Worth your time?
How to Make Hermes Agent 10x More Powerful (Memory & Skills) by The AI Architects | Tom Crawshaw
Watch on YouTube
Teaches: a hands-on, usable walkthrough that shows how to build memory-and-skill automations, with a visible sponsor and a modest paid-service plug.
Why this score
- Scored from the transcript, summary, and output signals.
The short version
This video shows how to turn Hermes from a one-off chatbot into a continuously improving, autonomous agent by configuring two features: memory and skills. Tom explains how Hermes' built-in memory files (memory.md and user.md), a second-brain file system (Obsidian or plain folders), reusable skills (/learn), and scheduled cron jobs combine so the agent remembers context, learns repeatable processes, and runs automations on a VPS 24/7. He demos an end-to-end automation that converts mentorship-call transcripts into Instagram reels and explains common pitfalls like bloated memories and messy skills.
What it teaches
- Build agents the same way you build teams: give persistent context (memory), codify repeatable processes (skills), and schedule/monitor work (cron). Use this when you want continuity across sessions and to scale repeating tasks. Do not use heavy memory for transient details or for things that exceed file/token limits.
- Convert workflows into skills only after you tidy the process. Apply the '/learn' action to automate a process you run regularly, then iterate V1->V2 based on failures. Avoid turning disorganized, ad-hoc work into a skill; automating a messy workflow replicates the mess and multiplies rework.
Key moments
- 0:24Why this matters: Tom identifies memory and skills as the single levers that turn Hermes from a disposable chat into a compounding assistant that improves with use.
- 1:35Why this matters: Demonstrates the practical reason to run Hermes on a VPS (Hostinger KVM2 example) so scheduled tasks and Telegram integrations run when your laptop is off.
- 3:30Why this matters: Explains the distinction between memory.md (session notes) and user.md (who you are), and why these should be concise "sticky notes" rather than a full journal.
- 8:39Why this matters: Shows the exact trigger to create reusable automation, use "/learn" after running a process so Hermes converts it into a repeatable skill.
- 10:02Why this matters: Introduces sub-agents so you can parallelize work (phase one, two, three concurrently) instead of running steps sequentially in one agent.
- 11:59Why this matters: Launches the demo that ties memory, skills, and cron jobs together, transcript -> idea extraction -> hooks -> scripts -> scheduled skill.
Skip the video. Keep the lesson.
More scored videos
