Aiden JohnstonInstagram

day 88 of posting every day until @curatedating has 3,000 daily active users

1:43English(auto-generated)7 segments469 words, 2 min read

Search inside any video

SavedThat transcribes your saved videos and lets you search across all of them instantly. Save this video and find any moment.

TL;DR

Learn how to set up AI workflows that operate while you sleep, maximizing productivity and minimizing risks in your projects.

AI agent workflowsautomating tasks with AIeffective AI usageproject management with AIsleeping while AI worksAI task delegationsafeguarding AI projectsmaximizing productivity with AI

Chapters

  1. 0:00Introduction to Overnight Agent Workflows
    01
  2. 0:15Establishing Credibility with AI Usage
    02
  3. 0:30Defining Tasks and Scope Clearly
    03
  4. 0:45Giving Sub Agents Their Own Copies
    04
  5. 1:00Baby Proofing Your Project
    05
  6. 1:15Having Agents Check Each Other's Work
    06
  7. 1:30Creating Clean Deliverables
    07
  8. 1:40Conclusion and Next Steps
    08

Transcript

0:00

If you pay for Claude, ChatGPT, or any other AI service, you are wasting your money unless you have agents running while you sleep. This is Overnight Agent Workflows 101, and by the end of this video, you should have an idea of how to have AI complete tasks while you're away doing other things. To establish credibility, here's my AI usage in the last 30 days, and the money that comes out of my pocket is only $200 a month to Anthropic. So, I run a startup and all the things

0:15

associated with the startup, code to review, marketing to do, videos to post, and last night before bed, I gave it the prompt, what can you do overnight that's useful and doesn't need me. And here's how to set up your workflow so it's that easy for you too. Number one, define the task and scope clearly before you start. It's like any other human you talk to. You can't just tell them

0:31

to do something when they don't know what you want them to do. Or in my case last night, it had a list of GitHub issues to work through. Number two, give every sub agent its own copy of the project. If you want things to happen in parallel, have your Claude tell you about Git work trees. Number three, baby proof your project. You've heard all the horror stories of on-call engineers getting

0:47

pinged at 3am because the database got deleted by the agents. Only give it the permissions that it needs and take away all the dangerous buttons. Cloud Code has a feature called hooks that blocks a command before it runs. For example, in my startup, agents can't merge code, can't post to GitHub, and can't push to my live database.

1:04

And trust me, you will sleep a lot easier when that no matter how bad the AI messes up, at least it's reversible. Number four, have other agents check the work. You can have recursive loops of design, build, test, review from different agents going back and forth so that you don't wake up to a lot of code that is complete and utter bullcrap. You will wake up to bullcrap that has been reviewed by at least a couple other agents.

1:20

Number five, make the deliverables either something clean and defined or decisions that you have to make once you wake up in the morning. For example, my morning report was seven questions that I answered in one line each. So if you're paying for AI and it's just sitting there while you sleep, that's just like a gym membership you never use. That was day 88 of posting every day into my dating app as 3000 daily active users.

1:38

Follow along for the journey. Peace.

Are you the creator or rights holder of this video? Request removal of this transcript.

Related Transcripts

Never lose a moment again

Save videos from YouTube, Instagram, and TikTok. Search everything that was said, and jump to the second.

Start Saving Videos - Free Trial