Jeremiah Lowin runs 5 Claudes at once. Here's the system behind it.

Hinzugefügt: 2026-05-30

[00:00:00]How do you leverage AI tools to increase your productivity to the to the scale that you have? I mean, right now, you know, people might get on Claude and they'll and they'll type a question or they'll say, "Fix my message." or something simple, but you've really taken it to a completely different degree.

[00:00:14]>> I think that my use of [music] it is aggressively parallelized and I put in the effort up front to be able to use a shorthand with it that gives a predictable result. So, for example, while I'm reviewing things, which is a very like [music] come and interact with it for a moment and then let it go off and think, I can take my attention and then go work on another thing. And so, it's like it's one of these things [music] where it's all well and good to parallelize things, but because it's a human in the loop process and you have to come back into the loop, I think my productivity exploded when I did that because I am now [music] running at every moment, right now as we're talking, there are five tabs open in my terminal as five different Clauds work in five copies of the same repo. They can't step on each other. They're all working across multiple branches.

[00:00:54]Eventually, I'll use PRs to sort out their work and make sure they don't step on each other.

[00:00:59]And so, that's given me this ability to work on so many different things at once.

[00:01:04]>> In terms of context switching, do you feel like the best CEOs and founders have this skill?

[00:01:09]>> I think you need some degree of it just for the practical reality of you and I will talk about productivity right now.

[00:01:15]Then we'll go back in the room and we'll have a talk about pricing and then go to market and someone will have a question about an open source thing and by the way, we have Prefect and by the way, we have FSMCP and and we can't I can't take a moment and be like, "Sorry.

[00:01:27]I can't talk about this right now because I need to remember how this happens and how this works."

[00:01:33]>> What advice would you have for someone that is trying to leverage this new world of AI agents and they're seeing you, they're like, "Oh man, this guy has built all these projects on his own. He has a family."

[00:01:46]>> If there's one thing that I've used to do more work and keep a level of quality, it's trying to [music] figure out like, "Oh, I can keep track of a a of things at once." which means I can dispatch, I can scale myself in those directions at once.

[00:02:01]But I need systems for like keeping track of that work and being on top of it. I think the secret sauce is [music] that there are many failures for every success. And I think that folks don't look at the hundreds and millions of lines of code that I've written [music] that no one's using today, that I'm not using today, that I've open-sourced and been like, this [music] might be useful and and nothing's happened to them. Um I don't think anyone is is good enough or knows the situation well enough or understands a problem well enough to to like one-shot a solution into the world.

[00:02:34]I think for folks who want to build software, the real question is to focus on what's hard. What big problems are out there that one-shot software is unlikely to solve, that we might need a framework for, or we might need collaboration for. And then deciding like, what is success? And so that kind of curiosity, and how do I just get to the next challenge?

[00:02:54]I don't know. I value that.