19d (edited) • General discussion
AI Developer Accelerator — Coaching Call - August 4th
AI Developer Accelerator — Coaching Call - August 04
Meeting Purpose
Key Takeaways
- Ty Wells's "Terrain" system creates custom UIs for existing apps in minutes by applying a front-end skin over their APIs, enabling rapid, personalized user experiences without touching core code.
- Morgan's Python driver revived a legacy Dymo printer in 2 hours, using AI to find the protocol and process print-test photos to auto-calibrate label boundaries.
- Patrick's team knowledge system uses Git as an agentic memory, hiding complexity from users and using Claude for "Fastlane Merging" of knowledge artifacts.
- Juan's AI photo booth is gaining traction, with an event coordinator becoming an impromptu salesperson; the immediate challenge is scaling operations and securing funding.
Topics
Business & Strategy
- Event Success: A recent event generated high engagement, with an event coordinator becoming an impromptu salesperson for the booth.
- New Lead: Secured a lead from a country club operations manager in Indio, CA.
- Marketing: Finalized a promo video.
- Scaling Challenges:
- Logistics: Transporting equipment (e.g., LED floor) in a Honda Pilot is a bottleneck.
- Funding: Exploring seed funding, possibly via Y Combinator.
- Research Y Combinator's YouTube channel for pitch guidance.
- Partner with industry advocates (e.g., event planners, venues) to systematize the pipeline.
- Business Model: Runs corporate workshops on CloudCode for ~$2k/session.
- Strategy: Focuses on a "vertical" approach, training one "AI Champion" per department to build automations.
- Revenue Model: Aims for recurring revenue via follow-up calls (e.g., 3 months at $X/month).
- Deal Closing (Real Estate Client):
- Challenge: A deal is stalled after 5 calls, despite the client sharing sensitive financial data.
- Paul's Advice: Define a paid "Phase 1" proof-of-concept (POC) to secure a smaller, easier-to-approve commitment. This de-risks the engagement for both parties and tests the client's operational readiness.
AI-Driven Development
- Concept: A front-end "skin" that provides a custom UI for any existing application without modifying its core code or APIs.
- How it Works: The skin sits on a separate server, consuming the original app's API endpoints to render data in a new, more intuitive way.
- Example 1 (Zendesk Replacement): Reskinned a ticket system into a 3D office layout, showing ticket status and employee locations at a glance. Built in ~30 minutes.
- Example 2 (Uptime Kuma): Applied a custom skin to the open-source monitoring tool.
- Use Cases: Multi-tenancy, personalized dashboards, modernizing legacy apps.
- Problem: A Dymo printer was unusable due to a lack of modern drivers.
- Solution: Used AI to find the printer's protocol documentation and write a new Python driver in 2 hours.
- AI Assistance: Fed print-test photos to the AI, which analyzed them to auto-calibrate label boundaries and resolution.
- Result: A fully functional CLI driver that prints text and images via UDP.
- Goal: Create a team-level knowledge base that acts as an agentic memory, not a queryable database.
- Architecture: Uses Git (Azure DevOps) as the backend.
- Individual Repos: Each team member has a repo.
- Team Repo: A central repo aggregates all knowledge.
- CI Pipelines: Push changes from individual to team repos.
- Hidden Complexity: Git operations are abstracted into natural language commands (e.g., "publish to team").
- Fastlane Merging: Claude performs automatic merges for non-conflicting knowledge artifacts.
- Conflict Resolution: Claude summarizes true conflicts for a human owner to resolve.
AI Tooling & Workflow
- Morgan: Finds Opus 5's responses "strange" and narrow-focused.
- Patrick: The key is to describe the problem broadly and clearly state the goal, but avoid over-specifying the solution. This allows the model to leverage its full capabilities.
- Ryan: Confirmed Codex is effective for security reviews, catching issues he would have missed.
- Patrick: Explained the workflow: An official OpenAI plugin hooks Codex into CloudCode. After a coding task, CloudCode sends the output to Codex for an adversarial review, then reads the report and acts on the feedback.
Next Steps
- Research Y Combinator's YouTube channel for pitch guidance.
- Explore partnerships with event planners and venues to scale the business.
- Hire a media person to film workshops for professional LinkedIn content.
- Send the draft "Phase 1" POC contract to the real estate client to formalize the engagement.
- Follow up with the retailer client regarding the digital screen project.
- DM Juan with additional scaling ideas.
Action Items
- Email retailer re: digital SEL screens; offer to include Ryan on call - WATCH (5 secs)
- DM Juan re: AI Booth scaling/funding ideas - WATCH (5 secs)
- Email real estate client Phase 1 contract; request signing authority; schedule demo - WATCH (5 secs)
- Write custom Mac scanner software for network all-in-one printer - WATCH (5 secs)
0
0 comments
AI Developer Accelerator — Coaching Call - August 4th
skool.com/ai-developer-accelerator
Master AI & software development to build apps and unlock new income streams. Transform ideas into profits. 💡➕🤖➕👨💻🟰💰
Leaderboard (30-day)
1
+4
2
+3
3
+3
4
+2
5
+2
Powered by