AI Developer Accelerator
Log In
Community
Classroom
Calendar
Members
Leaderboards
About
Log In
5
Patrick Chouinard
2d •
General discussion
AI Developer Accelerator — Coaching Call - June 2nd
AI Developer Accelerator — Coaching Call - June 02
VIEW RECORDING - 112 mins (No highlights)
Meeting Purpose
Discussing advanced AI agent development, corporate adoption challenges, and best practices.
Key Takeaways
Corporate AI Requires Justifiable ROI: The initial "go AI" mandate is over. Companies now demand clear ROI, making token cost management and deterministic, auditable workflows (via skills) essential for adoption.
Skills Over Agents for Corporate Use: For client work, prioritize deterministic skills over unpredictable agents. Skills provide auditable, repeatable results, while agents are best used as personal development tools or as orchestrators for other agents.
Agentic Development Requires Validation: Tools like /goal are powerful but require precise validation criteria (e.g., a specific test case) to prevent token waste. Use them for quantifiable tasks, not subjective ones.
Strategic Tooling is Critical: Combine tools for optimal results. Use Claude for creative tasks (e.g., UI), Codex for detailed validation, and token-saving utilities like RTK (Rust Token Killer) to manage costs.
Topics
Corporate AI Adoption & ROI
The initial corporate mandate to "go AI" has shifted to a demand for measurable ROI.
Dmitry's Experience: A major bank contract for an agentic AML platform proves the value of a deterministic, auditable approach.
Methodology: Uses a "Technique Engine" with adversarial reviews and deterministic workflows (via the Dboss framework) to ensure high-quality, auditable output.
Strategy: Build deterministic Python code for web scraping instead of using an agent's browse function. This is more reliable and cost-effective, as the code only needs to be fixed once if a site's DOM changes.
Elijah's Client Strategy:
Initial Approach: Record conversations and screen shares to analyze business needs and build point solutions.
Community Advice:
Focus on Skills: Deliver deterministic skills for specific tasks (e.g., research, proposal prep) rather than full agents. This provides clear value and avoids the complexity of agents.
Engage End-Users: Work directly with the people doing the day-to-day work to identify real problems, as their needs often differ from executive directives.
Agentic Development & Best Practices
/goal for PRD Execution:
Requires a highly detailed prompt with explicit validation criteria.
Use Cases: Best for small, quantifiable tasks (e.g., building a specific tool), not large, multi-feature applications.
Recommendation: Use the GRILL skill by Matt Pokok to clarify PRDs and create a canonical dictionary of terms, which helps operationalize the project for an agent.
Claude vs. Codex:
Claude: Strong for creative tasks and UI design.
Codex: Excellent for detailed validation, code review, and finding subtle bugs.
Synergy: Use Claude for the initial creative draft, then Codex for rigorous validation.
Token Cost Management:
High token usage is a major concern. Rod's $20 Cursor plan was depleted by a single dashboard update.
Solution: Use token-saving utilities like RTK (Rust Token Killer) to optimize command execution and reduce costs without sacrificing quality.
Personal Projects & Updates
Ryan: Scaling a retail screen software business (10 screens deployed).
SEO Strategy: Used an SEO agent to generate ~300 localized marketing pages, which are now ranking well and expected to drive inbound leads.
New Projects: Scoping a real estate "AI second brain" and building a custom CMS for a taxi company, including an invoicing backend and a MailChimp-like email system.
Patrick: Building a Hermes-based DevOps agent for a home lab.
Role: Acts as the "eyes and ears" of the network, listening to alerts from Prometheus/Grafana and managing inventory.
Future Goal: Enable autonomous actions like resource allocation and machine resets after a "learning mode" period.
Maksym: Optimizing a lead-gen funnel for an automotive sales tool.
Goal: Increase the test drive conversion rate from 37% to 50%.
Product: A SaaS platform providing WhatsApp chatbots with sales documents, financing data, and media content.
Next Steps
Ola Oyo: Contact Scott Rippey on the community site for advice on migrating to Google ADK 2.0.
Rod Morrison: Share the RTK (Rust Token Killer) repository link in the chat.
Juan Torres: Investigate Proxmox as a virtualization platform for a home lab, using a mini PC instead of a noisy, power-hungry data center server.
Dmitry Avramenko: Contact Maksym Liamin offline to discuss building on WhatsApp and product-market fit validation.
Elijah Stambaugh: Focus client proposals on delivering deterministic skills for specific business problems, rather than full agents.
Patrick Chouinard: Package the RAG project for community use and address issues in the GitHub repo as time permits.
Action Items
DM Maksym re: offline chat -
WATCH (5 secs)
Like
0
0 comments
5
AI Developer Accelerator — Coaching Call - June 2nd
AI Developer Accelerator
skool.com/ai-developer-accelerator
Master AI & software development to build apps and unlock new income streams. Transform ideas into profits. 💡➕🤖➕👨💻🟰💰
11.3k
Members
46
Online
8
Admins
JOIN GROUP
Leaderboard (30-day)
1
Purendeeswar Reddy
+5
2
Tom Welsh
+4
3
Dmitry Avramenko
+3
4
Juancho Torres
+3
5
Pasquale Salomone
+2
See all leaderboards
Powered by