WTF Is an "AI Agent Loop"? Genius or Hype?

S/o Coderabbit for sponsoring today’s vid: https://coderabbit.link/greg On this episode I sit down with Professor Ras Mic to break down agentic loops. We define what a loop is, explain why well-known builders like Boris and Peter swear by them, and stay honest about who they truly serve. Mic argues that human-in-the-loop remains the strongest setup today, and he walks through the one loop he runs every day for code review using Cursor, GitHub, and Greptile. By the end you will know when a loop earns its place and when your own hand belongs on the wheel. Timestamps 00:00 – Intro 01:23 – What is a Loop 07:59 – /goal Explained 11:32 – The Slop Machine 12:42 – Code Review as a use case for Agentic Loop 18:19 – Honest Take for Builders 20:42 – The Future of Loops 21:50 – Closing Thoughts Key Points * A loop fires once from a human, then the agent generates, reviews its own result, and feeds it back to keep building. * Human-in-the-loop keeps you directing, governing, and approving each step while the agent builds. * Wide-open loops make heavy assumptions and burn serious tokens; Michael cites Peter's tweet about $1.3 million worth of tokens in one month. * Reserve slash goal and similar loops for the $200/month plan, since the $20 and $100 tiers burn through fast. * Loops shine in confined, fixed-feedback work: code review, SEO pages, and other binary tasks. * Mic’s daily win is a closed code-review loop with Cursor, GitHub, and Greptile that chases a 5/5 score. Numbered Section Summaries 1. What a Loop Promises I open by asking Mic what listeners will walk away knowing, and he frames the whole episode: a clear definition of the loop, the reason top AI voices obsess over it, an honest read on who it fits, and a real loop he runs himself. This sets up a balanced, honest take. 2. Diagrams and Human in the Loop Mic kicks off with his beloved stick-figure diagrams and lays out human-in-the-loop: I prompt an agent in Cursor, Claude, or Codex, review the result, and iterate. He uses a to-do app as the example — landing page first, then authentication, then the back end — with me steering every move. 3. How Agent-Run Loops Work Mic explains the approach Boris and Peter describe: the human fires the loop a single time, then the agent checks its own output, feeds the result back in, and continues from a spec.md or PRD.md task list. In theory this points toward the future of building. 4. Where Wide-Open Loops Go Sideways Mic paints a startup analogy: hire a brilliant developer, hand over the spec, and let them build the whole thing solo. The developer fills every gap with assumptions, those guesses drift from the product vision, and the token meter runs hot the entire time. He reminds me that every plan leaves edge cases, so a free-running agent tends to guess wrong and spend big. 5. The Token Bill and the Slop Machine Mic notes that Boris and Peter operate with unlimited token budgets, which makes constant looping rational for them. For everyone watching a budget, he calls these meta-harnesses a money burner, and I land on the phrase that sticks: loops can turn into a slop machine. 6. The One Loop Mic Runs: Code Review Mic shares his daily loop: Cursor as the harness, GitHub for source control, and Greptile as the code-review agent that scores each push out of five. His skill, grep loop, tells the agent to read Greptile's review, apply fixes, push again, and repeat until it hits 5/5 or stops after five turns. He holds a firm rule: code ships to production only at a score above four out of five. 7. Why Code Review Fits and Apps Resist Mic draws the line: code review offers a fixed, defined feedback loop, so a loop thrives there. App building stays fuzzy because the full vision lives partly in his head, and the loop even cracks past 1,000 lines of code, where he splits work into multiple PRs to keep Greptile sharp. SEO pages and other binary jobs fit loops too. 8. The Missing Piece and the Future I point out the piece startup builders truly need: sharing the app for real feedback midway, the way you would pull over on a road trip rather than ride full self-driving from Miami to Charleston. We both agree the future holds fully autonomous loops, yet as of this recording on June 9, 2026, human-in-the-loop stays the best loop. The #1 tool to find startup ideas/trends - https://www.ideabrowser.com/ LCA helps Fortune 500s and fast-growing startups build their future - from Warner Music to Fortnite to Dropbox. We turn 'what if' into reality with AI, apps, and next-gen products https://latecheckout.agency/ The Vibe Marketer - Resources for people into vibe marketing/marketing with AI: https://www.thevibemarketer.com/ FIND ME ON SOCIAL X/Twitter: https://twitter.com/gregisenberg Instagram: https://instagram.com/gregisenberg/ LinkedIn: https://www.linkedin.com/in/gisenberg/ FIND MIC ON SOCIAL X/Twitter: https://x.com/Rasmic Youtube: https://www.youtube.com/@rasmic