Boris Cherny, the engineer who built Claude Code: "I don't talk to an agent anymore. I talk to a loop" a prompt is one instruction you babysit. a loop is a goal the AI works toward on its own: > it plans > does the work > checks itself... > fixes what's weak > and repeats until it's done you step out, the work keeps going that's the shift the best engineers quietly moved to. most people hear a line like that and have no idea what it means in practice this breaks it down without the hype: what a loop actually is, when it's worth building, when it's a money trap, and how to run one yourself today watch his clip, then read the article belowshow more

Most people are still thinking in prompts

once you see it, it's hard to go back

Thanks for sharing. I'm using Claude every day as well.

Good

that's where the AI agents are now. The only constraint to me is that it takes some time to train it, and you need to do that, otherwise the outputs will be bad But it's definitely worth it, and you need some degree of CS knowledge

24/7/365 loop. In work mode ai is disciplined worker, in autonomy mode it's way more free what it does, moods can impact his decisions. Meet Genesis. AI strapped to human like brain.

This feels like a shift from controlling AI to trusting a system.

Prompts you babysit loops you just set and forget

When tokens are free. Loops/goals are ecstasy.

Agents get better when they self-correct

already reading this article from Anatoli 10/10 to be honest

Bro that's basically AI working while you sleep crazy

@AnatoliKopadze's Loops explainer is great. Glad to see the conversation head towards getting real value out of agents. A thread:

Is this not just the agent loop with human gates? Plan, act, verify, repeat. That pattern has a name already. I think we keep rebranding the same thing and calling it a shift! or maybe I missed the point?

the "loop vs prompt" framing is right but the hard part nobody mentions is the "checks itself" step — self-grading is where loops quietly drift

if you value your time, you're already switching to loops

Dude, that loop mindset is next level. AI handling the full cycle solo? No more constant prompting, just set the goal and let it cook.

Auto loops cut tons of manual oversight hours, yet complex setups carry steep hidden maintenance costs.

i like this style keep going

loop mode is actually insane

bro why didn't this get 1 million views?

The key word is “checks itself.” Without that, the loop is just a long prompt with better branding.

Loops is directionally right, but without independent reviewers it will fall apart because agents will optimize for completion and not correctness.

That's a cool evolution! It's amazing how AI is becoming more self-sufficient.

yeah most people mean 'it tried again 6 times' and call that a loop

because the hard part was never just executing a prompt, but knowing what self-improvement looks like in an unpredictable system. boris is right to call it a loop, not just a task list