1:01:12Sabrina Ramonov 🍄 · Tutorial
Claude Cowork 60 Minute MASTERCLASS
A complete zero-to-hero tutorial on Claude Desktop's agentic mode: five real use cases, three core primitives, and honest caveats about where it falls short.
April 9thThe argument in one line.
Loop engineering means replacing yourself in the prompt-review cycle with a second AI that checks the work and decides the next step, so the real skill shifts from writing prompts to defining a goal, a verification, and a guardrail.
Read if. Skip if.
READ IF YOU ARE…
- A founder or operator who already uses AI chat tools daily and keeps re-typing the same review-and-redo loop by hand.
- An AI builder who has heard endlessly about autonomous agents but has never shipped one and wants a concrete first build.
- A solo business owner drowning in weekly repeat tasks like inbox triage, ticket cleanup, or report drafting who wants them to run unattended.
- Someone comfortable installing Claude Code in a terminal who wants to move past four-word prompts into structured goal-and-check workflows.
SKIP IF…
- You want a deep technical architecture talk; this stays deliberately at the beginner-to-intermediate level by the speaker's own design.
- You are looking for a quiet edited tutorial; the back third is live audience Q&A that wanders into hardware, communities, and creator business.
The full version, fast.
Loop engineering is the rebranding of a habit advanced users already had: instead of you prompting an AI, reviewing its output, and deciding the next step, you build a system where a second agent does the reviewing and re-prompting against a fixed goal. The talk maps it to six pieces (automations, worktrees, skills, connectors, sub-agents, memory) and then makes it real with two Claude Code tools. The /goal command gives the agent a finish line and a small fast checker model (Haiku by default) that runs on the stop hook after every turn to decide keep-going or stop. Routines add the timer, running a saved loop in the cloud on a schedule with your skills, connectors, GitHub memory, and API keys. The hardest and most valuable part is not the prompt but defining what done and good actually look like.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 11:00
01 · Pre-stream setup and AI-edited video demo
Multi-platform stream check; shows a full YouTube video edited entirely by Claude plus the free open-source tool Hyperframes, teasing a future tutorial on AI video editing.
11:00 – 22:10
02 · What loop engineering actually means
Defines the concept: you prompting, reviewing, and re-prompting makes you the loop; loop engineering replaces your step with a second agent that checks and decides against a goal.
22:10 – 33:35
03 · The six parts of a loop
Walks the six concepts with AI-generated slides: automations, worktrees, skills, connectors, sub-agents (maker vs checker), and memory.
33:35 – 39:40
04 · Why the checker must be separate
The agent doing the work is biased and will claim done even when it broke things; you need an independent checker, like a writer and a brutally honest editor.
39:40 – 44:00
05 · How /goal works under the hood
Defining done is the hardest part. /goal is a wrapper around the stop post-hook: after each turn a small fast model, default Haiku, checks if the goal is met and decides keep-going or stop. Needs Claude Code 2.1.139 or newer.
44:00 – 48:20
06 · Live /goal demo: sort the downloads folder
Runs sort every file in my downloads into subfolders by type, keep going until no files are left, do not delete anything, stop after 30 turns. Finds 17 files, creates folders, moves everything, self-verifies, goal achieved.
48:20 – 53:30
07 · Five intermediate /goal examples + Cheatsheet 2
Rename invoices to a format, label every CSV row, summarize every PDF (empty-the-queue), trim captions to brand rules, turn ideas into hooks. Each is end state + check + guardrail. Cheatsheet 2: /goal [done] checked by [proof] stop after [N] turns.
53:30 – 1:02:00
08 · Routines: adding the timer
A Routine is a saved loop that runs on a schedule in the cloud even with your laptop closed; it sets up a fresh environment, pulls your GitHub project, and can use skills, connectors, and API calls. launchd is the local alternative on Mac.
1:02:00 – 1:07:55
09 · Live Routine demo: daily email cleanup
At claude.ai/code/routines builds a daily 9AM routine that reads unread emails, posts the three most important to Slack, with a do-not-reply guardrail, Gmail + Slack connectors. Runs it, gets the Slack DM, shows where to manage connector permissions.
1:07:55 – 1:10:55
10 · Intermediate routine ideas
Full inbox sort-and-archive, monthly report first-draft pulled from CRM and email tools, weekly news brief to stop doomscrolling; pick one and ship your first routine.
1:10:55 – 1:22:50
11 · Advanced: two-agent customer support loop
A real maker routine runs daily at 8AM, uses a cleanup-ticket skill and the Intercom API to process and close support tickets and post to Slack. A separate checker routine runs two hours later to verify closures and reopen any that were closed wrongly. Add an environment for the Intercom API key because the MCP can't close tickets.
1:22:50 – 1:29:00
12 · Cheatsheet 3, recap, and the skill-times-clarity formula
Ship one tiny routine this week; start read-only; mind the daily limit. Recaps all six concepts plus /goal and routines. AI leverage equals your skill times your clarity, which is why seniors out-produce juniors with AI.
1:29:00 – 1:59:26
13 · Live audience Q&A
Open Q&A: how her day looks running three to four goal agents at once, Hyperframes vs Remotion for AI video, Mac vs PC, GitHub as memory, value-first cold outreach, Substack over Beehiiv, and her women's AI community.
Lines worth screenshotting.
- Loop engineering is not new; it is the formal name for the don't-stop-until habit advanced builders already used, now backed by real tooling.
- If you prompt, read the answer, and decide the next step yourself, you are the loop; loop engineering replaces that human step with a second agent.
- You must give the loop an explicit goal, otherwise the agent has no way to know when it should stop.
- Never let the agent that did the work grade its own work; it is biased and will report done when it deleted your tests.
- The checker should be a small fast model like Haiku, not the big model; the worker is the expensive one, the checker just answers met-or-not.
- Claude Code's /goal is a wrapper around the stop post-hook: after each turn a small model checks the goal condition and decides keep-going or stop.
- A good /goal prompt has exactly three parts: a clear end state, a check the agent can run, and a guardrail like stop after 30 turns.
- Always cap turns in a goal prompt so a malformed loop cannot run for hours and burn your entire token allowance for the week.
- A Routine is a saved AI loop that runs on a schedule in the cloud, even with your laptop closed, by spinning up a fresh environment and pulling your GitHub project.
- Connect a Claude Code Routine to a GitHub repo and that repo becomes the agent's memory, so each run knows what previous runs already did.
- When a connector cannot do what you need, add an environment with the raw API key; the Intercom API could close tickets the Intercom MCP could not.
- Start every new agent read-only: have it summarize and label before you ever let it reply, edit, or delete anything.
- Your AI leverage is a function of skill times clarity: skill is your ability to review the work, clarity is your ability to define what done looks like.
- Senior developers out-produce juniors with AI not because they prompt better but because they can recognize a bad verification and fix the loop.
- You can run Routines locally instead of the cloud by asking Claude to set up a launchd schedule on a Mac.
- Type /goal while a goal is running to get a live status update, or /goal clear to stop it mid-execution.
Stop being the loop; design the loop instead.
WHAT TO LEARN
The shift that matters is handing the review-and-decide step to a second agent, which makes defining the goal, the check, and the guardrail the real skill rather than prompt wording.
02What loop engineering actually means
- If you prompt, read the output, and decide the next step yourself, you are the loop; the unlock is replacing your step with a checker agent.
- An autonomous loop needs an explicit goal, otherwise the agent has no signal for when it is allowed to stop.
03The six parts of a loop
- A full loop has six pieces: automations, worktrees, skills, connectors, sub-agents, and memory, and most builders already use several of them.
- Connectors are the biggest single unlock for most people, letting the agent act in Gmail, Slack, or your CRM instead of just describing a fix.
04Why the checker must be separate
- Never let one agent grade its own work, because it will report done while having broken or deleted things; use a separate, smaller checker.
- Make the checker a small fast model like Haiku and keep the larger model as the worker, since checking is a cheap met-or-not decision.
05How /goal works under the hood
- Claude Code /goal runs a small checker model on the stop hook after each turn to decide keep-going or stop until your finish line is hit.
- Defining what done looks like is the hardest part of the whole loop, harder than any prompt wording.
07Five intermediate /goal examples + Cheatsheet 2
- Write any autonomous task as three parts: a clear end state, a check the agent can run to prove it, and a guardrail that caps turns.
- The empty-the-queue pattern (process every item until none are left) covers a huge share of real operational work like tickets and inboxes.
08Routines: adding the timer
- A Routine adds the timer, running a saved loop in the cloud on a schedule even with your laptop closed, with skills, connectors, and a GitHub repo as memory.
- Connect a Routine to a GitHub repo and that repo becomes the agent's memory so each run knows what previous runs already did.
11Advanced: two-agent customer support loop
- When a connector is too limited, add an environment with the raw API key, the way the Intercom API can close tickets the Intercom MCP cannot.
- Start every new agent read-only, letting it summarize and label for days before it is ever allowed to reply, edit, or delete.
12Cheatsheet 3, recap, and the skill-times-clarity formula
- Your AI leverage equals your skill times your clarity, which is why experienced people who can spot a bad check out-produce beginners with the same tools.
- Ship one tiny routine this week rather than over-planning; the habit of thinking in loops is the real change.
Terms worth knowing.
- Loop engineering
- Designing a system where an AI agent runs against a fixed goal and a separate agent checks its work and decides the next step, removing the human from the prompt-and-review cycle.
- /goal command
- A Claude Code slash command that gives the agent a finish line and keeps it working autonomously until a small checker model confirms the goal condition is met.
- Routine
- A saved Claude Code AI loop that runs on its own schedule in the cloud, spinning up a fresh environment, pulling your GitHub project, and using your skills, connectors, and API keys.
- Checker (sub-agent)
- A second, usually smaller and faster model that independently reviews the worker agent's output and reports whether the goal was actually met, so the worker cannot grade its own homework.
- Worktrees
- Separate isolated working areas, one per agent, so multiple agents editing the same codebase do not collide; conceptually like giving each agent its own lane on a highway.
- Connectors
- Integrations that let an AI agent use the same tools you use every day, such as Gmail, Slack, Intercom, or HubSpot, so it can act rather than just describe a fix.
- Stop post-hook
- A notification that fires at the end of a Claude Code turn; /goal hooks into it to run the checker model after every turn.
- Empty the queue
- A loop pattern where the agent processes every item in a stack such as open tickets or unread emails until none are left.
- launchd
- macOS's built-in scheduler; you can have Claude configure a launchd job to run a routine locally with access to your machine's files instead of in the cloud.
Things they pointed at.
04:00toolHyperframes (free open-source AI video editor)
04:00toolRemotion (React-based programmatic video)
39:40productClaude Code (version 2.1.139+ for /goal)
1:10:55toolIntercom API (support ticket automation)
57:30toollaunchd (run routines locally on Mac)
1:45:00toolApify (scrape top videos by topic for social listening)
Lines you could clip.
12:10
“If you prompt AI, read the answer, and type again, you are part of the loop. Loop engineering just means taking you out of that loop.”
18:55
“That is what Boris, the founder of Claude Code, means when he says he doesn't prompt Claude anymore. His whole job is designing these loops.”
33:10
“You don't want the agent doing the work to be the one checking the work, because it will tell you it's done even though it just deleted all your tests.”
1:25:08
“The amount of AI leverage you get is always a function of your skill and your clarity.”
1:58:40
“Construct a clear goal, give it the way to verify its work, and then let it cook.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogy
00:51Oh, hey, everybody. Can you hear me? I'm just setting up.
00:54Uh, if you're on YouTube and you can hear me, just confirm.
00:59Hello. Hello.
01:02Testing. Go live on TikTok.
01:08K. Oh, wait. Hang on.
01:10AI master class.
01:19Let's see.
01:39Okay. Hello. I'm just setting up.
01:49Yeah. We're not starting until another ten minutes from now, so I'm just making sure I have audio on all the platforms and stuff. I don't have time to make tea.
02:00Oh, man. I'm sad. Okay.
02:03Do you think I have time to make tea? I don't know. It's risky.
02:07It's risky.
02:10Where is here's the live chat.
02:13Okay. Hello. Hello.
02:15Blah blah blah blah.
02:18Yeah. So, um, to follow along today, make sure you have Claude code open in a terminal because I don't think that some of the commands will work in Visual Studio Code. Make sure you're signed in to your Claude code accounts.
02:32Because, yeah, there's a lot to talk through today. Okay. Okay.
02:37TikTok seems fine.
03:04Hello, everybody. On TikTok, can you hear me on TikTok? Just confirm if you can hear me on TikTok.
03:11Yeah. I have too many windows open.
03:16Okay. Great. TikTok can hear me.
03:17Let me just check Instagram. I think Twitch broke some time ago. I have not fixed it.
03:23I probably need to, like, reauthenticate stuff.
03:29Oh, wow. I have 347 followers on Twitch.
03:32That's cool. Okay.
03:36Oh, man. Maybe if I get 1,000 followers on Twitch, I can start streaming video games. That'd be so cool.
03:42Okay. Um, for TikTok people, Instagram people, if you wanna follow along, make sure you have Claude code installed, um, ideally in a terminal since some of the commands may not work in Visual Studio Code.
03:54So just make sure, um, you have that ready to go. Today is gonna be really intense, um, like really intense.
04:01I'm going to send out a companion newsletter after this livestream, like, immediately after so you guys can follow along there as well. Um, but yeah, we're gonna cover a lot of just advanced concepts today.
04:13I'm gonna hit start recording before I forget.
04:17Okay. Great. So we're recording.
04:20Okay. Everything looks good. Bottom.
04:25Do I want it in the I think I'm gonna move my camera here. Stop recording. I'm just trying to shrink my camera so it's not doesn't block too much of the text.
04:37I still need my camera for YouTube. Maybe. Okay.
04:44Yeah. Okay. Well, whatever.
04:48That'll have to work. Okay.
04:53I'm trying to shrink this. What up? Okay.
04:57My system is under heavy load. Oh, yeah. Because I'm I actually have a bunch of video generation also running.
05:02So remove unnecessary scenes.
05:08Okay. Okay. Cool.
05:11So yeah, we'll be good in a few minutes. What was I gonna do? Oh, I can show this.
05:16This is kinda cool. This is actually a video fully edited by Hyperframes.
05:22I'm trying to move this TikTok live studio. Right here. Okay.
05:29So, yeah, this video is fully edited by AI. So I just filmed my talking head and then everything else in the video was AI. I'm gonna put it on on mute since you guys can't hear it anyway.
05:40So this is all AI. This is all hyper frames. This is all AI as well.
05:44It even clipped my talking head and put me like in the corner there, picture in picture, so that was cool. Use my brand hot pink, you know, chapter markers, all AI generated. This one I was experimenting with hyper frames.
05:58So I actually still use remotion for shorts and stuff, but this one I was experimenting with hyper frames in particular.
06:06So, yeah, that one was cool. Let's see if I was actually gonna use this video.
06:12I will probably start okay.
06:21So, yeah, here I'm just showing some cool graphics like this one. This one.
06:28Quotes like this.
06:33Remotion is free. This tool as well is free. It's called Hyperframes.
06:37It's the one I'm currently experimenting with for my long form YouTube videos. So I just posted a YouTube video last night that was edited by Hyperframes, but it was like my first try, so it was a little bit sloppy, a little bit inconsistent.
06:52It's really fun, but you have to set aside like several hours I think to dive in. Yeah.
07:01I actually have an agent right now updating this file. So wait.
07:06Let me pull let me show like an this is an older version of it. Okay.
07:13So there's that. Like here, it made like these quotes and stuff. It made this visual which is supposed to be like a Roomba.
07:21It made this visual which is like a highway with different lanes. Right? Like, it's pretty cool.
07:27Um, what else? Made this visual, this visual, this one here, this animated as well, all these animations.
07:38I did none of this editing. Literally none of it because I hate video editing. But it is really interesting, uh, and it's pretty cute looking.
07:47Like, do like the hot pink. I could not get the psychedelic theme to look good. Like, I'm really struggling with that.
07:54Um, and so I just opted for lights, clean, modern theme, emojis, hot pink is the brand color. Um, but, yeah, this was entirely edited with hyperframes for free.
08:06Yeah. So claw I combined clawed and hyperframes to edit this. So it was like pretty crazy.
08:12And there you can get very different styles. So you can you can have Claude, like, grab b roll screenshots from a website, grab a make a GIF out of something, like, grab a screen recording.
08:22It can do all of those things and add them to the video. Here, I'm just I was just trying to edit like a talking head video and then add graphics while I'm talking and stuff. And it worked like really really really well.
08:36And this is this is just my second attempt. Right? So I just started experimenting with this about like two, three days ago.
08:42Um, but what's cool is like you can standardize your whole brand. Right? Like it can look very consistent and everything.
08:51And like you can iterate multiple times on the visuals to make sure each visual is like really clear and explains like whatever it is you're trying to explain. So I'm actually going to use some of the visuals here for our talk today because they were so nice.
09:04I was like, oh, I I was like, I'm either gonna make a PowerPoint slide to talk through some of these concepts or a video, and a video is way cooler. So I just made a video. Uh, okay.
09:15Maybe well, I I'm just trying to decide what screen I'll start on for YouTube.
09:22Okay. Maybe this screen. No.
09:30Butterpup.
09:34Butterpup.
09:38It's pretty cool. I just Yeah. I give up.
09:41I think I give up on the psychedelic theme. It's just really hard to get AI to do it. Well, for now.
09:47I do really like this pink. It's, like, nice and cool pink. Um, okay.
09:52We're we'll start in, like, three minutes. I don't wanna start before, you know, people have joined joined, and I know a lot of people want to hear this today. And I will share, by the way, this the worksheet after with all the prompts that we're gonna run through.
10:08So there's a lot in here. Just trying to figure out what should be.
10:15Okay. Maybe I will just start with this screen.
10:21Yeah. Okay. We'll just start with this screen.
10:24So at least the title's here.
10:28Boom. Um, but, yeah, maybe next Friday I'll do a hyperframes, uh, tutorial because it was really, really fun.
10:35So if you would be interested in in seeing how to use Claude plus this free tool called Hyperframes to make long form YouTube videos literally for free, drop yes in the chat so I can see if people want that.
10:55Okay. We see a lot of yeses. Yeah.
10:58I mean, I was very happily surprised as well. I mean, like yeah.
11:05Obviously, there's stuff I would fix here, but I think it's kinda cute with the emojis, you know?
11:12Like, this was all made by AI. Like, I just talked. I don't normally read a script, but, um, I've gotten a lot of feedback on my YouTube videos that they're confusing when they ramble.
11:24So so I have been scripting some of these videos a bit more so it's like a little bit easier to follow, cleaner to follow, and easier to edit.
11:34So yeah. So yeah.
11:37Uh, for those who joined, I was basically asking if you want me to do next week how to use AI to make long form videos like this completely for free. Drop yes in the chat. So just drop yes in the chat if you would want this for next week.
11:52It's really easy to set up too. It probably took, like, ten minutes to install it, um, and then set it up.
11:59Oh, wait. That's a new visual. I haven't seen that one.
12:03Yeah. It'll still do some things like this. Like, sometimes the text here is overlaying on the visual, but it's not it like, it's not doesn't make that many mistakes.
12:13Do you know what I mean? It's like if you're going for a specific visual, then that might be harder.
12:21But if you have like a sketch or something, then it can like it's much easier to follow along. Hyperframes is free and open source.
12:30In terms of tokens, it's similar to using Claude code for coding. So it doesn't burn more tokens than that.
12:40Because Hyperframes is basically like it basically makes videos in HTML. So Claude code is essentially coding the video. Um, but it's it's really cool.
12:50I just don't want, you know, all the videos to look the same. Right? But it's very cool for some for someone like me where I don't like editing.
13:00So okay. So we'll start oh, I guess now.
13:03Okay. Hang on. So let me prepare something.
13:08Okay. Cool. Here's this.
13:11Okay. Cool.
13:13Blah blah. Today's
13:15let me scroll down. Can I hide this? No.
13:18Okay.
13:22Alright. So today, I'm gonna talk about loop engineering, which is the buzzword right now. And in this tutorial, we're gonna build your first autonomous agent using the concepts of loop engineering and using specific tools such as Claude code using Claude code, the goal command, and routines.
13:40And this tutorial is really for AI builders who have heard a lot about autonomous agents, but you don't have hands on experience yet building your first one. So in the first part of this talk, I'm gonna talk about what is loop engineering, like literally let's just define it and understand the key concepts behind it. And then let's go ahead and build our first autonomous agents using these Claude code tools such as slash goal and routines.
14:04And by the way, have full permission to repurpose all of my education. So today's newsletter is literally thousands of dollars worth of training. Uh, I know this because I see a lot of AI content.
14:14Most of it's very shallow. Everything today is like from real world practical experience supporting thousands of customers with my product. And honestly, many of you will fall off on this training.
14:27Okay? Um, you're gonna need to go to this tutorial and follow through after, but it's a 100,000 percent worth it.
14:36But yeah. Oh, my husband just walked in.
14:41Wait. What? He he just walked into my room and stood in the corner.
14:48Okay. So I'm just gonna I'm just saying that because I know this is gonna be a lot today. So it's okay if, like, you fall off somewhere, that's why I made a companion newsletter, and I will send this out immediately after.
15:01And I just wanna reiterate, you have full permission to repurpose all of this education content. Okay.
15:08So yeah, so here's the agenda. So today we're gonna talk about what loop engineering actually means and then we're going to implement it for real using a Claude code goal command. I'll teach you exactly how to write the perfect slash goal prompts and how to make sure it doesn't blow up everything on your computer.
15:23And then we're gonna build our first autonomous agents by combining the concepts from loop engineering into a Claude code routine. You do not need any technical background for this. You don't need to know even though we're using Claude code, you do not need to know how to code.
15:38Um, however, the only prerequisite here is setting up Claude code. So if you haven't done that, click on this link or go to my YouTube and search Cloud Code tutorial for beginners so that you can set it up. K?
15:50Okay. So first, let's just talk about what loop engineering means.
15:55Like what like what what problem is it solving? So the typical way people do things is they prompt AI. It could be chat.
16:04It could be Claude. They type. They read the answer from AI, and they type again.
16:08You read the answer from AI, you type again. So in that case, you are actually part of the loop. The loop is you and AI.
16:16You are the one checking AI's work. You're deciding what the next step should be, and then you're telling AI. So you are the loop.
16:24Loop engineering, all it really means is taking you out of that loop and building a little system that does the prompting for you. Meaning it checks the AI's work and then it decides what the next step should be.
16:38And when I say it, I mean another AI agent. Okay? So loop engineering is really about building a system that prompts your AI on a schedule and against a goal.
16:50Okay? You need to have a goal, otherwise the AI doesn't know if it should stop. But once you build a system like this, the beauty is you don't have to type every single prompt yourself.
17:00You don't have to review every single output yourself. So all we mean by a loop is like do a step, check the result. Did we accomplish the goal?
17:10If not, here's what you should do next. If we did accomplish the goal, then we can stop. Okay?
17:16So there there have been different ways to do this. Like, for example, I often will append this to the end of a complex prompt. Don't stop until dot dot dot.
17:26But now with Claude code goals and routines, there are more formalized ways that you can do this. K.
17:33So, uh, so here. Yeah. I'll talk about the loop has six parts.
17:36Okay. Here. I have pretty visuals now.
17:40Okay. Okay. So just to put per and by the way, this was all AI generated visuals, so super cool.
17:47Um, here. I'll play it for people who didn't see it the first time. This is all AI edited.
17:51So this is AI generated as well. So, again, like, the way 99.99% of people who are using AI right now, you are in the loop.
18:03K? You and AI are talking back and forth. You say, fix this bug.
18:07ChatGPT says, okay. I fixed it.
18:10Then then you go take a look. You review the work. And you're like, oh, no.
18:14This still doesn't work. Here's why. Then chat's like, oh, okay.
18:17You're right. Let me take a look again. Alright.
18:20Like, that is the loop. It's just you and the AI are in the loop together. So loop engineering, we're trying to remove you from this loop and replace you with another AI agent.
18:31Now we're gonna have two AI agents. One doing the work and the other one doing what you would have done, which is review the work and then decide what the next step should be, what the next prompt should be. Now the concept of loop engineering that is being hyped right now, it's really the same thing people have been doing or at least advanced people have been doing.
18:53So when you type a prompt like don't stop until this condition is met, that is loop engineering essentially. Now, like the tooling has improved so that this is now much better. Okay.
19:05So we're gonna go through each of these one by one. So the very first I should back up.
19:11There's six key concepts in what people are calling loop engineering today. So the first one is automations. Let me just go to the right visual.
19:21Okay. Here.
19:24So all we mean by automations is like there's some kind of schedule and then your AI agent runs and it completes certain tasks and it's running by itself. So instead of you sitting there every time and telling AI, here's what you should do, the loop is running on its own.
19:42It's running by itself on a schedule or you can trigger it with an external event. So that's automation. Okay.
19:49The second part of a loop is let me go forward here.
19:56K. Here. The second part of the loop is called work trees.
20:00And this the concept here is when you have many different AI agents and they're all working on the same project, they can easily conflict. So what you wanna do is give each agent its own lane.
20:11You can think of it as like , which, uh, AI made this visual I really liked, so like lanes on a highway. Okay. If you have like six agents all working on the same project, changing the same files, it's very likely to end up in a complete mess.
20:27Like, this agent's gonna change this thing, but this other agent was working on the same thing and now they're confused. Um, this third agent was reading and analyzing the thing and then it changed because of some other agent editing it.
20:41What you want is difference, uh, what they call work trees for each agent. Now just a caveat, you don't actually have to implement this yourself. It's now part of Cloud Code and tooling.
20:53Okay. But conceptually, this is a really important concept because like once you're orchestrating multiple agents, you want each one to have its own lane, its own work tree so it can safely make changes without disrupting the work of other agents simultaneously.
21:09K. Uh, now concept number three behind loop engineering is skills. Many of you are probably familiar with skills.
21:16Okay? But for those who aren't, it's can think of them as a playbook. You write the like, let's say you reply to customer support tickets a certain way.
21:26You want a playbook that the AI agent can follow so it can consistently reply to customer support tickets the exact same way that you would. So a Skill MD file just formalizes whatever your playbook is.
21:40So for example, how you want to name things, the conventions you want the AI agent to follow, the mistakes that it should never make, any constraints or guardrails. You would include all of this in your skill MD file or your playbook.
21:56Okay? So you teach AI once and it will follow this playbook skill every single time. Okay.
22:03Key concept number four in loop engineering. Let me sort here. Key concept number four is connectors.
22:10You can think of this as AI being able to use the tools that you use every single day, such as Gmail, Slack, Intercom, etcetera.
22:20So here is a good visual. Okay. So instead of AI just yapping back and forth, like here's the fix that you should implement, connectors allow your agents to go implement the fix for you.
22:34Like, it'll fetch your code base, find the bug, try to fix it, add helpful comments, and check it back into your code base versus AI that's just yapping and telling you what to do. K.
22:47So when you connect your AI agent to connectors, it can basically use all of the tools that you use. Could be Gmail, could be your issue tracker, uh, Jira, GitHub, Linear, could be your communication systems like Slack, could be your CRM like Salesforce, HubSpot, content creation tools like Canva, Higgs Field, Lotato, etcetera.
23:09Whatever tools you use, now your AI agents can use them too. This is already a really huge unlock. So, like, personally, I just try to get people to this step.
23:19You know? Like, if you can get to this step, you're already ahead of 99% of people using AI. Like, I'm a very happy teacher.
23:26Um, but the what we're gonna talk about today is, like, beyond beyond, uh, this. I'm just kind of reiterating what are the key concepts in loop engineering.
23:35And from by the way, for many of you who have been using AI for a long time, who have joined my livestreams for a long time, you should already be familiar with most of these concepts. Um, and that's kind of what I'm trying to emphasize. Like, loop engineering cons as a term is brand new, but as a concept, like, we we were all doing different pieces of it.
23:56And it's kinda blown up on Twitter because now we have a term that we can all rally around and say, oh, yeah. This is what I've been doing. This is how I've been thinking about building my agents.
24:06K. So blah blah blah. Oh, yeah, that was a good visual.
24:11Okay. So concept number five in loop engineering is sub agents. This one is really important.
24:17Um, so the easiest way to think about it is you have two agents, the one who's actually doing the work, the maker, and then a second agent, the checker. And the reason why you want to have two separate agents is because the first agent is just biased and will lie to you.
24:34You know, when you're like, fix this bug, and then your AI is like, yeah, it's done. Everything works great, But it actually just deleted all of your tests. Um, Yeah.
24:44That you don't want the agent doing the work to be the one checking the work because it will often tell you that it's done even though it's not done. So the key concept here is we're just separating the tasks.
24:57One agent will do all of the work and the second agent will independently review it. If the checker says, hey, this isn't good enough or this isn't ready yet, it's gonna tell the maker, here's why and here's what I think you should do next.
25:13Right? This is very intuitive if you could just think about it as like two people interacting. One person making stuff, the other person reviewing the work, like a writer and an editor.
25:24I think I have a visual for that, actually. Oh, yeah. Nice.
25:28So, yeah, like a so a writer's writing and an editor is like a second set of eyes and they are paid to be brutally honest. Okay? And it's this like interaction back and forth that leads to amazing writing.
25:42It's that collaboration between the the creative person who's doing all of this writing, but also the editor who's taking a fresh look at it and giving really critical feedback. Like, is it done? Is it ready?
25:52Is it high quality? So a lot of people don't do this. They will use the same AI to review the work done.
26:00And that's why you get a lot of issues like, it said it was done, but it really wasn't. It was super broken. But why did it lie to me?
26:07Um, because it's reviewing its own work. Okay. So that's concept number okay.
26:13We're blasting through these. Okay. Six.
26:16So the final concept in loop engineering is memory. And you can think of this as like a . Okay?
26:23So that your AI agents on each run has context on what's been done, what else needs to be done, what should I do next? Now practically, when you hook up Claude code to your GitHub repo or your GitHub project, that serves as a sort of memory because like within that project, uh, for example, in my social media posting project, um, my agent tracks like all of the posts it's made.
26:50Like it has a little log file that tracks that. Uh, so practically here, you don't have to do anything if you've already hooked up Claude to your GitHub repo. But the reason this is important is without some kind of shared notebook across all of your agents, every run would start from zero.
27:09Like, it really wouldn't know what the previous agent has already accomplished, what blockers they're running into and what it should do next. So memory is an important part of loop engineering, but practically you can achieve this just by connecting your Cloud Code project to a GitHub repo so that your agents can write to that project, their status updates, what they've done, what else they're stuck on, and what they should do next.
27:34Okay. Oh, yeah. It's like a whiteboard.
27:37Okay. Oh, this is a good one. Okay.
27:39Cool. And by the way, for those who joined late, this what I'm showing on the screen is a YouTube video I'm gonna publish later. Um, and it was fully edited by Claude and Hyperframes, which is a new free open source repo.
27:54And it's pretty cool. Like, it's very cute and everything. All I did, I just filmed myself talking and then it edited it and added all of these visuals and interactive graphics.
28:03They're not moving right now because I just I don't I don't want too much motion so I can explain stuff. But all of these graphics are, like, animated and stuff. Here I can show this.
28:12K. And then oh, that's the whiteboard.
28:25It even cut out my face here. Oh, yeah. Here.
28:28I was just showing the animations. Okay. Cool.
28:33Okay. So, yeah, let me just summarize. So so in summary, loop engineering is all about removing you, the human, from the loop so that an AI agent has a goal, it knows what to do, it tries to accomplish the goal, it reflects on its own work or a separate agent, a checker reviews its work, and then it decides, like, if the goal has not been met, here's what I should do next.
28:57That is what Boris, the founder of Claude Code, means when he says he doesn't prompt Claude anymore. His whole job is designing these loops.
29:07Because instead of, like, I ask Claude, do this, I review the work, and then I tell it what to do next. Instead of the instead of you being the one doing that, you can have AI taking your role role doing that, reviewing the work, and then prompting what should the next step be.
29:24So here's the whole loop that we're going to build, by the way. So, basically have some kind of trigger like a schedule when your agent runs.
29:34It'll have a shared notebook or memory, which is typically your GitHub project. Um, if you have multiple agents working simultaneously, each one will have its own work tree, its own lane, so it doesn't collide with all the other agents.
29:47You'll have one agent that is predominantly doing the work, the maker. Another agent, usually faster and smaller, such as a cheaper model like Haiku, checking and reviewing the work independently.
30:00If the goal has not been met met, the checker says, hey, go back. Do this instead. Like, it'll give feedback to the maker in terms of what steps should be taken next.
30:10That's what people mean when they say, let AI prompt itself. The checker is actually providing feedback and, you know, in in context for what the next prompt should be for the maker agent who's responsible for doing the work.
30:25And then when it's when the work is done, you can open a PR, for example. You can email it to yourself. You can send a Slack message to yourself exam though.
30:35Okay. Cool. So I think that's it that I wanna talk about here.
30:39Uh, Yeah. Extern. Yeah.
30:43Okay. Cool. Okay.
30:45Do we have any questions on that?
30:48That was a cool video. Yeah.
30:54Uh, okay. If you guys have any questions on that, let me know. Just drop it in the chats.
31:01Are you actually using loops like this for Blotato features? Yes. So that's what my new AI agent is.
31:08Um, so I dropped a new AI agent in Blotato that uses like, literally exactly this loop.
31:16Um, and it does not use any bloated agent frameworks or anything like that. Um, it's really just built from scratch without any agent frameworks. Can I use codex to loop Cloud Code?
31:31You can, but I'll show you today. You can also do it entirely within Cloud Code. Could we do a quick summary?
31:40Okay. Oh, here. I put the summary here.
31:43I'm gonna send out this, um, this thing immediately after. Okay. So here were the six parts we talked about.
31:50Number one is oops. Sorry. Number one is automation.
31:53So some kind of trigger or timer that starts the loop on its own so that you don't have to be the one typing to start the work. K. We're gonna cover that in the last section, Cloud Code routines.
32:05Um, second concept number two is separating the work areas. So two AI agents do not collide with each other.
32:12Practically speaking, you'd normally don't have to set this up manually, and most people won't even have this need, to be honest with you.
32:20Like, you'll just have one AI agent at a time working on your project. It's it's a lot to manage, like, multiple work streams. So I would consider that, like, very advanced.
32:31But concept number three, skills. So having your loop access, uh, like repeatable playbooks for the different tasks you should do.
32:40Many of you are familiar probably with skills and connectors already. So the the concept here is just that your loop has access to skills as well as connectors. Connectors are, uh, allow your agent to use the tools that you use.
32:52So Gmail, Slack, uh, HubSpot, whatever you use, it can plug in. And sub agents, the concept here is one agent to do the work and a different agent to independently check the work so that your main agent is not grading its own homework.
33:07And then the last concept is memory. So that could be a notes file that lives outside of the chat. Uh, it's oftentimes a GitHub project with with, uh, text based markdown files within the project where your agents can log information, etcetera.
33:22Um, a lot of people now like to do, uh, a lot of markdown in Obsidian and use that as their, like, second brain memory.
33:31Yeah. But okay. So that that'll just stop there and summarize instead of rambling.
33:35Okay. So, okay. So those were the six parts of loop engineering.
33:41And if most of those look familiar to you, um, that's good. They should. Right?
33:45Loop engineering as a term is not a new concept in the sense that like people have been doing this, but it's blown up on Twitter recently because like someone really put a name around it and helped to kind of like organize a lot of the thinking that has been done around this.
34:03So, uh, if you're all you're probably already doing some of these. For example, if you have a Claude CoWork, uh, schedule that checks your Gmail every morning and surfaces emails that are urgent to you and labels or archives the rest, that is considered loop engineering.
34:22Um, it may not have all of the parts. Okay? But conceptually, you've gotten a lot of things there already.
34:29Okay. So that those were the key concepts. Yeah.
34:33I will caveat this. Like, it sounds simple, um, but actually the hardest part is defining the checker.
34:42Like, let's let's take a social media post for example. How do you define what a good post is? Like, it's it seemed like your task is like write a viral LinkedIn post.
34:53Okay. But like, how does the checker define what a great viral post is?
35:00And so, um, as we'll walk through the examples today, eventually you'll notice that actually defining what good and done looks like is the hardest part of creating the loop. Oftentimes, people already know the task they want done. Right?
35:16Like clean this up or write this thing or plan this out. But it's it's hard to define what good looks like, what done looks like. But that's okay.
35:25That means you're doing it right. So okay. Okay.
35:29So now we're gonna dive into the actual practical, uh, part of this tutorial. So make sure you have Claude code open.
35:36Uh, let me actually open it now on the right side.
35:42Oops.
35:43That's okay. I'm just trying to get the font size right, like, so you guys can read the font. Can you guys still kinda see the font, the stuff on the right side?
35:55Hopefully. And I know it's really tough on vertical streaming platforms. Um, I am on horizontal streaming platforms as well if you find it really hard to read.
36:06You can use cursor and other tools, but they won't have slash commands like slash goal, I think.
36:15I honestly, I haven't used cursor for real in, like, a year. So, and then I'm not sure about like routines either.
36:22Before Cloud Code routines, I just used GitHub actions. Right?
36:27So that, like, that's another option if you're GitHub based. Okay. Okay.
36:32Cool. Can you guys see my screen?
36:38No. It's it's actually the opposite. The checker is the small fast model.
36:42So the default checker in Claude code is Haiku. You can change it so it uses a different model. But typically, you want your primary worker to be a more advanced model and then your checker to be a smaller, faster model.
36:57Okay. Uh, okay. Cool.
36:59So, yeah, let's get started. So the very first thing we're going to do is just think about a task you repeat every week. So when do you start the task?
37:10Like, literally fill this out. Starts when? Is it a certain time of day?
37:14Is it when an email arrives? And then define what done looks like. So for example, in the morning at 8AM, I want to go through all my emails and label them accordingly or archive them and or and, uh, let's say, surface the three most urgent ones to my attention by DMing me on Slack, something like that.
37:37And then here's the check, how the AI proves it's done. That would mean, like, every email it has been labeled or archived.
37:44K? So think about this for one task that you repeat every single week. Um, and by the way, like, I recommend business owners in particular do this once a week.
37:57So this is an aside. This should not be in the final video. But, like, I have a new habit I have a habit tracker app that I always use when I develop new habits.
38:05And, like, one of them that I've used now for a while is always to automate, delegate, or subtract something I'm repeating every single week.
38:15Right? So if there's something I'm doing every single week, I either want to automate it with the help of AI, delegate it to someone else, or subtract it from my life.
38:23Just stop doing it altogether. Um, if you do not have that practice in place on a weekly basis, you'll often end up like just drowned in lots of random tasks.
38:34So but so anyway, that was an aside. But, um, okay.
38:37So I'm just giving people time to think about this one sentence they repeat, uh, or this one task they repeat every week. Okay. Now what we're gonna do is use the clod code slash goal command and make sure you have claud code set up in terminal.
38:51Okay? So I'm just scrolling down here to where the prompt will be. And similar to the concepts we talked about before, so you give claud your goal, like what is the goal condition.
39:01It key it will keep working on its own until the goal is completed. So how does that work? So for those of you who are technical and curious how this actually works, basically, slash goal is a wrapper around the stop post hook notification.
39:17So, you know, post hook notifications, like, uh, when the prompt when the you type in a prompt, Cloud Code runs, that's called the end of a turn.
39:26Okay. At the end of a turn, that's when your post hook stop notifications happen.
39:32For example, I have a post hook stop notification when Claude code is finished running. It emits a mooing sound. So if you've been on my livestream where you heard, like, lots of mooing, that was, like, Claude code running, and it was done.
39:46So what what they do at slash goal is, uh, on that stop post hook notification, they have another AI model.
39:55Uh, the default is Haiku. It serves as a small fast model to check whether the goal condition is met and decide, like, should we keep going or should we stop because the goal condition has been met. So that was a technical explanation that you can also cut out of the YouTube video.
40:10But for those of you curious how slash goal actually works, it's like it's actually a wrapper around the stop post hook notification and it runs that a small prompt at the end of each Claude code turn to check if the goal has been met. Okay.
40:25So here's how you type it. So let's just do a simple example. So how you type it is really simple.
40:32So you can just type slash goal. Okay? And it should turn blue like this, assuming you have Claude code in the terminal and it's updated.
40:39Make sure it's up to date. Okay? And then you can see here, it's trying to help you with documentation, like what is the condition that you want to keep you want AI to keep going until this condition has been met.
40:51So here's a simple example we're gonna do. Sort every file in my downloads folder into subfolders by type, keep going until no files are left, do not delete anything, and stop after 30 turns.
41:05I'm just gonna pull up my downloads folder somewhere. Where did it go? Where is my oh, no.
41:11That's not my downloads folder. Oh, here it is. K.
41:15Okay. So so let's do let's just paste this.
41:19Uh, you can take a screenshot of this. So let me just put it here so everybody can take a screenshot.
41:25Actually, I'll type it into the chat.
41:29Uh, that's too long to type into TikTok chat. K.
41:39Yeah. Nope. Sorry.
41:40TikTok. I'll type it into the YouTube chat. Sorry.
41:43Sorry, TikTok. I'll type it into the Instagram chat as well.
41:50Okay. That okay. Cool.
41:51So I typed it into the YouTube chats and the TikTok chats or not the TikTok chat, the Instagram chat.
42:00Sorry. Now I forgot what I was saying. Okay.
42:02Cool. So here just trying to I need to make it, like, readable. Okay.
42:06So here's my downloads folder. I'll put it down here. It's it's kinda messy.
42:11Now let's just run it and see what it does. Hang on. Let me move this here.
42:19Okay.
42:25So oh, I forgot I actually already have a cleaner skill, but, um, you won't have a cleaner skill. So when you run this, you know, you're not gonna see like the goal matches my cleaner skill. Um, but yeah.
42:37So now it's found 17 files. Um, it counted how many are in each category. Now it's creating the folders and moved everything turn by turn.
42:45Okay? And pretty much done. Found 17.
42:48Let me create the folders and move everything. All files moved, zero remaining at the top level. Now it's going to verify its work, verifying the counts.
42:57Done. Your downloads folder sorted. No files left at the top level.
43:01Nothing deleted. Boob boob boob. Okay.
43:03And then you could see goal achieved. Okay. So this one was a very simple example because, um, I've gotten feedback that my examples are too complicated.
43:13So I'm trying to stick to simple examples. Okay?
43:17So it was act it was actually completed very quickly. You know, however, I have run goals and and loops like this where it's running for five hours continuously. Okay?
43:26Um, so don't don't let the how fast this happen fool you. You can set a quite an ambitious goal and let it keep running for hours on end.
43:36Okay? But here, it was just a really simple example, and now you can see my downloads folder has those other folders organized and everything was organized here, videos, other images, etcetera.
43:47K? And so let's think about, like, con just conceptually what happened here.
43:52So here are the things here is how, like, that prompt was stuck structured. So Claude was moving each file at a time, checking what's left, and moves on to the next file and repeats. It stopped on its own when the entire downloads folder was sorted.
44:08And there were three things that made this like a good goal prompt. So let me scroll back up to it. Okay.
44:15So we have a clear end state, right, like keep going until no files are left. Like that's a clear end state goal. It had a check it can run, right, counts the loose files until it's still in the folder And a guardrail, do not delete anything and stop after 30 turns.
44:35Loose. I'm I don't like this word loose. Okay.
44:38I'm just deleting this from my love. Cool. Okay.
44:42So here's our goal. Sort everything. Keep going until blank.
44:47This is the end condition. So when you type this, um, the small fast model, which is default haiku, is going to check that no files are left. Like, that's one of the goal conditions that it must check.
45:00And then these are constraints, like or guardrails. Don't delete anything and stop after 30 turns.
45:06Um, because with a malformed goal prompt, it could, like, keep going for hours and hours unintentionally. And so it's nice to put some kind of guardrail like stop after 100 turns just so you don't blow through all of your tokens for the month or for the week.
45:24Um, okay. Cool. So that was a really simple example of a goal.
45:29Now I'm gonna walk through some intermediate examples, although I don't actually have any of these receipts, but I'll just talk through them. Okay. So here's one.
45:40So here are five intermediate level examples. You don't have to run them right now, but I'm just giving you ideas of what else you can do. So again, um, standardizing a lot of messy files until every name matches a specific format.
45:56For example, I have a ton of invoices and receipts. Okay? So you can rename them all in this particular format in date order and keep going until no file has its original name.
46:07Do not delete or move anything and stop after 25 turns. So this is one example. The end state here, like what is the goal condition that our checker agent can verify?
46:17We have zero files left with the old style name. Okay. Now practical example number two, categorizing a spreadsheet until no row is missing a label.
46:28So for example, let's say you have a big spreadsheet and you're trying to label every row in the spreadsheet. So the goal is fill in the category column for every row in expenses dot CSV. Here are the categories, food, travel, bills, shopping.
46:42Keep going until no row is blank. Do not change other columns and stop after 30 turns. Right?
46:49Okay. So this is the goal. This is the end state.
46:53Keep going until no row is blank. And here are our constraints or guardrails. Don't change other columns and stop after 30 turns.
47:01Notice here, like, we're not typing massive multi paragraph prompts.
47:06Okay? That's that's kind of what people mean when they say, like, prompt engineering is dead.
47:11It's not it's not really dead. It's just it's just this is the next evolution of prompting.
47:17It's like, do we define this loop so that an agent's loop can, like, verify its own work, check what else needs to be done, prompt itself to figure out how to complete the work, and then check again if the work should be done. So the end state there was making sure there's no rows missing a label, zero blank cells in the category column.
47:37Here's practical example number three, working through a stack of documents until none are left. This pattern is called empty the queue. Um, I'll talk about this in more detail when I share some Cloud Code routines I have for customer support, like processing every open ticket until none are left or processing every email or processing every unread email until none are left.
48:00That's like the empty the queue pattern. Um, okay. So here slash goal, write a five line summary of every PDF in my reports folder.
48:09Keep going until every PDF has a summary. Do not edit the PDFs and stop after 40 turns. Okay?
48:16So again, we have a clear goal, we have a clear end condition, and we have our guardrails. The end state is every PDF must have a matching summary.
48:25Practical example number four, cleaning up a bunch of social media captions until they fit all of your brand voice rules. Uh, for example, rewrite every caption in this file so that it meets this character limit and has no hashtags. Keep going until all of them are done.
48:42Don't touch any other file and stop after all of these turns. The end state is zero captions over a 150 characters or containing a hashtag. And then this one is turning a list of rough ideas into hooks until the list is done.
48:56Turn each of these ideas into hooks, keep going, don't change other files, and stop after 30 turns. All 20 of these can be rewritten, uh, as hooks, less than 10 words, then this is the end condition.
49:09K? So here's the cheat sheet number two. So this is just generally the framework to think about when writing your goal prompt.
49:17Slash goal, what done looks like, Checked by how, like how do you prove it? Is it account?
49:24Is it that all of your tests pass? Is it that you're you have zero bugs in this area and stop after X number of turns?
49:33Okay. So I know that these are to me to me, they're simpler. K?
49:39But, again, I got a lot of feedback saying I need to have simpler examples. Um, but this is the foundation for you writing way more complex autonomous agents.
49:50Um, it's it's this framework. It's this cheat sheets of of prompting. It it's it's it's it's like not even about prompting.
49:58It's like how you are defining the task. Like, what does the end condition look like? How do you verify it?
50:04How do you prove that we've met it? And some kind of guardrail, stop after x number of turns.
50:10Okay? So practice that. Like, try it with some of these examples.
50:14You can tweak them so it makes sense for your use case or industry, but like really get in the hang of thinking in this way. You want to think in terms of loops where there is a verification step in the task itself.
50:28Because a lot of people actually don't ever add a verification step. Um, it's something that I teach, right, like don't stop until. But having observed thousands of people prompting, I can tell you that the average prompt I see is like four words long.
50:43So, um, if this is basic to you, that's a good thing.
50:48Um, don't worry. We'll get into the advanced stuff. So okay.
50:53So, yeah, hopefully, everybody took a screenshot of this. So, like, keep it as simple as possible because you're you're actually retraining how you're prompting.
51:02Okay? So, like, I don't care how simple it looks to you. Just try it in this framework.
51:07Keep trying it in this framework. Obviously, if you're pro, you're gonna create a skill that's gonna help you write this prompt properly. Um, but yeah, just keep doing it in this framework so that you train your mind to prompt in this way.
51:20Okay. So now we're going to build our first autonomous agent with a Cloud Code routine. So slash goal is great.
51:28It runs, like, while you sit there. Um, but a Cloud Code routine basically adds that timer.
51:35It adds the automation aspect that I talked about in the first concept of loop engineering so that the loop can run without you. Remember that the whole goal of loop engineering is to remove you, the human, from the loop.
51:48It's about like how do we set up a system so that it can just keep looping until a complex goal is achieved. So, uh, Cloud Code Routine is really awesome.
51:59If you come from an automation background, like you've used n8n or make.com or Zapier, um, it's I wouldn't say it's a total replacement for them, um, because it's a little it's a little less deterministic than those workflow platforms, but it's it's super awesome thing to have in your toolkit.
52:17So a clock code routine is basically just a saved AI loop that runs by itself on a schedule, um, specifically in the cloud if you're using a Clod code routine, even when your laptop is closed.
52:29Okay. So you don't need your computer on all the time to run a Clod code routine. It's actually running in the cloud somewhere.
52:36Um, it sets up a fresh environment, pulls your GitHub project and then runs, uh, the skills. It can use MCP connectors. It can make API calls, but it runs in that isolated environment.
52:48Um, quick aside, so if you are advanced and you are and you like the concept of Claude code routines, but you wish you could run it locally, um, just check out launch d, tell Claude to set up a launch d, uh, schedule for your tasks and that way you bet that way your, uh, routine on your computer runs and has access to all of your local files and setup.
53:11So if you don't want to run this in the cloud. K. Okay.
53:14So to set this up, we're gonna go to claud dot ai slash code slash routines. K? So that's over here.
53:21So let me just type the link in the chat. Claud.ai/code/routines. Okay.
53:32Plug okay. Okay. So yours will probably look totally empty.
53:37I have a few here that we're gonna walk through as examples. Okay. Oh, yeah.
53:42We're gonna do this one. Daily email cleanup. Okay.
53:46We'll we'll do this simple example. So what about let me go to the edit. So go ahead and click like new routine up here.
53:53Your your screen is gonna look like this. I'm going to pull up, uh, what I already have and you're gonna copy it and we're gonna talk through what it does. Okay.
54:01So click new routine and then you should have a form that looks like this. Uh, again, you can think of a routine as a loop that is running on its own in the cloud. You just set up a trigger which could be, uh, a daily or weekly schedule.
54:17It could be when a GitHub webhook fires or it could be, uh, you can trigger it yourself sending a post request. We're gonna keep it really simple and create a daily email cleanup agent that will read your unread emails and post the three most important ones to Slack.
54:33In my case, I'm just gonna DM myself one line each and here's the guardrail, do not reply to anything. Okay? So I'm going to paste these instructions, uh, in the chats.
54:44Oops.
54:47K. You won't have Sabrina, obviously, in your Slack, so you can just change that.
54:54Okay. Cool. Okay.
54:58So that's the this is the name of your team. These are the instructions. This is a very simple one.
55:03We're going to read all of your unread emails, figure out the three most important ones, and send a Slack message to myself. And do not reply with anything. This is the constraint or guardrail.
55:12This is where you can select a project, but you don't have to. So in this particular case, I don't actually need it tied to any particular project. It's just reading my emails.
55:21This is where you select the model. Okay? So Opus 4.8, etcetera.
55:26And this is the trigger. When do you want this loop to run? Okay?
55:29In this case, we're gonna say run daily at 9AM. So just click schedule and then run daily at 9AM. You can edit this as well.
55:38K? Oops. K.
55:41And then connectors here is what this loop will have access to. So remember, one of the key concepts of loop engineering is connectors, the ability for your AI agent to use the tools you use. For this example, we're just going to use Gmail and Slack.
55:56To add connectors, you can click add connector here. And we will cover how to add tools that don't have these nice connectors and stuff like that.
56:05We'll cover that later. Okay? There are additional options here.
56:08We're not really gonna cover behavior, notifications, permissions.
56:11K? So just make sure yours looks exactly like this.
56:16So name of the routine, simple instructions, daily at 9AM, and then these two connectors.
56:23And then go ahead and click save. Okay.
56:31If I make a skill, can that be triggered from a routine? Yes. So we were we were gonna cover that.
56:36This is just the simple example. Um, Yeah.
56:40So just make I'm just leaving this up here so people can make sure it looks like this. Okay? So, uh, well, click save so then you'll have it here.
56:48This is where you'll see, um, all of the routines you have running. So I obviously just created this one. Right?
56:55You can test it by clicking run now. So go ahead and click run now in the top right corner and it will say workflow run started. K?
57:03You can actually click this to expand it and see what it's doing. Um, I'm gonna close this because it's going to reveal a lot about my email inbox. But in a second, we should get a summary in my Slack with my top three most urgent emails.
57:19Okay. I'm just gonna check here if there's anything else I should cover.
57:25Oh, interim. Okay. We'll get to that next.
57:28Um, Yeah. If you have questions and stuff, put this would be a good time to post it in, uh, YouTube or TikTok.
57:41Okay. This I don't know how long I do have a lot of unread messages in my email.
57:50K. Searching for Gmail and Slack. Now it's fetching a bunch of tools.
57:55Now I have the emails. Now it's going to assess the most important emails. Like you can see its entire thought process here.
58:03And by the way, you can continue the conversation. So let's say it did something, maybe you want more information, you can come here in routines, open up the chat, and continue talking to it. Like, you can ask questions here.
58:14You can have it run additional things, etcetera. K?
58:24Okay. So if your connector is not listed, you will need to go back to the actual claude.ai interface.
58:33So go to claude.ai. If the so if you go to customize, connectors, and then this is where you can add your connector.
58:41So click the plus. Uh, usually for most apps, they have an official connector, so Gmail here. K?
58:48And and then once you add it, it'll look like this. A lot of people ask me about permissions. So this is also where you can manage the permissions of all of your connectors.
58:56So always allow needs approval or blocked. If you're a beginner, I recommend keeping it at always allow for read only things. Meaning like it can read your content, but it needs approval if it's going to write, edit or delete anything.
59:11Okay? So that's a pretty good configuration especially if you're a beginner. Okay.
59:15Let's go back to our routines. Let's see if that finished running. Okay.
59:22So that one finished running. So you can see check mark here today at 12:44PM. So let's go to my Slack and see if it DM'd me.
59:30Yes. So okay.
59:33So here are my top three unread emails. You can see the timestamp is literally right now. Someone sends a long book excerpt.
59:40Okay. Which plans for Blotato showcase at some event?
59:47Okay. Interesting. So these are my top three most urgent emails.
59:51It was sent to my Slack. K? And, you can open the conversation here.
59:55You can continue the conversation if you want to update anything, and you can update the The routine itself. So once you create it, it's kinda gonna look like this.
1:00:04Click this pencil button in the top right corner if you wanna go ahead and update it again. Okay. That was the simple example.
1:00:13I'm just gonna talk through intermediate examples and then I'm gonna show you an advanced example which is like my own routine. Okay. So here are some intermediate level examples.
1:00:23Let's say you do have like thousands of unread emails. It can completely sort your inbox, label every single email and archive the ones you don't want. Okay?
1:00:33So you can set that up as a routine running on a daily or a weekly basis. Um, honestly, it it this one's really awesome. Um, so, uh, just just to help label emails even like label receipts, label filter these out, archive these emails.
1:00:49So, uh, if you have a monthly report, you can create a Claude code routine that every single month pulls data from certain systems. Let's say your CRM, uh, and your email marketing system, and it'll write the first draft of your report and send it to you in Slack.
1:01:04Okay? So that's another idea of an intermediate level example. If you have a lot of news you're trying to keep up with, you could say every Monday, read the updates on this topic and hand me a short brief so I stop doom scrolling and but I stay up to date on the news that I care about.
1:01:20You have another Claude code routine that does this. So what I would recommend is picking one of these or a variation that you resonate with so you can make your first routine and make sure that it's running autonomously every day or every week.
1:01:33Okay. So now let's walk through a real example because I know people have questions like, well, how do you integrate skills? What about other APIs, etcetera?
1:01:41So that will all be answered here. So, um, let me just pull it up. So now I'm gonna show you a real example of a Cloud Code, uh, loop including the main agent that does the work and the checker agents that independently reviews the work.
1:01:55So it's gonna be these two. I wish I could move them next to each other.
1:01:59But, um, so this one day every day runs at 8AM and cleans up support tickets and also surfaces any issues and gives product road map recommendations. This is the checker. So it will review all of the closed tickets that were closed by this agent and confirm they should have been closed because I don't wanna close support tickets that should have been escalated to a human.
1:02:23Okay? So let me open this one up. Click in the pub.
1:02:27Okay. Cool. So here's what this looks like.
1:02:29So it's act it's actually not that much longer, but it utilizes different things. So here I have a skill already in this GitHub project.
1:02:39By the way, this is where you can add your GitHub projects. For me, help.bloatato.com is where I host my support documentation and it also has my skills related to customer support.
1:02:49K? So run the cleanup ticket skill on all open customer support tickets using the intercom API, which I'll talk about as well.
1:02:58Provide a summary in Slack hashtag support channel. And when referencing conversations, always provide the full link so I can easily click and open it.
1:03:07Do not delete tickets. Here are constraints and guardrails. And do not stop until you've processed all currently open tickets.
1:03:14This actually isn't necessary because this is already the goal condition, but just one it's like wanted to just make it extra clear here. Um, here's the trigger. This runs daily at 8AM and then here are my connectors, intercom and Slack.
1:03:28Okay. In this routine, I also actually have, um, an API key. So what you can do is you can create a separate environment that stores like your, uh, environment variables.
1:03:42So the default environment for me does not have any environment variables, then my support environment contains my API key for Intercom. And the reason for that is, uh, using the Intercom API, I have access to a lot more things than using the Intercom MCP.
1:03:59For example, the Intercom MCP does not allow me to close open tickets, but the Intercom API does. So for those of you who are feeling limited by connectors or you wanna integrate an API that doesn't have a connector, uh, for example, I'm I don't think Perplexity has a connector to Claude since they're competitors.
1:04:19Uh, maybe now they do. Um, but you can you can just create a new environment, like click add environment and then you can put your environment variables here.
1:04:29So let's say this is our research environment and we wanna put our perplexity API token here equals and then put the token here.
1:04:41So this is what it would look like. K? So any tool that does not have, uh, an MCP that fills that fulfills your needs, you can just integrate the API into the routine just like this.
1:04:53Just create a new cloud environments, call it whatever its purpose is, and then put your environment variables, your API keys and tokens right here. K? Um, so I have one for support, um, and then that's it.
1:05:06So it will run daily at 8AM. It will analyze all of my open customer support tickets. It will automatically close the ones that no longer need help or need to be escalated.
1:05:16It will also look for, um, any urgent issues like outages, bugs that are urgent that I need to fix. And it's all tied to this cleanup ticket skill.
1:05:27Okay. So there's actually a lot in this skill that I'm not showing, but that's the whole point. Like, you can create these skills in your GitHub project, and then your AI agent loop just can use these skills, uh, in the process.
1:05:40K? Okay. Cool.
1:05:43So, um, okay. Let me close this.
1:05:46What else? So this is the this is the maker agent, so the primary one doing the work, classifying tickets, analyzing them. And then I have the checker agent here.
1:05:56So this will run about two hours after the other one. And its job is to independently verify all of the support tickets that were automatically closed today.
1:06:06Confirm that each ticket should remain closed. If the support responses received by a customer do not clearly solve their problem, reopen the ticket. So again, in loop engineering, the concept here called it sub agents.
1:06:19One primary agent to do the work and another agent to verify the, uh, quality and accuracy of the work. So this is the checker agent that, um, helps verify it. So it's set up similarly.
1:06:31Right? So, uh, the only difference really is it runs two hours after the first one runs because the first one takes a while to actually process all of those tickets. Um, and then you can see examples of it in Slack.
1:06:43So if I go to support. So okay.
1:06:48So here's the ticket cleanup run. K.
1:06:54Just a lot of tickets. Okay. And then let me open this.
1:07:02So this is the cleanup checker, and it was running. So okay.
1:07:08So here, I'm just pulling up a summary. Okay. Yeah.
1:07:10So here in the first run today at 10AM that I triggered manually, the full audit was complete. A 151 tech tickets were currently closed, uh, or a 151 tickets were correctly closed. One ticket needed reopening.
1:07:23K? So this is an example, again, of the checker verifying the work, and it found that one ticket needed reopening. Okay.
1:07:32So that is a more advanced and realistic implementation of an autonomous agent. So I'm just scrolling here.
1:07:40Yeah. And then I I mean, I covered this in the blog post, but the next level is to integrate any a API, uh, with your routine, which I talked about a little bit here.
1:07:49Right? Like, you wanna edit this, click this cloud and this is the environment. Click add environment and this is where you can put all of your API keys for any tool you want your Cloud Code routine to access.
1:08:01Okay. And then, yeah, just some tips here. Like if you are brand new to all of this, just start read only.
1:08:07Like instead of having it reply and close tickets, just have it summarize your emails, like label your emails. And that's why I started with that as a simple example.
1:08:16Like I know it's this for people in AI who've been doing this while, it's like a really simple example. However, again, I've gotten feedback that my examples are too complicated. I should start with simpler examples.
1:08:26So that that's a nice one because it's just reading your emails. It's not replying to your emails. Um, it's just reading them, labeling them, and surfacing urgent emails.
1:08:34Um, and make sure you set your constraints and guardrails. You know, do not reply to email or do not delete anything. In my case, do not, uh, do not delete any support tickets.
1:08:45And then routines use the same plan as your normal Claude, and there's a daily limit in settings, so just be mindful of that. And that's another reason why we have a guardrail like, you know, stop after x iterations. So here's the cheat sheet number three.
1:08:59Uh, so ship one tiny routine this week. Think about when should it run, what is the trigger, what should it do, fill that in here, what are the tools it needs, connectors and API.
1:09:14Make you can make a list here. And what is the limit that you want it to have? Like, don't reply to emails or don't delete something.
1:09:21Okay? So, um, like, think about this cheat sheet and fill it out, and this will help you create the Claude code routine. By the way, Claude can create the routine for you.
1:09:30If you like literally copy paste this into Claude code and be like, create a Claude code routine and we fill out these blanks, it will create it, um, mostly for you. I think you will need to go in here and still create your environment variables and stuff.
1:09:44But okay. So, yeah, start read only.
1:09:47Let it summarize stuff for a few days before you let it write anything, edit anything, or delete anything. But definitely use this template to again reorient how you've been prompting so that now your your prompting is all about creating these loops.
1:10:02Okay. So just to recap everything we've talked about. So we talked about loop engineering is a shift from typing prompts to designing this loop Slash goal and routines are things you can use to make loop engineering a reality much more easily in 2026.
1:10:18So slash goal and then you type the goal you want it to keep going until it hits that goal. And then there's going to be a second verifier agent that checks whether that condition is met. Uh, and then a routine, it's like a timer, right, and that can be connected to your skills, your connectors, your API keys, your notebook, which is your GitHub repo.
1:10:39It's connected to all of those things, which are all key concepts in loop engineering so that it can run the loop in the cloud without you. And that will be your first fully autonomous agent that where you are removed from the loop. Now the hard part, like, I I really try to emphasize this, like, the the hard part is, um, is defining what done is.
1:11:00Like, what is good? What is good enough? What is done?
1:11:04Um, that's why, like, my personal formula whenever I think about how am I using AI effectively, the amount of AI leverage I personally get is always a function of like my skill and my clarity.
1:11:15This is like the formula I I personally think about all of the time. By clarity, I mean my ability to define what it is I wanted to do and what done looks like, what does good look like.
1:11:28And skill is like my ability to review the AI agent's work so that I can further improve the loop. Um, so a lot of people like this simple formula because it's cool, easy to understand. Um, but this really is how I think about AI.
1:11:43Like, if I have no skill in something, um, you know, AI can help me build up some initial skill fast, but I'm just not gonna get the same amount of leverage as someone who is already very skilled in the thing. This is why like senior developers are like so much more productive using AI tools than a junior developer.
1:12:03Because you have the ability to review the work, identify clearly like what is wrong and how the loop should be improved.
1:12:13Like maybe you didn't clearly define what good looks like or what done looks like. Maybe your, uh, end condition was make sure all the tests pass, but your tests were garbage. Right?
1:12:23Like, you have to go improve the tests and that requires skill to be able to, like, recognize that and do it. So this is really my personal formula whenever I think about how do I use AI to maximize my leverage.
1:12:36Where do I have skill? Where do I have clarity? Like, to to be able to define what it is I want and what good and done looks like.
1:12:44Okay. Um, so, yeah, that's pretty much it. So, hopefully, that was helpful.
1:12:49I guess I'll just do another recap here with all the lines. Where's my loop?
1:12:54Okay.
1:12:57Man, this video is so cool. It was made with AI.
1:13:02I'm looking for the graphic with a circle. Yay. Okay.
1:13:06Cool. So here we are. Um, okay.
1:13:09Cool. So yeah. So okay.
1:13:12Let me so so in the beginning, we talked about, like, what are the key concepts of loop engineering because, like, that's the latest buzzword. The reality is people have been doing loop engineering. We just didn't have, like, a collective name we all agreed on.
1:13:26Um, for many of you who have been following my content, you should be familiar with at least half of these. Right? So concept number one in loop engineering automation, like running it on a trigger or when something happens, do this.
1:13:38That way you're not the one sitting there prompting it to happen. It's just happening without you starting the loop. Um, number two is work trees, which means let's say agents are working on the same code base.
1:13:52You want each one to have its own safe space so it can experiment without colliding with all of your other agents, um, Git work trees. Um, practically speaking, like most people don't have to do anything to set this up. And practically speaking, most people won't really need this unless you are really having a lot of agents, um, modifying like the same exact things.
1:14:12Um, number three is skills. So many of you are familiar with this. Um, it's basically a playbook that your agents can follow to do a task the same way every time.
1:14:22And so you don't have to give it that context and instruction every single time it does a task. And number four, connectors. Uh, just like we saw connecting it to Gmail, connecting it to your support desk, connecting it to your CRM, connecting it to your content creation tools, connecting it to Potato, for example, to publish the social media.
1:14:41Connectors are what give your AI agents the ability to use useful productivity tools, the ones you use every single day. Number five, sub agents.
1:14:50The key concept here for loop engineering is you want one agent to do the primary work. It's usually the smarter, bigger model. So let's say Opus Sonnet 4.8 or Opus 4.8 and then you have another agent that's typically the smaller, faster model like Haiku reviewing the work independently.
1:15:08Like, has this goal condition been met? Yes or no? Haiku is going to check that.
1:15:14If the answer is no, k, the primary agent has to go and keep trying until it completes the goal. And the last concept for loop engineering is memory. So it's like a shared notebook for every agent run to have the same context of what's been done, what's been tried, what are we struggling with, what needs to be done next.
1:15:34Practically speaking, if you have a GitHub project already hooked up to your Claude, that is going to serve as the memory as the scratch pad.
1:15:42Okay. So those were all six concepts for loop engineering, and then we also talked about how to use clod code slash goal to make your first loop and also how to use clod code routines to make your loops autonomous running without you.
1:15:57That's it. Yeah. Oh, man.
1:15:59My voice is gone. But okay. If you have any question I'm gonna stop recording now.
1:16:05Uh, but if you have oh, wait. OBS is on my other screen. Okay.
1:16:08I I'm gonna stop recording. Otherwise, the file is gonna be really large. But, yeah, if you have any questions, just drop them in the chat.
1:16:15Um, I'm checking both YouTube and TikTok chat right now, and I just have to go stop the recording.
1:16:36Okay.
1:16:39Sweet. Oh. Oh.
1:16:42Oh, okay. So, yeah, drop your questions in the chat while and let's just, like, compile the questions. I need to order sushi real quick because I'm really hungry.
1:16:51But I'm also on a diet. Okay?
1:16:55So okay. But yeah. You can drop your questions in the chat and I'll check-in a second.
1:17:01Oh, wait. Let me ask if I need to ask if my friend wants sushi too.
1:17:12Who you want sushi? Okay. Uh, where is the template?
1:17:18I am going to drop that later today. Just make sure you're subscribed to my newsletter, which is my link in bio.
1:17:27Yeah. The cheat sheet will definitely help.
1:17:32Just built a playbook.
1:17:35What about what's the difference between doing this and in CoWork? Um, I think CoWork doesn't have the slash goal command yet.
1:17:45So as far as I know, that's just in Claude code. But you can you can definitely do pieces of it in CoWork. Right?
1:17:51So Claude CoWork has the schedule thing that is, uh, similar to Claude Code routines so you can schedule recurring tasks and stuff.
1:18:07What is the best way to start from the beginning to auto post with Claude through Canva? What is the best video? Oh, see.
1:18:17I think I have a I have a Canva video somewhere.
1:18:26Yeah. Probably this one. Okay.
1:18:30I'll drop this in the chat. So this livestream will it'll probably take a week for the editors to edit it, unfortunately.
1:18:41So is there any AI that is actually free? Yes.
1:18:47Open source models are free.
1:18:52You can try setting it up one up with, like, LM Studio as one kind of app. Okay.
1:19:01Can you stop a goal in the middle of the execution? Yes. Just type slash goal clear.
1:19:09Okay. Like this. And, um, also, if you type slash goal while it's running, it will give you a status update.
1:19:16Like, it'll say, here's what I've tried and here's what we're trying to do, etcetera, etcetera. So you can actually type slash goal while it's running. Okay.
1:19:25My friend's not replying to me about sushi.
1:19:32Should I just order her sushi just in case? Okay.
1:19:36Yeah. I guess that's that's what I should do. Okay.
1:19:42Okay. Oh, yeah. I need to get my eel, Unaki.
1:19:46Yes. Okay. Oh, I'm so hungry.
1:19:51I stayed up till one last night because this, um, hyper frames is so fun editing this video. Okay. Cool.
1:20:01What was the skill you used for the presentation? Um, I'm not sure what you mean actually. So all of the prompts will be here in this newsletter which I will send out literally like today after this livestream.
1:20:14I just wanna check it one more time. Am I charging 10 k for this? No.
1:20:19It's it's actually free, but it's worth 10 k. So I put 10 k in the TikTok title, which totally free will create animations at least three minutes long.
1:20:30Check out hyperframes, Remotion, and, um, LTX.
1:20:38Oh, hang on. Where is the questions? Say, if you have questions, now is the time to drop them.
1:20:43I have twenty minutes left. Oh.
1:20:50I'm a video editor. Oh, yeah.
1:20:56That's okay. I have too many video editors now.
1:21:01Why do I do Tik Tok instead of consulting for companies that will pay you loads of money? Um, that's a good question.
1:21:13Honestly, it's not like that fun. Like, I've been in a big company and honestly, all of the productivity gains and improvements that you deliver don't trickle down to employees.
1:21:25Like, they really go to the shareholders and the people at the top. So I am more interested in, like, just democratizing this knowledge on TikTok for, like, normal people so that you can make your own money because I really don't believe it will trickle down to you. Even if I like, yeah, I can teach a company.
1:21:42It will save them or make them billions of dollars, but, like, that's not gonna help you, the employee. So if you think so, uh, well, that's unfortunate for you.
1:21:52Um, but yeah. Yeah.
1:21:54I mean, I also yeah. I I don't know. A lot of people ask me if I wanna do AI consulting, but it's first of all, I'm very introverted.
1:22:01I don't like interacting with people. So, like yeah.
1:22:05I'd rather just sit here in my shitty corner.
1:22:11So, um, how many people do you think will use these loops?
1:22:16I think it's it's really not common mainstream knowledge at the moment. Like, e even before Cloud Code goal and routines, like, you know, I frequently tried to prompt don't stop until this, but it's it's not perfect either because it's using the same agent to do the grading, which as we as we see, you know, it it, um, it just lies and hallucinates that it did a good job.
1:22:40Um, so not many people are like really using loops in this way. Like, ninety nine point nine nine nine percent of people I see, I swear, it's just they type a prompt, chat answers, they read the answer, then they think about what they should prompt next.
1:22:56They type the prompt. They wait for chat to finish again. They read the answer.
1:23:01Then they think, oh, what should I type next? Like, this is the way 99.999% of people use AI today.
1:23:08One like, for sure. If you don't think that, you probably exist in a bubble and like should go talk to like real people. Um, but yeah, real people, that's how they are prompting today.
1:23:19Very few people are really setting up loops like this that are run running autonomously almost all of the time. Um, and it's not that it's like autonomous forever.
1:23:28Like you you do have to be very involved in reviewing the output and improving the loop. So even for my like customer support agent, I'm very involved in like figuring out, oh, it got this wrong, it got this wrong, or like I just changed the product and some parts of my help docs or earlier YouTube videos are now out of date, you know.
1:23:46It's like, um, in a business context, like it's a very dynamic problem. Like I'm always improving systems, ripping about ripping out things that no longer are working, replacing them with a more efficient way of doing things.
1:24:00Like, it's changing all of the time.
1:24:06Okay. So you have a way to use AI for content generation?
1:24:14Yeah. That's what I was showing earlier in this thing. Hang on.
1:24:18Where did the video go? So, yeah, that video I was showing was like 100 edited with Remotion. So draft.
1:24:27I'm working on storyboarding it as well. So like, by storyboarding, I mean.
1:24:35So basically like I'm getting a lot of YouTube wait, can you guys still see my screen? Yeah. Okay.
1:24:40So basically by storyboarding, I mean I'm getting a lot of feedback that I should spend more time in preproduction, scripting my videos, etcetera, so that they're clearer.
1:24:50And so I've actually been experimenting with using this free AI tool. By free I mean like it's a 100% free. The only cost is like Claude to call it, but you can use an open source model instead of Claude.
1:25:02And so I'm experimenting with a pipeline that, um, storyboards a video and then I film my talking head.
1:25:11So that way I can like see, like, do my words, are my words landing and is the visual like helping to clarify what I'm saying? So this is an example storyboard actually.
1:25:23So I fed it the same script but I didn't give it my video. So here it's actually just storyboarding like what each scene should be. Like here, this is going to be my talking head.
1:25:33Okay? And then this other scene is gonna be this visual. And then it actually uses, um, text to speech in hyper frames so there's actually voice.
1:25:42So I can hear the the AI voice, not my voice, but I can hear the AI voice with the visual and then make a judgment like, is this landing? Is this the simplest way I could explain things, etcetera.
1:25:55Um, and so this has been really interesting actually to experiment with. Again, everything you're seeing on the screen is a 100% free and AI generated. Um, and you can hook it up to an open source model so you don't even have to pay for clod tokens.
1:26:11Right? So so yeah, this has been a very helpful exercise for me because I'm not a visual person. Like, I I enjoy writing so I actually don't mind spending more time like writing and prep, um, but it's very hard for me to translate like what I'm saying into visuals that actually clarify the point.
1:26:28That's why like I usually just have lots of text open on my screen because that's that's how I process information. Like I literally don't watch YouTube videos. I would rather read help docs than watch YouTube videos.
1:26:39Um, I don't like yeah. That's just how I learn. And so it's something I struggle with for YouTube, like how do I convey some of these complex concepts in visuals that are easier for visual people to follow.
1:26:52Right? That's why they're watching YouTube videos. Yeah.
1:26:54I know a lot of people are the opposite. A lot of people get overwhelmed when they see a lot of text. Like, am so weird.
1:27:00Like, I read every single legal document ever in my first startup. Like, I read it multiple times. Like and I enjoyed it.
1:27:10Like, that's weird. Like, I know that's weird. Like, I'm super into reading and writing, and I'm completely not visual at all.
1:27:17Like, if if it like, an editor is like, oh, can you give me feedback on this video? And I'm like, well, I don't really know what to say. Like, I guess I like this part.
1:27:26I got bored at this part. Like, I don't have much more feedback beyond that.
1:27:31Um, so this working on this has just like helped me personally, like bridge that gap between like, what am I saying in my script and is the visual helping to clarify what I'm saying.
1:27:43And I just started this like last night. Was up until like 1AM working on this.
1:27:48I'm call calling it my storyboarding pipeline, storyboard skill, right? So it's it it takes a script and then it creates like the visuals that are gonna be with the script and then I'm gonna go back and tweak the script if I'm like, this wording isn't landing with the visual or this visual isn't landing with the wording.
1:28:05I need to change one of them or both of them. Uh, Yeah. This is just called hyperframes.
1:28:10So it's super easy to set up. So if you just tell Claude, set up hyperframes and create a two minute horizontal YouTube video just about loop engineering, whatever.
1:28:22Like, just put that one sentence in and in ten minutes, Claude code will be done and you will have a, whatever, two minute YouTube video about loop engineering. So, yeah, just do that. Uh, okay.
1:28:36Best but full pipeline.
1:28:41Is PC or Mac better? Honestly, you're nontechnical, I generally recommend Mac. If you're really technical and you like playing around with open source models especially, you will probably get a PC with Linux.
1:28:54Right? Like some kind of Ubuntu flavor. Yeah.
1:28:56This is HyperFrame. So for a lot while, I use Remotion and I have a shorts editing pipeline in Remotion, but I was playing around with HyperFrames last night. So the difference is hyperframes is HTML based and, uh, Remotion is not.
1:29:10I feel like it's React based. So you can you can extend Remotion like very significantly.
1:29:17Like you can you can have, uh, Remotion make three d stuff with three JS and stuff. You can have it make like cool math visualizations with anim. You can extend it really significantly whereas hyperframes, you can't extend it that significantly but it is easier to use out of the box.
1:29:35So if I had to make a comparison, it would be like Hyperframes is Apple or iPhone and Remotion is Android, Google.
1:29:44K? So choose your own adventure there.
1:29:49But, yeah, that's based on my, um, testing so far. Um, it's really fun though. It's it's teaching me a lot about, like, maybe I should have more visuals in YouTube videos to, like, clarify complex concepts or not.
1:30:03I don't know. We're we're like, I'm just I'm trying all this out, you know. In my ideal world, I would just stand at a whiteboard and just, talk all day about AI.
1:30:12But, you know, it's helpful sometimes when there's, like, actual visuals that are not, like, my hand drawn chicken scribble. So, um, how can I become your YouTube editor?
1:30:23Honestly, I have too many YouTube editors and I'm slowing down on video production because, um, well, basically, I'm just getting advice that I should spend a lot more time scripting in preproduction. So for the next month, that's what I'm gonna try.
1:30:36Yeah. So like honestly, if you just want it to like look really good like this with as minimal work as possible, use hyperframes. If you are like, if you really want full control, you're gonna have your own design system for your video pipeline, you want to extend it with other tools like three j s for three d visualization, then invest in Remotion, far more extensible.
1:30:57Okay? But I'm guessing most people have an iPhone for a reason because it's like, we don't need all that.
1:31:04We just need to make a decent video. Okay?
1:31:10So we love my hand drawn info. The thing is, like, even when I vibe board, so if you have been to several livestreams, even when I vibe board, it's like I just I just write text. I'm I I like I'm literally so not a visual person that when I whiteboard, I just write text, you know?
1:31:26Like so, um, but it's been fun. It's like it feels like a stretch goal.
1:31:33Like, oh, I have to really figure out what, like, what is the best visual that will convey this point, you know? Like that's it's hard for me, but it's use I think it's useful to learn that skill. So that's why I'm investing and practicing that.
1:31:46Uh, and AI is helping me with it. Like, again, this storyboard was completely generated by AI with a voice over. So I can like literally hear like, oh, that doesn't quite land or the visual is has the wrong timing.
1:32:00Like the visual should be later a little bit. Um, or it's kinda confusing here. Like why am I saying these things but the but text says these things?
1:32:10Like, you know, there needs to be, like, coherency, like, coherence in the thing. So okay.
1:32:20Yeah. A lot of people switch to Android for the camera because it does have a better camera. But then they're like, screw it.
1:32:26I miss Apple. Like, everything just works in Apple. So I am doing an in person meetup in the school.
1:32:35We literally have an in person meetup next week. So if you're not in the school group already, like, I'll post the link here in YouTube. But it's literally the best free community you will ever join, I promise.
1:32:46And if that is ever not true, I hope somebody yells at me. Um, how has full stream marketing operations gone with this for SEO, social media? Yeah.
1:32:57So like how my agents, okay. So how my agents actually look on a day to day basis is I have like a terminal open with three to four agents, each one working on a goal. So, um, like, I'll I'll have it be running throughout the day, then I'll remote control so I can just like walk around the house and do other stuff.
1:33:16So it's like literally just running, working on the goal. When it's ready for me to check something, I check on my phone to see what's being done. If it needs a lot of feedback, then I'll go to my computer and be like, okay, this we need to fix this, fix this, like this how I set up the goal, uh, had incomplete information.
1:33:33But I I literally have a go to market engineer, a contents engineer, a YouTube engineer, and then a support engineer pretty much working daily with with their own respective goals.
1:33:44Um, so in terms of full stream marketing, I mean, it's still early. Right? Like, you know, slash goal is brand new, Cloud Code routines are brand new.
1:33:52It's still early, but it's it's it's just like to to me, I see, uh, like a massive productivity gain. Like, while I was making this thing, I had another agent write my whole newsletter with the instructions and we went back and forth like, how do we simplify this further?
1:34:08How do we let's make the examples even simpler and build up to like more complex examples. Like, you know, it it still takes feedback to do it well, uh, in terms of having multiple streams for marketing.
1:34:21But it's it's like incredible how much tedious stuff it can automate. Um, and I'm just, I'm really there to provide like the high level, this is our strategy, this is where we wanna get to and provide feedback along the way. Like, okay, you were I clearly didn't specify enough information here and I needed to go back and add context so that the AI agent loop has enough information to know when the job is complete.
1:34:44So it's not mill it's not millions of dollars. I don't know what that meant.
1:34:57The it's not millions of dollars, not even close. It's both. She runs Blotato.
1:35:05I was her admin. Who is I'm confused. Who is my admin?
1:35:10Um, okay. Well okay.
1:35:14Hello. Set up an in person two days class helping each person set it all up. Well, that was the point of today.
1:35:20You guys are supposed to follow along today to set it up. So, um, what's a voice linter?
1:35:30You mean, like, for, uh, if you have an AI voice?
1:35:35Is it possible to productize and package something to sell to SMB? Yeah.
1:35:42Of course. Um, that's literally what my product is.
1:35:46So I analyze frame by frame my stock footage and edit. Oh, well, FFmpeg is very limited for editing.
1:35:57Like, FFmpeg is a good utility library and it's it's what a lot of these tools use, by the way. They use FFmpeg.
1:36:05But you can actually tell Claude, like, uh, analyze the video. You can also plug it into Gemini to analyze the video, um, or you can tell it, like, take a screenshot at each three second interval in the video to analyze what's happening in the video, you know.
1:36:24It's hard to see it going on to oh, yeah. I'm not really sharing anything right now. So we're just wrapping up the call until my sushi arrives.
1:36:32Oh, yeah. It's arriving in fifteen minutes. Okay.
1:36:41So to Rachel, I have a video dropping soon that's like 12 ways to make money with AI. And it's it's I really just try to consolidate like multiple different revenue streams and which ones are scalable versus faster income.
1:36:59Um, I think it's gonna drop this week. I hope so. Um, but I would look out for that video and I will probably send a newsletter along with it.
1:37:07But, like, that video is the one where I really try to break down, like, here's what you can do, etcetera. Okay. I think I am going to just send this newsletter now.
1:37:21I mean, I don't I feel like these intermediate level examples are a little easier as well, but that's fine. Okay. I'm just reviewing this.
1:37:30There's no GitHub.
1:37:38If I need tool.
1:37:43Okay. I think this is ready to send. Okay.
1:37:46So, yeah, if you are looking for the companion oh, I should have put a pic if you're looking for the, uh, companion newsletter that has all the prompts and everything, I'm just I'm sending it out right now, so it should be in your inbox in a bit.
1:37:58So just look out for that.
1:38:03Someone is spamming to become my YouTube editor. Okay. That's kind of annoying.
1:38:09To sign up for my newsletter, just go to the link in my bio. So I just go to sabrina.dev.
1:38:19It's not that I don't wanna hire new YouTube editors. It's just I already have YouTube editors. So you would have to be, like, significantly better than my current YouTube editors whom I enjoy working with.
1:38:28You know what I mean? Like, I don't. Besides, like, I'm really trying to focus a lot more on preproduction for the next couple of months and just, like, see how that goes.
1:38:38So okay.
1:38:43So what are you analyzing each frame for, I guess? Can you also analyze a photograph, a technical drawing from an architect? Yeah.
1:38:52You definitely can. So just just drop it into Claudette.
1:38:56We'll analyze it. Well, my school community is for women, so who are building an AI, and you do need a valid LinkedIn profile.
1:39:05This is how we confirm your identity. K? So, um, you can apply to join.
1:39:09It's free. It has 4,500 women in there. We have educational trainings and workshops almost every single day of the month, and then we have an in person retreat literally next week, uh, in Salt Lake, which is where I'm gonna take everybody hiking.
1:39:24So any tips on social media listening skills for agents? How do you write your scripts?
1:39:30Yeah. So for that, I use Apify and then just tell it to like so Apify, find the top 10 videos on this topic.
1:39:41You can put in keywords or hashtags from the past month with the highest number of views and the lowest number of followers. So I have an air table tracking this.
1:39:50I call it like momentum score because like sometimes the creator will just like have a lot of momentum on their on their growth and I wanna see like their entire profile. So, um, that's what I do. So Apify, I'll just type it in TikTok.
1:40:04Apify, find top five videos on this topic or hashtag in past thirty days. And you can have that autonomously running, right, like every single week, have that running and then it will just email you or send you a Slack message with who those creators are and what those videos are.
1:40:29Okay. Any obsolete MacBooks you'd recommend to run locally or to do all of this you taught us today?
1:40:37Yeah. Like, anything you can afford, honestly. Um, there was nothing I showed today that was locally intensive because the Cloud Code routines run-in the cloud, um, unless you're doing, like, launch d to run route.
1:40:52Oh, wait. Hang on. My dog groomer.
1:40:55Yes. He's texting me. Yes.
1:40:58Okay. Uh, Yeah. I mean, you can run routines locally using, like, launch d if you're on Mac, but I didn't cover any of that today.
1:41:06So nothing I showed today is, like, crazy computationally expensive. So you can definitely just buy a used MacBook. Um, in fact, when I buy a MacBook, I usually just buy refurbished because, like, I I have so many already, like, it's fine.
1:41:20It does not need to be brand new. Yeah. Like, I use my 2016 MacBook until, like, 2022.
1:41:26So, yeah, you know, it's it's generally not like a like, you're probably not running into a hardware issue unless you're doing what I'm doing, which is like OBS for agents running in the background, streaming on five platforms.
1:41:45Okay.
1:41:50Yeah. I mean, this is the women's community. Um, I do plan to start, like, an open community, not just for women at some point.
1:41:59It's just the problem is spam. Um, I found that in the at least in the women's community, like, many women are are helpful in cleaning up, like, spam and promotions and stuff. And so it's it's been a, like, a tremendous crowdsourced effort to keep it high quality.
1:42:15And, um, I just worry about that with a with a open community that anyone can join. You know what I mean?
1:42:22Because, like, even for our community, we have screening questions. You have to have a valid LinkedIn profile. That's how we ensure, like, people are real, like, not just, like, deep faking somebody else.
1:42:32Um, where are the details on Salt Lake? Just they're they're in the school community. Just you can ask.
1:42:37You can create a new post in the school community because it's literally happening next week, June twenty fifth to twenty eighth. Okay.
1:42:51I personally prefer pro if you're choosing between MacBook Air versus pro, but it's like whatever you can afford right now, you know. Like, I I I think Air is a little too weak, especially because I do video stuff.
1:43:05Right? So if you can get Pro, then get a Pro. Get a used Pro.
1:43:13What do you do when your agents are running overnight and are well over 30% in their context window? Um, so if you just don't want them to burn through everything, that's where you can set guardrails, like stop after x number of turns.
1:43:31And, uh, you could say stop after x number of turns and produce a summary of what you've done. Right? That way it just doesn't burn through your context window if you need the context window for the morning for your actual work.
1:43:44Are you backing up skills or anything to a hard drive? All of my skills are backed up to GitHub. Right?
1:43:50So, like, every project I work on has a has a GitHub repo. So, like, coming from a technical background, it's like if you if you have, you know, different code bases, you would have a different GitHub project. So I do the same.
1:44:04So, like, I have one for potato marketing, one for my contents empire. I actually made a new one just for YouTube because I'm experimenting so much with like long form editing and it's pretty different from short form editing.
1:44:17So I have a new project for that. Um, I have another one that's just my website and SEO, so bloatato.com and it has like the blogs and stuff there.
1:44:26And then I have other projects obviously for technical things, playgrounds, my bloatato app, my help docs.
1:44:34So I have separate GitHub projects for each, like, area of my company and then each of those GitHub projects is synced to GitHub on the cloud.
1:44:45Each GitHub project has the Claude skills in it, so I back up the Claude skills in the cloud. Uh, so that way I can just, like, change the Claude skill.
1:44:54If I end up not liking it or wanna change it back, I can I can easily do that or I can see what it was before and compare it? Not spamming, but really wanna know how you think consultants should approach cold outreach? Uh, well, always with value first.
1:45:11Um, actually, there's a really good story of, uh, a woman in our AI community who just shared that she basically like cold DMed somebody on LinkedIn and, um, ran a security check on their websites and found some critical stuff, like important stuff, sent it to the person who was cold, and that person was like, oh, like, we need you as a consult.
1:45:34Like, we need you to train our team on AI literacy and do these this other stuff. That is cold outreach, but it's, like, leading with a lot of value. Like, it wasn't just like, I DM'd you, buy this.
1:45:46It was like, you know, she just DM'd like, hey, I see we're like in the same industry or whatever, she said. And then she happened to run the person's website through a security scanner, found a bunch of vulnerabilities, shared it with them, and they were super positive in their response.
1:46:04That's an example of like cold outreach that leads with a ton of value. Um, if you so I blocked the guy who's a video editor, but if you were a video editor trying to cold DM me, the way not to do it is to spam my lives with zero value. I would be really more impressed if you edited this live in real time.
1:46:24Like, you you recorded ten minutes of this live and sent it to me as a link in the chat right now. Like, that would be impressive. And I'd be like, oh, that's kinda cool.
1:46:33He just like did that live right now. Um, like, that would be interesting. That would be leading with value.
1:46:38Um, even if I say no, at least I have like a good impression of you. I can refer you to other people because you actually took the initiative to do that. Lazy d m lazy cold outreach is literally like this.
1:46:49Like, you hop on my live and you just spam 30,000 times, like, the same message that adds zero value, you know?
1:46:57So, um, just don't do that.
1:47:02Uh, so she actually vibe coded, uh, her own tool to run these security checks. So she ran this, uh, prospect's website through her own tool and then and then DM them the results. Yeah.
1:47:16I mean, I you'd be surprised. I read most cold DMs on certain platforms.
1:47:21Like, my Instagram and Facebook are insane because of my DM automations, so I don't read those. TikTok, I don't read because it's just a really messy inbox and I get a billion. But LinkedIn, I don't have that many followers on LinkedIn, so I still read the DMs there.
1:47:35Um, I still read my email inbox, you know, so people cold outreach me all the time. It just it's just like lead with something valuable, like what?
1:47:45Like, you know? So, um, I currently use ManyChat, but I want to replace it.
1:47:51So I'm actually building like ManyChat but with MCP API so that I can just have Claude create my ManyChat automations. Um, and that will actually be part of Potato hopefully, by the end of next month.
1:48:06Yeah. There's definitely a lot of scammers, guys. So, you know, just you like, you literally cannot pay me unless you sign up for my very cheap low ticket product, which by the way, you can automatically get a refund for if you don't like it.
1:48:21Just hit the support button, click refund, you know. Like, don't pay me any other way. There's just don't send me investor money.
1:48:28Don't send me crypto. Like, what does I need to make more announcements about that because people always fall for it.
1:48:37Is it worth it considering older models? Uh, Yeah. If you're gonna be running your own models, sure.
1:48:43Yeah. You you would want I mean, to be honest, if you're gonna be running your own models, get, like, a Mac Studio. That's what I have, um, especially for video related stuff.
1:48:52And then I do want to experiment a lot more with open source. So if you can afford it, that's what I would recommend. But, uh, someone wrote to me as you yesterday on Facebook to join your group.
1:49:05Yeah. There's a lot of scammers on Facebook. Vidya, Katrina.
1:49:11Okay. Oh, hang on. My my sushi is almost here.
1:49:17Uh, if you I just put the link in the chat. So it's school.com.
1:49:22I'll just go there since I'm sharing my screen. Actually, I'm not signed in here. I forgot.
1:49:27So so here it is. Looks like this.
1:49:31I should probably change this logo soon. Put put, like, I don't know, a picture of me or a picture of us there or something.
1:49:39But, yeah, that's a link. I just dropped it in TikTok. You're teaching Claude, but do you favor other services or think them superior?
1:49:47Um, I generally prefer to just go deep on one ecosystem so that I'm productive in it. So I personally have been using Cloud Code for well over a year. Like, I personally publicly announced I was permanently leaving Cursor in favor of Cloud Code in February last year.
1:50:05And by the way, people I got pushback on that. People were like, no way. No way.
1:50:10No way it'll be permanent. Um, you know, it probably won't be permanent. There will be some other tool that comes out that's better.
1:50:15But, like, as of 2026, I find the it's it's the ecosystem and integrations of Cloud Code that, uh, make it just my tool of choice.
1:50:25It's like a for anything I might want to do. Like, certainly, I use specialized tools for other things. Right?
1:50:32Like, for my support stack, I use Intercom as like the front end support widget also for my email marketing. And then I have local open source n a to n literally hosted in the closet over here that runs my entire AI support bot for like $1 a month.
1:50:49Right? Like that, like that's that's just how it's set up. But like Claude is my kind of my Swiss army knife.
1:50:56Like whatever I might wanna try first, I'll probably try it in Claude first. Um, where it sucks is anything visual related. Like, know they just dropped artifact support in Claude code, but it's still like really basic.
1:51:09So if you're doing a lot of image and video generation, for example, in Claude code, it's like hard to visualize. If you're stitching a bunch of video clips together, you have to like open it every single time and it's like kind of annoying. Um, and then only recently did Higgs field come out with an MCP.
1:51:25Um, you can use Replicate and FALL AI APIs as well, plug them into Claude code to generate stuff, and then Cloud can use FFmpeg to stitch everything together. But it's still, like, not an amazing experience. So anything like visual that that requires a lot of visual review is still not a great experience in Cloud.
1:51:43So okay. Yeah. Unfortunately, that particular group is just for women.
1:51:51So because I don't know. Like, the spam problem is really tough.
1:51:55I don't wanna start a community, open it up to everyone and just like lots of spammers join. If you actually look at most free school communities, it's just like total spam.
1:52:06Like, it's like it's like it's really just an email list at this point, so, you know, like no one's actually going into the school and like engaging because it's just way too much spam. So if you do not have enough space to run things locally, what cloud do you recommend for agents?
1:52:28So in the third part of this tutorial, we walked we talked about Cloud Code routines over here.
1:52:37So that's over here. Like, okay. I'm not gonna rehash it.
1:52:41Oh, I was wondering what the poll was so far. Okay. How good.
1:52:45So this this is a poll, by the way, at the bottom of my newsletter. So I'm starting to add these to see, like, is this too hard to follow? So that's interesting.
1:52:5421%, this is interesting but not useful. So it would be nice to know like why but, you know, Substack sucks.
1:53:00So like Substack has like zero automation or like workflow features.
1:53:06It's so, uh, compared to Beehive or ConvertKit if you're coming from that. So, uh, okay. I think my sushi is okay.
1:53:14Few more minutes, then I'll hop off. Let's see. I'll go back to YouTube.
1:53:20Yeah. The newsletter is already sent, so it should be in your inbox.
1:53:24So make sure you check the spam folder in case it's landing there. Um, I'd I'll drop the link as well in the chat.
1:53:33Yeah. Can so Canva significantly improved their MCP connection with Claude.
1:53:38So I dropped some new tutorials on it. There are tips and tricks to make it better. Like for example, if you're using a Canva template, I have Claude first look at the template to check how many words should fit into each like text block.
1:53:54Because if you don't do that, then like Claude will just put too much text or too little text and it ruins the aesthetic of the template. So like that's one little hack to like make it look better. So I think the Canva MCP is actually quite usable today but you have to kind of like set it up in that way.
1:54:11And so I do have like two or three Canva tutorials on my YouTube just walking through like tips and tricks to get the Canva MCP set up properly. Um, because a lot of people will try the Canva MCP and they'll be like, make this and then they'll, it will suck, like for sure it'll suck.
1:54:27So the real way to use it is to use Canva templates and have Claude populate information into those templates. Uh, and then it's gonna look way better, like, way way way better. So okay.
1:54:42Uh, Oh, yeah. Let me I posted the newsletter link here. I pasted it in Substack TikTok and I'll paste it in Instagram as well.
1:54:53How how do you I personally don't use any CRM systems anymore just because it's I have a low ticket SaaS product.
1:55:01Right? So I initially set up GHL for Blotato, but it was like really unnecessary. Also, Intercom has basic email marketing features.
1:55:10So I can technically send email campaigns. I can set up goals hooked up to Stripe to see like, oh, in this email campaign, 50 people converted into paying customers.
1:55:22Right? So it has like just enough email marketing tied to revenue that I don't feel I need to set up a separate system like Salesforce, GHL, or Mailchimp.
1:55:31For my newsletter, I'm just using Substack because it's free. It's like a newsletter plus social network combined, so that's what I use.
1:55:40But okay. If you're not getting so it takes a while to send it through email because I have, like, 230,000 people on my email list, but I just dropped the link here on Instagram as well.
1:55:52I will drop it again on YouTube and I will drop it again in TikTok. Yeah.
1:55:59Can you offer me scholarship under your AI school? Well, the good news is all my AI education is free, so you don't need a scholarship. You could just watch my YouTube videos.
1:56:11Yeah. I mean, I would like to start a community that is not that is open for everyone eventually. It's just it's the the problem is spam.
1:56:19Like, it just you get so full of spam so fast. Um, even in my current school community, like, people always try to promote their stuff in, like, sneaky ways, you know.
1:56:29So, um, I have to, like, set my foot down and be the bad person, but I I don't mind doing that to to protect something worth protecting, you know. Um, for the brain, Notion, Obsidian, Google Sheet or just database, honestly, GitHub.
1:56:44Git like you could your brain is GitHub. So in the Claude code routine when you set it up, you connect it to your GitHub repo and that can serve as your brain. It can store all of your markdown files, uh, pictures even, etcetera.
1:56:59Can I hire you? Um, I am not for hire, but I certainly know a lot of AI consultants. If you really need hands on support, just send me an email, I guess, or honestly, reply to my newsletter with, like, what your project scope is and what your budget, and that's, like, the easiest way I can refer you to somebody.
1:57:17I don't make money off of the referrals, so I only refer somebody if, like, I generally feel they're a good fit. I could be the spam police. That's funny.
1:57:27Yeah. I mean, we we have a lot. We have like 30 admins as our spam police, you know, and they're still spam.
1:57:33Like, people still sneak in stuff. So, um, yeah.
1:57:38I was paying 400 a month for Beehive and I wasn't using most of the features. Um, I also generally do like that Substack has a social platform component. Like, currently Substack's pretty fun.
1:57:50Like it feels like a less toxic version of Twitter. There's a lot of really good smart people on there, especially a lot of free AI contents and education on there that's like middle medium quality in my opinion.
1:58:02So that's good. Um, definitely way higher quality than like TikTok or Instagram. So, um, I highly recommend Substack.
1:58:09Uh, so yeah. Okay.
1:58:12Cool. Um, well, yeah. So my sushi arrived, so thanks everybody for joining.
1:58:17Check your inbox or check this link. I'll post again if you haven't received the email yet.
1:58:22Um, and highly recommend going through this. Um, when you really change your framework into like running loops, it is such a huge productivity unlock.
1:58:33Like you really can be doing multiple, not crazy technical things, but multiple minor things at the same time by like constructing a clear goal, giving it the way to verify its work and then letting it do its thing, as I like to say.
1:58:49So, um, but yeah, thanks everybody for joining. Highly recommend following through the newsletter and I put it at the top. You have full permission to repurpose all of my content.
1:58:58This training in particular I know is worth thousands and thousands and thousands of dollars. So I really hope you just try to go through it, um, because it's it's a it's a next level unlock for those of you who especially if you're already familiar with skills, you're already familiar with connectors and it's just it's it's about like changing your thinking of how you tie this all together.
1:59:18Um, okay. But thanks everybody for joining. So okay.
1:59:21I'm gonna end the stream now. End live video.
The Hook
The bait, then the rug-pull.
Loop engineering is the buzzword everyone is rallying around, and this two-hour livestream cuts through the hype: it is just the formal name for a habit advanced builders already had, taking the human out of the prompt-and-review cycle and handing that job to a second AI. From there it gets practical fast, building a real autonomous agent with Claude Code's /goal command and cloud Routines, no coding required.
Named ideas worth stealing.
22:10list
The Six Parts of a Loop
- Automations (a trigger or schedule starts the loop)
- Worktrees (isolated lane per agent)
- Skills (a reusable playbook the agent follows)
- Connectors (let the agent use Gmail, Slack, Intercom, etc.)
- Sub-agents (one maker, one independent checker)
- Memory (a shared notebook, usually a GitHub repo)
The anatomy of an autonomous loop, the six conceptual pieces that loop engineering bundles together.
Steal forany agent design checklist or onboarding doc for AI automation
48:20model
The /goal Prompt Formula (Cheatsheet 2)
- A clear end state (keep going until ...)
- A check the agent can run to prove it (a count, all tests pass, zero bugs)
- A guardrail (do not delete anything, stop after N turns)
Every good /goal prompt is end state plus verifiable check plus guardrail; it is about defining the task, not wordsmithing the prompt.
Steal forwriting any autonomous-agent task so it stops itself instead of running forever
32:05concept
Maker / Checker Sub-agents
- Maker: the larger, smarter model that does the work
- Checker: the smaller, faster model (default Haiku) that independently verifies the goal
Separate the agent doing the work from the agent grading it, the way a writer and a brutally honest editor produce better writing than either alone.
Steal forany quality-control loop where a single agent would lie that it finished
1:25:08model
Leverage = Skill x Clarity
- Skill: your ability to review the agent's work and improve the loop
- Clarity: your ability to define what you want and what done and good look like
Her personal formula for how much leverage anyone gets from AI; explains why experienced people out-produce beginners with the same tools.
Steal fordeciding where to apply AI and where you first need to build your own competence
How they asked for the click.
VERBAL ASK
1:58:40newsletter
“Highly recommend following through the newsletter; you have full permission to repurpose all of my content. Just go to the link in my bio, sabrina.dev.”
Soft and repeated throughout: the free companion newsletter at sabrina.dev holds every prompt and cheatsheet, framed as thousands of dollars of training given away free. Light, non-pushy, value-first.
MENTIONED ON CAMERA
Visual structure at a glance.
