My Waterloo Intern went back to school. I'll miss him dearly but here's how I replaced him with Hermes

I was talking to Mike Bird, who podcasts about AI tools. He asked me if I was using Hermes.

"Of COURSE! Everyone knows HERMES. Who doesn't love a good... Hermes!" I said, slyly googling under the camera.

Soon, I was having PTSD flashbacks to my experience with OpenClaw.

Setting that up felt like getting a Roomba to stop trapping itself under the couch. I eventually got it working out of pure spite.

Then I was terrified to type anything into it in case it shut down like a McDonald's ice cream machine in July.

But apparently, everyone at the Builders Club is running these new agents all day long. They’re hooked up to browsers, auto-organizing hackathons, and probably running the office dishwasher.

So I watched some videos, downloaded Hermes, and let it loose.

It’s basically Claude Code, if Claude lived forever and had access to my whole online life.

It can use a browser. Search. Click. Call tools. Create MCP servers…. CHANGE ITS OWN CODE…

The mobile UI is TERRIBLE, so I hooked it up to Discord. I text it from my phone; every message spawns a new conversation thread.

At first, I thought: “Cute. Another ChatGPT wrapper.”

Then I gave it Trello access. Not by editing config files. I literally asked it to “connect to my trello”. Then I asked it to organize my boards. Then I told it to clone my repo and just... implement the features on the to-do list.

It stopped feeling like a chatbot and started feeling like I had an intern again, except I only pay this one $3 a day in DeepSeek credits.

Now I have it handling:

• Deep-dive research reports
• Meeting summaries
• LinkedIn coaching reports
• editing photos, creating motion design videos
• fixing downed websites

I even accidentally fell into the "Second Brain" cult. Turns out, it's not a mystical state of productivity. It's literally just an folder where the agent dumps whatever weird report I ask it to make. It syncs to Obsidian on my phone using SyncThing.

“Please watch 100 youtube videos on pizza making, transcribe them, then create a textbook with at least 150 pages, written in the style of Dave Barry. Include grounding in academic research. Make no mistakes.”

The UI layer for all of this is still a massive opportunity. Right now, interacting with these agents feels like playing command line roulette.

But the shift is real. You don’t need to use a computer any more. There is just one app, it writes itself, and the only hard part is learning to manage your army of agents through the godawful UX

Honestly, I’m having trouble figuring out what else to use it for. Any ideas?