the chat in a separate copy of the agent. It saves small facts to memory and longer ways of working as skills. A short Memory updated line tells you when it saved something. The next session starts with what it saved.

## Example

Leo corrects Hermes twice on how he wants commit messages. The review saves the rule, and the next day it writes them his way.

## Why it matters

An agent that learns from corrections needs fewer of them as time goes on.

## Common mistake

Expecting the new fact to work in the same session. It loads at the start of the next one.

## Three principles

- **Runs after a turn.** In the background.
- **Facts and procedures.** Memory and skills.
- **Shows in the next session.** Not this one.

## Related terms

- [Hermes MEMORY.md](https://learn.benibix.com/term/ai/hermes-memory-md)
- [Hermes Skill](https://learn.benibix.com/term/ai/hermes-skill)
- [Hermes Agent](https://learn.benibix.com/term/ai/hermes-agent)
- [Agent Memory](https://learn.benibix.com/term/ai/agent-memory)

![The Hermes learning loop: a session runs, a background review saves small facts to memory and long procedures as skills, and the next session starts with both.](https://learn.benibix.com/diagrams/hermes-learning-loop.svg)

---

Source: https://learn.benibix.com/term/ai/hermes-learning-loop
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-17
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
