https://hermes-agent.nousresearch.com/docs/user-guide/features/rl-training

Hermes is an evolutionary approach. Compared the DSPy agent model, which is programmatic / declarative.

DSPy for many of the use cases, is a lot of work (fine-tuning, debugging). DSPy excels at finance tasks or data analytics where you know all the details. But that is phase 2. Phase 1 is getting these details, and getting them fast. That is imho where Hermes comes in.

I can see myself using Hermes for

* Content tooling: auto-briefing, strategic foresight assessments.
* Intelligence analysis: method approaches to trend analysis or signals
* Individualization of content. OSINT, social media scraping, trend analysis.
* Data sensing, estimation on datasets, start simple
* Autoresearch (later), for example, for optimization solutions or performance debugging

#dspy #hermes #declarative #evolutionary #ai

The related top HN comment is also worth reading: https://news.ycombinator.com/item?id=47491023

"You're comparing [DSPy] downloads with Langchain, probably the worst package to gain popularity of the last decade. It was just first to market, then after a short while most realized it's horrifically architected, and now it's just coasting on former name recognition while everyone who needs to get shit done uses something lighter like the above two."

Preach! 🙌

#dspy #langchain #hackernews #genai #llms