3 sharp questions. 1 brilliant win.

"Totally new to this — which AI model should I use to get Hermes running?"

Welcome to the room, Travis. New Mac mini, fresh subscription — perfect starting point. Here's the shortest path from zero to working Hermes:

• Step 1 — Install Hermes. Open Terminal, run the install command from the Hermes docs. About 60 seconds.
• Step 2 — Pick a model. Go with Owl Alpha on OpenRouter. It's free, it has 1M tokens of context, and it's built for agents. Zero financial commitment to start. Full setup is in The Goldie Sovereign Stack →.
• Step 3 — Test it. Run hermes in your terminal, type "hello" — confirm it responds. That's your "I'm in" moment.
• Step 4 — Stop trying to learn everything at once. Pick one job for Hermes (writing first drafts, summarising your inbox, drafting outreach), use it daily for a week, then add the second job.

The mistake most newcomers make is reading 20 hours of docs before starting. Don't do that. Run a real prompt within 10 minutes of install. The learning happens through use, not preparation.

If you want a 30-day path from this moment to a fully wired Agent OS — every step in order, no guesswork — that's what's inside AIPB. The four weekly coaching calls are the fastest way to skip the early stumbles too.

"Hermes vs OpenClaw — should we switch?"

Honest answer, Jeremy — they do different jobs, and the right play is to run both.

Here's how I think about it:

• OpenClaw is local-first, always-on, lives across WhatsApp / Slack / Discord. Great as the assistant that's just there when you need to ask something or capture a thought.
• Hermes is the agent harness — kanban-aware, tool-using, skill-extensible. Better for jobs that have a defined output (videos, content, research, outreach).

You're right that Hermes feels easier and breaks less in 2026 — Nous Research has been shipping rapidly and the agent loop is rock-solid. But OpenClaw still wins on always-on availability and cross-platform reach.

If you have to pick one for daily-driver work today, pick Hermes. If you've already got OpenClaw running and it's stable, keep it for the ambient/notification surface and add Hermes for jobs that need an agent loop. Both live in the Agent OS sidebar next to each other — no need to choose.

Full picture on Hermes specifically: The Goldie Hermes Five → covers the five use cases where it punches hardest. For Hermes + computer use (the new free feature): The Goldie Shadow Worker →.

"Hermes for video generation — easiest cloud setup, no hardware?"

Olga — great question, and the answer changed recently. There are now three good paths depending on what kind of video you're making:

• HyperFrames (free, HTML-first) — best for explainer videos, landing-page heroes, faceless YouTube. Hermes drives the renderer directly. Free, no external video model needed. Output: deterministic mp4. See The Goldie 4-Hour Rank → for how I rank with these.
• Higgsfield / Seedance via MCP — for AI-generated motion video where you actually need a model to imagine scenes. Seedance is cheaper than Higgsfield right now. Connect via MCP and Hermes can call it as a tool. Pay per generation, no fixed cost.
• HeyGen / Synthesia avatars — for face-to-camera style content without a face. Hermes drives the script, the avatar service handles the render.

Cheapest cloud path: run Hermes itself on a Hetzner CPX31 VPS (~$11/month) and call out to whichever video service you need. You don't need any local hardware — Hermes orchestrates, the video service renders, the mp4 lands back in your Agent OS workspace.

For the "absolutely easiest setup" you asked about: start with HyperFrames. No external API. No model selection. No per-generation fees. Free, on whatever machine you already have. Get one video shipped end-to-end before you add the more expensive options.

And yes — there's more native integration with other video tools now than there was last month. The video-model MCP ecosystem is maturing fast. Full setup details for all three paths are inside AIPB.