Joel Yi

TEDx Speaker 👑 OpenClaw | Hermes Agents 🏆 #1 AI Award Winner 2018 ⚡️Ex US Army Cyber Officer 🇺🇸 Top Secret Clearance at 19

[Joel Yi @aiclawbots]

Here's the most underpriced AI setup available right now

Setup:
1. Grab a free API key at https://t.co/8gWqbLoET1
2. Set base_url to https://t.co/DrAuAAuRtZ
3. Connect it to Hermes
4. Route your low-priority workflows there

You're not replacing your main model. You're running a second engine for free on everything that doesn't need heavy lifting.

$0 in additional monthly spend. Real reduction in token costs on your main account.

[Joel Yi @aiclawbots]

NVIDIA just opened 80 free AI models through their API.

Most people saw the announcement and did nothing.

What you actually do with this: route them directly into Hermes Agent.

You get free inference on Kimi 2.5, DeepSeek 3.2, GLM 5.1, and a dozen others. Free. Daily. No credit card.

Your Hermes workflows just got significantly cheaper to run. Today.

[Joel Yi @aiclawbots]

This OpenClaw bot finds restaurants with outdated menus, rebuilds the menu as a live web page, and mails the owner a postcard. All on autopilot.

How agencies are using this to land recurring contracts:

• Scrapes every restaurant in a city in real time
• Filters by review count, rating, and last menu update date
• Samples the brand colors and visual identity from existing materials
• Renders a branded, mobile-friendly menu hosted at a QR-accessible URL
• Writes a personalized postcard referencing a real dish and a real reviewer
• Mails it addressed to the owner by first name

The owner gets a postcard showing their own menu, rebuilt. Most call back the same week.

[Joel Yi @aiclawbots]

The reason most business owners aren't using OpenClaw yet isn't that they don't want to.

It's that the installation, the security setup, and the first workflow feel like too much to figure out on top of everything else.

The technical barrier is real. It's also a 2-hour problem if you have the right guide.

What's stopping you is the 2 hours, not the complexity.

[Joel Yi @aiclawbots]

The founders who get serious about OpenClaw in the next 90 days are going to look like geniuses by the end of the year.

Everyone else is going to spend 2027 trying to catch up to the infrastructure those founders built in 2026.

The window is not closing tomorrow. But it's also not open forever.

[Joel Yi @aiclawbots]

This OpenClaw bot closes $50k+ landscaping deals on autopilot.

How it works:

• Scans satellite imagery for homes with empty backyards in target zip codes
• Filters by lot size, sun exposure, and recent ownership change
• Pulls homeowner contact info from public records directly
• Renders a luxury pool dropped into their actual yard
• Calculates build cost and property value lift for that specific zip
• Mails a personalized before/after postcard with a QR code to book a consult

Every step from lead sourcing to outreach runs without anyone touching it.

The only human step is taking the call when it comes in.

[Joel Yi @aiclawbots]

The single smartest cost move for anyone running OpenClaw at scale: route your low-stakes workflows to local models.

Meeting summary? Local model. Support ticket categorization? Local model. Draft replies? Local model.

Complex reasoning, strategy, anything that needs nuance? That's when you bring in the cloud.

You're not choosing between local and cloud. You're using both for what each is actually good at.

[Joel Yi @aiclawbots]

I don't care what hardware you're running. If you're not using local models with OpenClaw, you're spending money you don't need to spend.

Here's the actual setup:

1. Download LMStudio
2. Ask your OpenClaw/Hermes agent what local model fits your hardware
3. Let it walk you through the install
4. Tell it to start routing smaller workflows to the local model

The privacy case alone is worth it. The cost savings are a bonus.

[Joel Yi @aiclawbots]

@NousResearch @Teknium

Proud to bring a 100+ Hermes Agent hackathon to Seoul with amazing builders 🇰🇷 🔥

[Joel Yi @aiclawbots]

Someone built an OpenClaw agent that SELLS pool installations on autopilot.

It finds $500k-$1.2M homes that don't have pools.

Renders one into their backyard using satellite imagery.

Mails the homeowner a before/after postcard.

The pool contractor does nothing until a warm lead calls back.

[Joel Yi @aiclawbots]

Hermes does something Claude can't do out of the box: it remembers.

Every call, every support ticket, every lead interaction gets logged to a knowledge base. Over time that base becomes the source of truth for the entire business.

You can point any model at it and ask anything about any client, any deal, any problem.

And it's 10x better than Claude at it.

[Joel Yi @aiclawbots]

Most people treat AI agents like another basic tool.

The people getting the best results treat it like a new hire in their first 90 days.

They onboard it properly. They give it context about the business. They tell it what good looks like. They check the outputs. They correct it when it's off.

Think of it like onboarding an employee, and it will become one.

[Joel Yi @aiclawbots]

The 5 Hermes workflows that save the most time per week for solo founders:

1. Daily brief (saves ~40 min/day)
2. Meeting prep (saves ~20 min/meeting)
3. Lead qualification (saves ~15 min/lead)
4. Support triage (saves ~2 hrs/week)
5. Weekly KPI report (saves ~90 min/week)

If you're doing any of these manually right now, you're working for your business instead of running it.

[Joel Yi @aiclawbots]

The underrated Hermes skill: The Humanizer.

It reads any draft and rewrites anything that sounds like AI. Not just filler words. The cadence, the structure, the openings.

And you absolutely should run every piece of client-facing content through it before it goes out.

Output goes from "clearly AI" to "sounds like a real person" in under 8 seconds.

[Joel Yi @aiclawbots]

People building with Claude are mostly building prompts.
People building with OpenClaw are mostly building systems.

Both groups think the other one is doing it wrong.

Neither is wrong. They're just solving different problems.

[Joel Yi @aiclawbots]

The OpenClaw use case I keep seeing go viral is always some version of the same thing:

An agent finds people who need something, renders proof of what they'd get, and reaches out with that proof automatically.

Pool bot. Menu bot. Roof inspection bot. Solar panel bot.

This pattern works everywhere, especially for service based businesses

[Joel Yi @aiclawbots]

Here's what nobody tells you about running Claude and OpenClaw together:

Claude handles the thinking. OpenClaw handles the doing.

You describe the goal to Claude, it maps out the steps, you paste those steps into OpenClaw, and it runs the whole sequence across your actual tools.

So instead of comparing these two beasts, combine them

[Joel Yi @aiclawbots]

9 things a well-built Hermes bot can do that your current VA can't:

1. Run 24/7 without a single sick day
2. Handle 12 conversations at the same time
3. Pull live Stripe data before answering a billing question
4. Remember every interaction across every channel
5. Escalate the right problems to you before they become fires
6. Draft email replies in your exact voice
7. Log everything to Notion automatically
8. Update your CRM without being asked
9. Send you a Sunday night summary so Mondays start clean

The question isn't whether to build one. It's what you build first.

[Joel Yi @aiclawbots]

Claude vs OpenClaw isn't really a comparison. They're different tools.

Claude is the brain. OpenClaw is the nervous system.

You talk to Claude. OpenClaw acts in the world.

Once that clicks, you stop thinking about which one to use and start building with both.