GLM 5.2 + AI Harness Gets Fable 5 Results (No Subscription Needed)
I ran a full SEO/AEO audit on an open-source model — GLM 5.2 through an agentic harness in the OpenCode desktop app — and the whole thing cost about $9. The same job on Opus would've run me closer to $40.
You don't need another AI tutorial. You need a community full of people who are actually building. Join the AI Captains Academy: https://skool.com/aicaptains
Ready to go deeper? Free 5-day email courses:
→ AI Foundations Sprint: https://foundations.aicaptains.ai
→ Command Line Bootcamp: https://bootcamp.aicaptains.ai
But the audit isn't really the point. The point is this: when your harness is good enough, you don't need frontier models anymore. And nobody's talking about that.
In this video I clone a multi-agent harness (a "rig" — a structured team of AI agents with instructions, skills, and quality gates), point it at a real prospect's website with one line, and let it run autonomously for about an hour. Eleven agents across six phases. It scrapes the site, checks it against free APIs, writes a 25–35 page report, and the design-reviewer agent even catches GLM 5.2 using the wrong color and re-spawns the builder to fix it. That's the difference between slop and actual quality from an AI team — and it works no matter which model you run underneath.
Then I build a second harness live from a template (a thesis-writer rig — 131,000 tokens, ~$1.44), and at the end the audit goes out to the prospect as a cold-email lead magnet. A free basic audit up front, a $500 full audit and $2,000/month implementation behind it.
The future belongs to builders who can create their own solutions. Stop depending on others' platforms and start building your own digital empire. Become an AI Captain.
🔗 Resources:
AEO Audit: https://github.com/AI-Captains-Academ...
OpenCode: https://opencode.ai/
Venice API (get $10 free credits): https://venice.ai/chat?ref=PA5RHk
Apify: https://apify.com/?fpr=jordanurbs
Thesis Writer Github: https://github.com/jordanurbs/thesis-...
⏱️ TIMESTAMPS:
0:00 What this is — GLM 5.2 + an agentic harness in OpenCode
0:30 Why it matters: you might not have frontier access forever
2:00 The lead-getter rig — 15 leads scraped for under $5
3:00 How the AEO audit harness is built (11 agents, 6 phases)
4:00 What it writes: reports, design, build, send
5:00 The setup wizard — answering the harness's questions
7:00 Pricing the offer ($500 audit / $2,000-mo implementation)
9:00 The run begins — exploring the site structure
11:00 Sub-agents at work: researching competitors
12:00 Generating the lead personas
15:00 The real point: you don't need Cloud Code to do this
16:00 Building a SECOND harness live from a template
20:00 Cloning the StorySystems template into a thesis-writer rig
22:00 "My gift to you" — the rig pushed to GitHub
23:00 The agent team in action — proofing each other's work
23:51 The design-reviewer catches GLM's wrong color
24:51 The score comes in: 47/100 (and my "I am a moron" moment)
26:00 Reading the deliverable — the prospect's AEO score is C
27:00 The offer: free audit as a cold-email lead magnet
27:44 The numbers: ~1 hour, fully autonomous, ~$9 total
28:45 Be sovereign over your AI — wrap-up
This is for fellow builders who want AI leverage without renting it from one company forever. If frontier access gets gated behind your ID someday — and it might — you want to know you can run open-source AI agents and still ship frontier-grade work. The model isn't the moat. The harness is.
Say hi on X: https://x.com/jordanurbs
Substack: https://jordanurbs.substack.com/
For the algorithm:
Glm 5.2
Glm 5.2 ai
Claude Fable 5
AI Harness Engineering
AI Harness
Harness engineering ai
What is harness in ai