👑
⭐
2h • 💰 Competitions
🏆 WEEKLY COMP #4 RESULTS: THE AGENCY 🏆
I thought last week was hard to judge, This was the hardest comp to judge so far. Not even close.
Picking the top five took longer than picking the winner. Working apps. Animated explainers. Voice acting. Full Instagram brands. Three-contract architectures. Cloud automation that actually runs. People showed up this week.
Before I name the winners, I want to be honest: this was tight enough that a few of these slots could have gone the other way. The gap between 1st and 5th was smaller than the gap between 5th and the rest of the pack.
Let me run through the runners up first.
➖➖➖
🌐 Team portal: https://six8coffee.github.io/diana-calloway-real-estate-ai/
The only submission with real automation wired up. Two Cloud Routines running on Anthropic's servers. One processes the leads@ inbox hourly via Gmail MCP. The other sends a daily team briefing at 8am. Plus a hosted team portal, a confidence-and-trail audit log on every handoff, and a 6th specialist (Nurture Coordinator) for leads that aren't ready yet.
The "Jordan's day-one guide" framing in the README addresses a named hypothetical new agent and walks through the three daily tasks concretely. Real Austin context (Mueller neighborhood, 512 area codes) shows attention to brief details that a lot of submissions skipped.
➖➖➖
🌐 Live page: https://insightfulminds.github.io/cold-start/
Best diagrams in the comp. Best single-page packaging. The presentation is the most readable explainer of any submission this week. If you want to see how to communicate an ICM system clearly to a non-technical reader, study Sean's page.
Design choice that stood out: a back_to field in every handoff envelope so deals that stall (lender delays, inspection issues, financing problems) can route back to the right specialist as a normal operation. Plus a language flag baked into every handoff so the system already knows what language to draft in before the work starts. Real product thinking.
➖➖➖
🌐 Live demo: https://atxrealstate-production.up.railway.app/
He has submitted something every week that could/should win. I have already sat down to look at recruiting him for my main company.
Virgilio literally built a working application.
Not a mock. Not a concept. A Railway-deployed app with 16 real Austin listings (hero photos, galleries, prices), five named agents matching Diana's described structure (Diana, Elena, Carlos, Marco, Sara), Slack-as-interface design where the chat is the data entry, and a Guided vs Operational mode for ramping new agents. He shipped the actual thing Diana described needing.
Third week in a row Virgilio has shipped at this level.
He's not slowing down. . If you want to see what consistent compounding looks like in this community, watch his work.
➖➖➖
🌐 Live site: https://agency-os-tan-five.vercel.app/
Ruby was within an inch of taking it. And honestly was going to be the winner as I was looking through everyone's.
The Loom video she recorded for this submission is the cleanest voice work I've heard on any community submission. Full-on infomercial quality. Voice acting level. If you haven't watched it yet, go watch it. It's a master class in how to present a build.
Beyond the voice: an animated handoff explainer on the companion site, a WRITEUP.md that compresses the entire submission into three paragraphs of clean argument, and a design philosophy made explicit.
She refused typed schemas in favor of human-readable Handoff Cards with seven sections including a required Gaps field ("if Gaps is empty, you're not looking hard enough").
One continuous narrative thread runs through all 27 files of her repo.
Ruby and Ariel were close enough that the final call took longer than the rest of the judging combined.
➖➖➖
📸 Instagram: https://instagram.com/_agency_os
Ariel won because the whole package is the whole package.
Three root contracts (Handoff Protocol, Case State Schema, and the standout: an Emotional Intelligence framework drawing on Morgan Housel's Psychology of Money that formalizes how the same property data should land differently for an anxious first-time buyer, a relocating family, and an investor).
Seven specialists including a Daily Deal Desk for morning operational scans and a BD Coordinator for long-pipeline relationships.
Five end-to-end example flows.
But the ICM build alone wasn't the differentiator in my head.
It was everything around it.
A full landing site with FAQ, scenarios, guide, and market research pages.
An entire Instagram page, yes literally a brand new Instagram page.
A Calendly link on every corner. Ariel didn't submit a competition entry.
They submitted a product launch packaged inside a competition entry.
The Emotional Intelligence contract specifically is the kind of architectural choice I want to see more of in this community. It treats the human being in the transaction as signal the system should read which I think will be an ROI as time goes on.
Ariel takes the $325.
More importantly, they take the title of strongest builder in the community this week.
Remember In a world were AI is making it easier to build its the unique opinions and styles that become valuable.
➖➖➖
🎖️ Honorable mentions that deserve your time:
🛠️ — built a workflows/ persistent memory layer and shipped 10 actual test cases. Production-engineering instinct shows. 🎨
— added a 6th specialist for quality review and built a beautiful hosted onboarding page. 📺
— recorded an entire YouTube playlist explaining the build. Putting that level of teaching work into a comp submission is rare.
— nine specialists with structured handoffs plus a live page. Vendor intelligence and compliance as their own specialists was a smart call. 🏗️
— _shared/ canonical store plus Diana-as-product framing was top-tier architecture work. 📐
— cleanest typed JSON schemas plus conditional handoffs plus real database CSVs.
Every name above shipped work this week that would have won earlier comps. That's how steep the curve got.
➖➖➖
🚀 One last thing.
The brief said this comp was the on-ramp to The Lyceum. Multi-folder ICM, real handoffs, real architecture. After watching this work come in, I'm more convinced than ever that this community is ready to go deeper.
This week was the surface layer. The Lyceum is twelve weeks of building production-grade ICM systems with me walking you through the parts a six-day comp can't teach. Orchestration patterns. Context evaluation. Edge cases. What to do when the system breaks. How to scale it from one team to a hundred.
If this week was the on-ramp, the Lyceum is the highway, stayed tuned for updates as we build it out.
26
48 comments
🏆 WEEKLY COMP #4 RESULTS: THE AGENCY 🏆
skool.com/cliefnotes
Jake Van Clief, giving you the Cliff notes on the new AI age.
Leaderboard (30-day)
1
+991
2
+901
3
+815
4
+699
5
+426
Powered by