Everyone is talking about agent loops and loop engineering right now, but most of the advice assumes you are a hardcore coder running fleets of agents around the clock. In this video I break down what an agent loop actually is (reason, act, observe, repeat), why the verification step matters more than the architecture, and how to think about a "done" criteria that your agent can actually check. I walk through a few real loops I ran, including thumbnail scoring, a three.js plane, and an Abbey Road recreation, and explain why loops are about getting you closer on the first try, not perfect output. If you have been feeling behind because you are not running five agents at once, this one is for you.

I switched Claude Code over to GLM 5.2 and ran it all day. It's a 756 billion parameter open source model you can route straight into the Claude Code harness for about five times cheaper than Opus, and for most of my knowledge work it held up fine. In this one I show you what it can build, where it beats Opus and where it doesn't, and exactly how to set it up so you can switch between models per project. Here's the config I use. Drop this into your .claude/settings.local.json and swap in your own Z.ai key: "env": { "ANTHROPIC_BASE_URL": "https://api.z.ai/api/anthropic", "ANTHROPIC_AUTH_TOKEN": "your-z-ai-api-key-here", "ANTHROPIC_API_KEY": "", "API_TIMEOUT_MS": "3000000", "ANTHROPIC_DEFAULT_OPUS_MODEL": "glm-5.2", "ANTHROPIC_DEFAULT_SONNET_MODEL": "glm-5.2", "ANTHROPIC_DEFAULT_HAIKU_MODEL": "glm-5.2", "ANTHROPIC_SMALL_FAST_MODEL": "glm-5.2", "CLAUDE_CODE_SUBAGENT_MODEL": "glm-5.2" }

Some of you have never heard of the AIS+ community. Others have but the part that trips you up is the actual difference between the two. Either way, this post will give you clarity. This free group is a bundle of quick resources pulled from my YouTube videos, plus a massive open community that anyone can join. It's a great place to get your bearings and see what's possible. But it's open to everyone, it can be noisy and overwhelming, and there's no path through it. You can get help from other members, but I rarely answer questions here. AIS+ is the opposite: - A step by step roadmap with a clear order, so you're never guessing what to do next - A much smaller community of people who are seriously committed to building and selling AI agents - I answer questions every day and run a weekly Q&A call where you can get direct access to me For the course material: The roadmap takes you from zero to building and selling AI agents, and the whole thing is built on the latest tech like Claude Code and Codex. We update it constantly. The old n8n material has been archived. It's still there if you want it, but it's no longer the focus, because the way you build today has moved on and the courses moved with it. Here's the actual roadmap inside, in order, with when each piece opens up: 1. Start Here (opens the moment you join). Gets you oriented. How the community works, the path ahead, and how to get help when you need it. 2. Build Your Portfolio (opens the moment you join). Why a portfolio matters, beginner level tutorials, and what types of projects to focus on. You end up with real work you can show a client. 3. Claude Code (opens the moment you join). This is now its own dedicated course. Build faster, turn ideas into working automations, and go deep on the tool serious builders are using right now. This takes you from beginner to advanced, step-by-step. 4. Get Your First Clients (opens after 30 days). Getting your first clients is hard, because you don’t have any case studies yet. So, we analyzed all of the success stories from our members and found they get their initial clients with two different techniques: warm outreach and Upwork. So, we teach both techniques in detail with exactly what to say, exactly how to position yourself when you have no proof.