{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"NC Tweener Talks","title":"Robbie Allen @ Tweener Claw: \"3 to 5 of My Agents Fail Every Day: Here's What I've Learned\"","html":"<iframe width=\"100%\" height=\"180\" frameborder=\"no\" scrolling=\"no\" seamless src=\"https://share.transistor.fm/e/c4ba4d41\"></iframe>","width":"100%","height":180,"duration":1464,"description":"In this episode of Tweener Claw, Robbie Allen, the Founder and Managing Director of Automated Consulting Group and General Partner at NC Tweener Fund, shares a live talk from the June 10th Tweener Club meetup in Research Triangle Park.\nRobbie has spent the last year and a half deploying AI agents inside real mid-market companies, and he brings three lessons from just the last 60 days: how to build reusable AI skills that compound over time, how to monitor agents when 3 to 5 of your 25-plus will fail on any given day, and why using multiple LLMs to check each other's work has become a standard part of his workflow. Each lesson comes with a real-world example: a three-skill meeting pipeline, a monitor agent that watches the others, and a Codex Opinion skill that consistently finds gaps in Claude's output.\nThe sharpest takeaway: AI build costs have dropped dramatically, but support and maintenance costs have not. Non-deterministic systems are inherently brittle, and anyone telling you otherwise hasn't run them in production long enough. This is a practitioner's talk for practitioners; specific, candid, and immediately actionable.\nTimestamps\n01:31 Scot's intro \n02:03 60 days of lessons \n02:19 Three lessons preview \n03:08 Robbie takes the stage \n03:25 About Automated Consulting Group \n04:13 Reality vs. the AI hype cycle \n05:04 The Microsoft / Copilot rant \n07:13 Lesson 1: Skills that build skills \n08:41 Skills as reusable packages \n09:21 \"The sawdust of business\" \n10:58 Building a skills pipeline for meetings \n11:57 Skills security risks in the enterprise \n12:47 Lesson 2: Agents watching agents \n13:44 Build costs are down\n14:24 \"3 to 5 of my agents fail every day\" \n15:44 Log everything\n17:33 What causes agents to fail in the wild \n18:02 Lesson 3: LLMs checking LLMs \n18:13 The LLM Council concept explained \n19:44 The case for a multi-LLM strategy \n20:36 The \"Codex Opinion\" skill 21:33 Gemini Opinion & LLM Council in action \n23:46 Wrap-up & credits\n--- \nThis episode of...","thumbnail_url":"https://img.transistorcdn.com/UEzoK7N1siD9YRaSVBWfZ8B3suSS2aonEIZ5NwBH1Gs/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS80NjBh/YWVhZTA4ZTUyY2Fl/MDdhNzQwZWFhNDI4/MDc3Zi5wbmc.webp","thumbnail_width":300,"thumbnail_height":300}