Hey guys! I just sat down with @Patrick Chouinard for one of the coolest deep dives we've done in the community. Patrick has become our community's go-to AI expert and one of the most helpful resources we have. He's quietly been building something most teams pay a vendor $50k a year for. He's turning every community call we've ever recorded (two and a half years of two to three hour conversations) into a queryable "community brain." You'll soon be able to ask it anything that's been discussed and get a real answer with citations. Here's the part that broke my brain. Standard RAG completely falls apart on transcripts. The question gets asked at minute 6, the conversation drifts, and the real answer shows up at minute 41. A normal chunker has no idea those two moments belong to the same idea. Patrick solved it by adding an LLM analysis layer BEFORE chunking that restructures the transcript into self-contained units of knowledge. You can watch the full breakdown above! Here's everything we covered: ✅ Why traditional chunk-and-embed fails on non-linear data ✅ The LLM analysis layer that turns raw transcripts into RAG-ready knowledge ✅ How Patrick picks the cheapest model that's still smart enough for each step (Kimi K2.5 for restructuring, Sonnet 4.6 for the signal extraction) ✅ Why he chose LanceDB over Pinecone (and when you'd flip that decision) ✅ Running the whole thing locally with Ollama, Open Web UI, Gemma 4 4B, and gpt-oss:20B for more complex retrieval ✅ Using Claude Code to build a custom chunker instead of fighting with a library ✅ Real cost math. About 40 cents per two hour episode and under $100 to process the entire archive The wildest part is the price. Once the embeddings are built, querying is free forever because everything runs on your own machine. No SaaS, no per-token cost, no IT review. Patrick is also planning to open source the full pipeline once a few rough edges are ironed out. So if you've been wanting to build something like this for your own team, agency, or client, you'll have a working blueprint to start from.

Took the plunge and deployed it on my own AWS infrastructure. For those who've been curious but haven't pulled the trigger yet - it's real and it works. My setup: - Isolated VPC with private subnet (no public IP) - Access via Telegram only - Zero exposed ports - SSM for admin - Encrypted storage, locked-down permissions First conversation hit and Claude responded through Telegram. Wild feeling having an agent just... waiting for me. Security was my main hesitation. Solved it by putting everything behind NAT with no inbound routes. The agent can reach out (APIs, Telegram) but nothing can reach in. If you're on the fence - the infrastructure side is more approachable than it looks. Happy to compare notes with anyone else who's deployed.

VIEW RECORDING - 180 mins (No highlights) Key Takeaways - Claudebot (now Moltbot) is a powerful agent, but requires strict security isolation. Run it in a dedicated VM with its own "service accounts" (email, calendar, GitHub) to prevent data leaks. - SOC 2 & HIPAA compliance is a major moat, but a huge time sink. It costs ~$20k/year and requires weeks of manual work, making it a key differentiator for SaaS businesses. - Productize your AI skills by building a portfolio. Create a few free projects to establish expertise, then offer them as a productized service (e.g., voice agents via LiveKit/Bland.ai). - Use AI for rapid prototyping and idea validation. Before coding, use tools like GSD (for deep planning) or a simple Claude Code sandbox (for quick tests) to refine concepts. Topics Claudebot (Moltbot) & Agent Security - Core Challenge: Claudebot's file system access is a major security risk. - Solution: Service Accounts & Isolation Dedicated VM: Run the agent in an isolated Ubuntu VM (desktop version for GUI apps). Service Accounts: Create a separate Google account for the agent with its own email, calendar, and Google Drive. GitHub: Give the agent its own GitHub account with read-only access to your repos. It works on clones and submits pull requests for review. Memory Vault: Instruct the agent to sync all internal thoughts to an Obsidian vault in its GitHub repo, providing a transparent "window into its brain." - Use Cases & Demos Daily AI News Digest: Patrick's agent uses Brave and Perplexity APIs to email a daily news summary, including source citations and token costs. Training Harness: A Claude Code skill that auto-documents development sessions, generating summaries and artifacts for creating courses. Remote Coding: Scott coded an entire app via WhatsApp, demonstrating the agent's ability to run tasks in the background on a remote machine.

Hey everyone! Quick update on how coaching calls are working going forward. 🔄 WHY THE CHANGE I'm going heads-down on EMS Soap for the next 6 months. We're getting SOC2 and HIPAA compliant, building out the marketing funnel, and a bunch more work. It's a lot, and I need to focus. But I don't want to abandon the community. So Patrick and Paul are stepping up to help lead calls. I'll be on about once every four weeks, and they'll rotate the other weeks. 📞 HOW CALLS WILL WORK These calls can run 2+ hours. I want to make sure we're respecting everyone's time. Especially those of you who actually show up. Here's the new structure: 👉 Reply to this post with your questions before the call 👉 If you submit a question and you're on the call, you go first 👉 We work through questions in the order they came in 👉 Then we open it up for everyone else If you can't make the call but want your question answered, drop it in the comments. We'll get to it. But priority goes to people who are there. The goal is simple: if you're taking the time to show up, you shouldn't have to wait behind questions from people who aren't even on the call. 🔗 NEW ZOOM LINK (save this) https://us06web.zoom.us/j/81995207847?pwd=Xe6u6LmIQOmCP5VTnOwWYjDBfZNKGB.1 📅 WHEN Tuesdays at 6PM ET (same as always) Looking forward to seeing you on the calls!

VIEW RECORDING - 165 mins (No highlights) Key Takeaways - Automated Training Creation: Patrick built a system that generates a full training course (NotebookLM, slides, video) from a single Claude Code project by auto-extracting patterns from the dev conversation. - Advanced RAG Architecture: Scott implemented a 3-layer RAG system in NeuralSpark, using a Recursive Language Model (RLM) to have the LLM intelligently "hunt" for information, solving context rot and improving accuracy. - Enterprise AI Consulting: A new business opportunity was identified: helping companies securely adopt agentic AI (e.g., Claude Code, Cowork) to bypass productivity-crippling internal IT policies. - Agentic Workflow Updates: Brandon is releasing new Shipkit videos and commands (e.g., worktree, task review) to optimize agentic development, enabling parallel feature work. Topics The Future of Work: System Creators vs. Task Doers - Morgan's Workflow: Automated a non-programmer's task (creating product sheets) from days to <10 minutes using skills in the Windsurf IDE. Process: Markdown → HTML (styled) → PDF + meta.json for product pages. - Brandon's Cowork Example: Resolved a complex Excel matching problem (50 parts x 50 centers) in 5 minutes by pasting the problem description and file into Cowork. - Consensus: The future workforce will be system creators and managers, not task doers. People must "level up" to think at a systems level, as AI will automate individual tasks.