AI Developer Accelerator
Log In
Community
Classroom
Calendar
Members
Leaderboards
About
Log In
5
Patrick Chouinard
4d •
General discussion
AI Developer Accelerator — Coaching Call - April 14th
AI Developer Accelerator — Coaching Call - April 14
VIEW RECORDING - 82 mins (No highlights)
Meeting Purpose
A coaching call for AI developers to share progress, discuss challenges, and exchange insights.
Key Takeaways
Enterprise AI Strategy: For Microsoft shops, Azure AI Foundry is the recommended path over Copilot Studio, which is seen as a legacy tool. The new Microsoft-Anthropic partnership is expected to bring more advanced, managed agents to the platform.
Agentic Development: Ty Wells demoed a full user-driven development loop: narrated feedback → AI plan generation → mobile PR review and merge. For RFP scraping, Paul Miller suggested analyzing public meeting minutes for early project signals.
Idea Generation: To manage the overwhelming number of AI-generated ideas, the group recommended using Andrej Karpathy's "LLM Wiki" prompt to structure conversations in Obsidian, enabling the AI to help prioritize.
Community Project: Patrick Chouinard is building an open-source project to create a "digital intelligence" from the community's collective knowledge, with a demo planned for the next few weeks.
Topics
Enterprise AI Strategy: Microsoft vs. Google
Microsoft Stack:
Copilot Studio: Not recommended for new agent builds. Its Power Automate-style interface is considered a legacy paradigm.
Azure AI Foundry: The preferred path for enterprise development, offering more power and model choice.
Anthropic Partnership: The new integration of Claude into Copilot is expected to bring more advanced, managed agents to the platform, making it a space to watch.
Google Stack:
Gemini Pro: Excellent for consumer use cases.
Gemini Enterprise: Currently has significant feature parity issues and limitations compared to Pro, making it difficult to use for serious enterprise work.
Agentic Development & Use Cases
User-Driven Development (Ty Wells):
Workflow: User narrates feedback in-app → AI generates a plan → Developer approves/reviews PR via Telegram.
Key Feature: The recording widget captures console logs, providing critical debugging context.
RFP Scraping (Alex Rojas):
Goal: Build an agent to scrape government RFPs ("tenders") in Mexico.
Insight (Paul Miller): Analyze public meeting minutes for early project discussions to get a head start on opportunities.
Recommendation (Patrick Chouinard): Use a managed agent service (e.g., Anthropic's) for projects with legal implications, as self-hosted agents carry higher risk.
Event Management Platform (David):
Functionality: AI-powered platform for event planners.
Process: User describes event → AI suggests vendors → Platform sends automated RFQs and negotiates.
Demo: Scheduled for next week.
Tools & Techniques
LLM Wiki (Andrej Karpathy's Gist):
Purpose: Structure raw AI conversations into a navigable wiki in Obsidian.
Benefit: Enables the AI to help prioritize ideas and find market value.
Patrick's Use Case: Using a modified version as a preprocessor for RAG databases to enrich metadata before chunking.
Ghost Database:
Description: A free, CLI-only Postgres database designed for agents to programmatically create, load, and destroy schemas.
Alternative (Patrick Chouinard): For portable RAG, LensDB is preferred as it can be packaged directly with an application.
Tablet Peripheral Solution (Ty Wells):
Problem: Avoid expensive, proprietary hardware peripherals (e.g., signature pads).
Solution: Use a low-cost tablet (e.g., Fire tablet) to take over a session for specific tasks (e.g., signature capture, check-in).
Business & Career Insights
Finding High-Value Use Cases:
Method: Instead of asking for problems, ask clients about their most painful, time-consuming, and unavoidable tasks.
Rationale: Solving these pain points creates immediate value and builds trust.
The New SDLC:
Shift: Developers are becoming "markdown developers," focusing on strategy and managing AI agents rather than writing code.
Example: Patrick's team used Claude Code as a "USB stick" to transfer context between developer instances, solving a bug via inter-agent communication.
Presidential AI Challenge (Elijah Stambaugh):
Elijah and his son won the Ohio state challenge.
Project: An N8N automation with Google Gemini 2.5 Flash that creates personalized lesson plans from a standard plan and a student's interest inventory.
Status: Pitching for the regional win, with a potential trip to the White House and a $30,000 prize ($10k each for Elijah and his son, $10k for the school).
Next Steps
David: Demo the event management platform next week.
Morgan Cook: Demo the completed NDA project once cleared.
Alex Rojas: Evaluate Anthropic's Managed Agents for the RFP scraping project.
Patrick Chouinard: Demo the community intelligence project in the next 1-2 weeks.
Patrick Chouinard: Schedule Brendan for a future call and announce it on the community site.
Action Items
Review Anthropic Managed Agents for RFP/tender use case -
WATCH (5 secs)
Present RFQ/negotiation system demo next week -
WATCH (5 secs)
Set up Obsidian LLM Wiki w/ Andrej Karpathy gist -
WATCH (5 secs)
Clear NDA; share released project w/ group next week -
WATCH (5 secs)
Review Ghost CLI Postgres for agents -
WATCH (5 secs)
Like
2
0 comments
5
AI Developer Accelerator — Coaching Call - April 14th
AI Developer Accelerator
skool.com/ai-developer-accelerator
Master AI & software development to build apps and unlock new income streams. Transform ideas into profits. 💡➕🤖➕👨💻🟰💰
11.2k
Members
43
Online
8
Admins
JOIN GROUP
Leaderboard (30-day)
1
Tom Welsh
+5
2
Burhan Pithawala
+3
3
Lydia Wilmsen
+3
4
Juancho Torres
+2
5
Arcchippus Inkum
+1
See all leaderboards
Powered by