@Cander_zhu: In the past two days, @AnatoliKopadze posted two blockbuster contents, and I read both carefully: 1. His ultra-detailed long article "Loops explained: Claude, GPT, Mira and what actually works" (currently over 8...

Over the past couple of days, @AnatoliKopadze dropped two重磅 pieces of content, and I read them both carefully:

1. His ultra-detailed longform piece, “Loops explained: Claude, GPT, Mira and what actually works” (now over 8.9 million views)
2. Peter Steinberger (creator of OpenClaw) delivering a 19‑minute talk at AI Engineer Europe

Combining @addyosmani’s article “Loop Engineering” with both of Anatoli’s pieces yields a huge amount of insight.

Anatoli’s article lays out the low-level logic of Loops with exceptional clarity:

• A prompt is “you give step-by-step instructions”; a Loop is “you give a goal + acceptance criteria and let the Agent run the result on its own”
• A reliable Loop usually needs these five parts: Automation (trigger), Skill (reusable instructions), Sub-agents (separate the coder from the code reviewer), Connectors (actual execution of actions), Verifier (hard‑core acceptance checking)

And Peter, as the creator of OpenClaw, shares his real-world workflow in the video:

“Each loop is so much faster now that I ship more than ever with way less effort.”

His core takeaways:

• The goal isn’t writing the perfect prompt — it’s designing a closed‑loop that can verify itself
• The Judge (acceptance criteria) determines the Loop’s success or failure; best to use deterministic checks like tests, type checks, diffs, etc.
• Structure your project for the Agent, not against it

I really hope everyone reads these three pieces — ideally together (Addy’s Loop Engineering + Anatoli’s long article + Peter’s video). Theory + practice + framework complement each other, making the overall effect much stronger.

I’ve personally been building a small Loop around “structuring new notes in Obsidian” — still early days, but I can already feel the shift from manual organization to automated workflows.

For anyone interested in Loop Engineering, I strongly recommend reading all three. Theory + real‑world examples + systemic thinking — that combination works best.

Are you leaning toward learning the theory framework first, or would you rather dive straight into practical workflows?

#LoopEngineering

Anatoli Kopadze (@AnatoliKopadze): Peter Steinberger, the guy who built OpenClaw just shared his actual workflow:

“Each loop is so much faster now that I ship more than ever with way less effort.”

19 minutes from the person who knows more about AI agents and loops than anyone else.

Watch it, then read the full