The bait, then the rug-pull.
The promise is bold: a 24/7 AI employee doing work for you around the clock. But the real argument is quieter — Hermes Agent is not self-organizing, and most people using it are getting 10% of its value because they treat it like a chat window. Six use cases later, the case is made.
Where the time goes.
01 · Hook and intro
Bold claim, pain framing, promise of 6 transformative use cases.
02 · Use case 1 — /goal with meta-prompting
Meta-prompt first, then launch /goal. Detailed brief equals hours of autonomous build. Demo: Godot 3D shooter.
03 · Use case 2 — Kanban board morning routine
hermes dashboard opens Kanban. Drop daily tasks into Triage, go do human work, come back to finished output.
04 · Sponsor block
HubSpot AI Agents Cheat Sheet via Futurepedia — 7 agent tools, setup guides, copy-paste starter prompts.
05 · Use case 3 — Competitor technical research
Hermes browses a competitor site, inspects the console, builds full stack breakdown. Feed output to coding agents.
06 · Use case 4 — Memory wiki
Agent builds a self-hosted site of all past conversations and daily logs — diary for the user plus memory reinforcement for the agent.
07 · Use case 5 — Tailscale multi-device admin
Install Tailscale free on all devices; agent becomes administrator of the whole device network. Most-used use case for the host.
08 · Use case 6 — Morning priority prompts
Agent messages you at 9 AM, asks your top priority, generates tasks, updates memory. Compounds over time.
09 · /goal result reveal and CTA
Playable 3D stealth game with loot and enemies built autonomously during the video. Subscribe CTA.
Visual structure at a glance.
Named ideas worth stealing.
Meta-prompt before /goal
- Tell AI what you want to build
- Ask AI to write the /goal prompt
- Paste meta-prompt into /goal and launch
Ask the AI to write the task brief before launching a long-running agent run. Produces dramatically more detailed prompts than manual writing.
Kanban morning routine
- Write full to-do list
- Put agent-executable tasks in Triage
- Do human-only tasks
- Return to finished agent work
Separate agent work from human work at the start of each day. Triage auto-assigns to sub-agents.
Memory wiki setup prompt
One prompt: tell the agent to build a site cataloguing all past conversations and daily logs, clickable by subject or date. The agent builds the full site.
Morning priority prompt
- Every morning at 9AM ask me my number one priority
- Come up with tasks to help that priority
- Update memories about me accordingly
Daily check-in that compounds agent context over weeks. Better context leads to better recommended tasks.
Lines you could clip.
"If you do slash goal and just say build me an app, you might as well not written slash goal."
"The CEO of all your devices is what it turns Hermes into."
"The more you do this, the more your agent will self improve and be custom for you."
"You wake up, go to your computer, open up your Kanban board, get all your to do items for that day, and start putting them in triage."
Things they pointed at.
How they asked for the click.
"If you learned anything at all, leave a like down below. Make sure to subscribe to the channel."
Brief, sincere, no hard sell. Host expresses genuine gratitude before the ask.
Word for word.
Six habits that turn an AI agent into a real employee
Hermes Agent degrades to a chatbot when given vague prompts; structured daily inputs — a meta-prompted goal, a task board, a memory system — are what separate a power user from someone who forgot the tool exists.
- /goal only unlocks its potential with a meta-prompted, detailed brief — ask the AI to write the goal prompt before you run it, not after you see weak output.
- The Kanban morning routine separates work by who should do it: agent tasks go in Triage at the start of the day, human-only tasks stay on your list.
- Competitor research via browser control produces a full technical stack breakdown — stack, pricing, analytics events — in one prompt, exportable to Markdown.
- A self-hosted memory wiki serves double duty: a searchable diary for you and a long-term memory reinforcement layer for the agent across sessions.
- Tailscale (free) unifies all your devices into one private network, letting the agent retrieve files, run LLMs, and test local builds on machines you are not sitting at.
- Daily morning priority prompts compound: each answered question makes the agent's context more specific to your actual work, not generic AI defaults.
- Agentic tools are not self-organizing — the structure you impose (rituals, boards, wikis) is what makes the difference between a tool you use daily and one you abandoned.































































