A two-person breakdown of OpenAI's top-tier model that skips the game demos and goes straight to code audits, nine money-making agent prompts, and a Raspberry Pi speaker built and shipped in about 30 minutes.
GPT-6 Astra costs more per token than any other OpenAI model, but it finishes tasks in fewer steps, and the real payoff is auditing live code and turning a hardware idea into a working prototype in under an hour.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
A developer or founder who already has a live app and wants two free prompts, a performance audit and a security audit, that a top-tier model can run in one sitting.
Someone with an idea that felt too big for last year's AI, whether it is a company-wide agent audit or a hardware side project.
A builder curious what a coding agent can actually do with computer use, browsing, and a 3D tool like Blender.
SKIP IF…
You want game demos, 3D toy renders, or a straight spec sheet comparison against other models.
You are looking for a plain feature list instead of applied use cases.
TL;DR
The full version, fast.
GPT-6 Astra is OpenAI's most expensive model, priced near the top of the market, but it needs fewer steps per task so the cost stays reasonable under a Codex subscription. The hosts skip the viral 3D demos and test it on real work: a performance audit that cut one app's response time from 800 milliseconds to 20 to 30 milliseconds, a security audit that caught a real vulnerability in a live payments app, and nine agent prompts for bill negotiation, competitor tracking, and QA. The clearest proof is a hardware side project: guest Ras Mic asked Astra to design a Raspberry Pi speaker and went from idea to ordered parts to a merged pull request in about 30 minutes. Their conclusion: ask the model for the projects that felt too ambitious a year ago.
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.
Greg sets up the episode: skip the viral 3D toys and demos, focus on real use cases for making money and getting more productive with GPT-6 Astra.
01:53 – 04:14
02 · Astra Overview
Astra sits at OpenAI's top price, on par with the priciest model on the market, but it needs fewer steps per task so the value holds up. Ras recommends a performance audit and a security audit for any live app.
04:14 – 11:48
03 · 9 Astra Prompts
Greg shares the list of nine Astra prompts he posted publicly, later reposted by Greg Brockman: a bill renegotiator, a service-to-software converter, a marketplace deal watcher, a one-person operator dashboard, an agent-opportunity audit, a browser operator, a nightly QA agent, a competitor spy, and a lead-generating browser game.
11:48 – 16:21
04 · Jarvis Speaker Idea
Ras pitches Astra his dream of an AI speaker built into a HomePod-style shell. Since jailbreaking a real HomePod was not possible, he had Astra design the whole build from scratch in Blender.
16:21 – 18:29
05 · Think Bigger with Astra
In about 30 minutes Astra generated a parts list, Amazon links totaling $561 before tax, a wiring diagram, supplier names for the shell, and the code for a pull request Ras merged into his agent repo.
18:29 – 21:16
06 · Vibe Coding to Vibe Manufacturing
Ras turns a spare PS4 into a Linux learning project with Astra's help, and the hosts compare the moment to 2024's vibe-coding wave (Lovable, Bolt, Replit, v0), this time reaching physical hardware.
21:16 – 22:52
07 · Closing Thoughts
Ras estimates his agent burned through more than $24,000 of model usage in a week on two $200 subscriptions. Both hosts frame it as a subsidized window worth using before prices catch up.
Atomic Insights
Lines worth screenshotting.
GPT-6 Astra costs as much as OpenAI's top-tier pricing, but it finishes tasks in fewer steps, so the cost per finished task is not as extreme as the sticker price.
A single performance audit prompt cut one live app's response time from 800 milliseconds to 20 to 30 milliseconds.
A security audit on a live payments app surfaced a real vulnerability before it caused damage.
Astra designs better UI when it is handed a reference image instead of an open-ended prompt.
Pointed at a government website, Astra's computer-use agent found pages a Google search missed.
A guest asked Astra to design, price, and order a Raspberry Pi speaker build, then wrote and merged the integration code, all in about 30 minutes.
The parts order for a DIY AI speaker prototype came to about $561 Canadian before tax.
One guest's AI agent tallied over $24,000 in a week of model usage against two $200 monthly subscriptions, a sign of how subsidized current AI pricing is.
The gap holding people back is not model intelligence, it is bravery: asking for bigger projects than another landing page.
The same shift that turned everyone into a vibe coder in 2024 (Lovable, Bolt, Replit, v0) is starting to reach physical hardware.
Takeaway
Ask The Model For Bigger Projects
AGENT PLAYBOOK
The ceiling on what an AI agent can build is set by how big a request you're willing to make, not by the model's intelligence.
02Astra Overview
Run a full performance audit prompt on any live app before assuming the code is fine; one audit cut response times from 800 milliseconds to 20 to 30 milliseconds.
Run a security audit on any app real users touch, even a side project, since it can catch vulnerabilities before they turn into an incident.
Give a coding agent a reference image instead of an open text prompt when you want a specific UI look; the results land closer to what you pictured.
039 Astra Prompts
Point an agent at your own bills, subscriptions, and provider chat support and have it negotiate or cancel what you no longer use.
Turn a service business into a software product by reverse engineering the workflow into steps, then pricing a simplified version at $500 to $5,000 a month.
Before hiring anyone, prompt an agent to audit which of that role's tasks it could already handle, and estimate the time and cost saved.
04Jarvis Speaker Idea
When a product you like has weak built-in AI, like a smart speaker with poor voice assistance, treat that gap as your opening instead of a dead end.
If the official route is closed, like jailbreaking hardware, ask the agent to design an equivalent build from parts instead of dropping the idea.
05Think Bigger with Astra
Let an agent turn a vague hardware idea into a concrete parts list, budget, and purchase links before you talk yourself out of trying.
A 3D modeling tool like Blender can double as a way for an agent to explain a physical wiring layout you would not otherwise follow.
Once the hardware works, ask the same agent to write and open the pull request for the software that connects to it.
06Vibe Coding to Vibe Manufacturing
Old hardware sitting unused, like a spare game console, is a low-risk way to practice giving an agent an open-ended physical project.
The models that make it trivial to launch a landing page are being used for far bigger, more ambitious builds by people willing to ask for more.
Expect the shift that made everyone a software builder in 2024 to repeat for physical products, favoring small, specific niches over mass-market ones.
Glossary
Terms worth knowing.
Vibe coding
Building software by describing what you want in plain language and letting an AI agent write the code, rather than writing it by hand.
Vibe manufacturing
Applying the vibe coding approach to physical products: describing a hardware idea and letting an AI agent handle the design, parts list, and wiring.
Computer use
An AI agent's ability to control a computer directly, clicking through websites and filling forms the way a person would, instead of only answering text prompts.
Blender
Free 3D modeling software used here by the AI agent to lay out and visualize the physical speaker build before any parts were bought.
Codex
OpenAI's coding agent product, referenced here as the subscription that gives access to GPT-6 Astra and covers most of the model's usage cost.
Resources
Things they pointed at.
12:51toolBlender
13:16productRaspberry Pi
13:45linkAmazon.ca
15:10toolGitHub pull request workflow
17:33toolOmachi (Linux distro mentioned for the PS4 project)
Quotables
Lines you could clip.
05:56
“The saddest thing that people can do with a model like GPT-6 Astra is generate landing pages.”
contrarian one-liner that reframes the whole video→ IG reel cold open↗ Tweet quote
05:40
“Intelligence has gone up, but the bravery for people to try new things hasn't gone up.”
sharp thesis statement, quotable on its own→ newsletter pull-quote↗ Tweet quote
17:33
“I could run Linux on my dust collecting PS4. I would have never imagined, thought, or even tried to do this.”
concrete before/after moment about ambition→ TikTok hook↗ Tweet quote
21:35
“I would cancel my Anthropic subscription. I would cancel Claude. Astra's that good.”
“In the last week, I've spent over $24,000. Mind you, I just have a couple of $200 subscriptions.”
shocking cost stat that proves the subsidy point→ IG reel cold open↗ Tweet quote
The Script
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.
17px
metaphoranalogy
GPT -6 Astra is here and I got good news for you. By the end of this episode, you're going to understand what are the different startup ideas? What are the different use cases?
How do you use it to be more productive? How do you use it to save money, to make money? Everything all in here.
I'm going to show you how you can use it with hardware to create hardware businesses. And this is an episode not just focused around cool 3D demos. for people who want to make money to be more productive with AI and how you can use it specifically with Asha.
What are the different prompts? How to think about it? I brought on my friend Ross Mike, who's just amazing at clearly explaining all this sorts of stuff.
So enjoy the episode, like, comment, and subscribe for more of this in your feed. And I can't wait to see you at the end.
Ross Mike on the pod. This isn't your your average GP6 Astro review, is it? No, we don't have no games.
We don't have any trade demos. We don't have there's some weird ones, right? Some weird 3D ones.
Although those things are cool, they're not really useful to myself, you and Greg. So we're actually going to talk about some useful, cool stuff. And particularly, we're going to get you guys to dream big with hardware.
yeah exactly so by the end of this episode what I hope you take away is getting your creative juices flowing around really some ideas around how you can build a business using GPT -6 Astra and actually how you can have a more productive code base, have better products. So by the end of this, you're not going to get some glitzy, cool video game thing, but you're actually going to have, okay, here's what I can actually create today.
Yep, that's what we're going to do. And we'll just get started. So a couple of things about the model.
First and foremost, this is OpenAI's most expensive model. It's on par with Fable in terms of pricing. But if you have a Kodak subscription, very subsidized.
And not only that, it's also very efficient, meaning it doesn't take as many steps to complete a task. So even though, yes, the price is high, you're still getting a good bang for intelligence. If you have an existing app or web app, mobile app, something that you've worked on, software, I highly encourage you to do two things.
Tell the agent, tell your Codex agent, I want to run a performance update. My pages aren't as fast as they need to be. My API calls aren't as fast as they need to be.
Please review my entire app and suggest a performance update. I did that for one of my apps and speeds went from 800 milliseconds to literally 20 to 30 milliseconds. Right.
That's an insane increase in speed. Highly suggest you do that. Number two, something we couldn't do with Fable is a security audit.
I was telling Greg this. I have some sort of payments startup, if you will, in the background for a specific niche. And I had a couple of people using it.
I ran the security audit and. Astra was basically like, my boy, you would have been cooked if I didn't read all this. So run a security audit if you have an app that people are actually using.
In terms of UI, I mean, I could just show some examples. This is one that it designed. Really nice.
The reason it was nice was because I gave it a reference image. And the reference image, if I have it, is the old Microsoft XP network icon. I really like the OG network icon.
Very swaggy. For anyone over 30, you remember this. I gave it that.
I'm like, use this as a reference image. And yeah, it looks pretty nice. Without a reference image, it kind of does too much, right?
So do what you will with that. Number four, computer use. The agent Codex Astra, insane with using.
your computer insane with doing tasks for you in the background, using your local machine, highly encourage you to do that. But what Greg wanted me to talk about. Wait, before you get into this, before you get into this, I want to share some things of my own, of my own that I think you should use Astra for.
So let me share my screen real quick. Please. So this, I tweeted this, Greg Brockman retweeted it and he said it was good.
It's good enough for Greg Brockman. It's good enough for all of us. So I just wrote, here's some nine cool Astra prompts we're trying.
One is the bill renegotiator. Go through my internet, phone, software bills, jump into each provider's chat support and negotiate them down or cancel what I'm not using. I don't pay for a lot of stuff, I'll tell you that right now.
I'm obsessed with how do you turn agencies into software right now. So here, pick one service business in a niche. You've got to pick the niche and reverse engineer the exact workflow they sell to clients.
Break it into steps, tools used, inputs, outputs, human judgment points, and places where the work gets slow or expensive. Then design the simplest AI product that could replace the first version of that service and charge $500 to $5 ,000 a month. Now, is it going to be perfect?
No, but it's going to get your creative juices flowing around what the software could look like. And I want to add to that. I think one thing that I do a lot that a lot of people should do is I prototype the heck out of everything, Greg.
Let's say I'm building an app and I'm using Daytona, their sandbox provider. I will literally, before I jump into Daytona, I'll spend 15 minutes trying to build a clone. It's not that I'm going to use the clone, but I'm going to allow building the clone to just get me to think.
Because I think one thing that I've realized is intelligence has gone up, but... The bravery for people to try new things hasn't gone up, right? Like I said, like I said this the other in a video, the saddest thing that people can do with a model like GPT -6 Astra is generate landing pages.
Literally. It's just like you can use GPT -4 or whatever we had, like we're past that, like really think bigger. So I like that idea.
Number two is fire. Number three, I almost called it the Gary Vaynerchuk. prompt, but I refrained.
Watch Facebook Marketplace and Craigslist in my city for cameras, furniture, bikes, what have you. Listed way under market and text me the second one's mispriced with the link. This is a really good idea for people who just want to make a few hundred dollars a day or something.
Maybe they can't get a job. A lot of our young people can't get jobs right now. I'm going to screenshot that and put it in my church group chat because it's true.
Jobs for the young kids is really, really hard right now. Yeah, it's really, really hard. So that's a good one.
And also for a lot of people, it's fun to do this sort of thing. So young or old, this is an interesting one. Number four, if you are a founder.
This is a no -brainer. Basically, it's creating the one -person company dashboard. Look at my docs, notes, Stripe exports, analytics, customer calls, and project lists, and build a weekly operator dashboard.
I want to know what is making money, what's wasting time, what customers are asking for, what I should stop doing, and the three highest leverage actions for the next week. Be blunt and show your work. Prove.
You should set that up for your project. I was literally just going to say that.
Okay, we got a few more and then we'll get to your hardware thing using Blender, which I think is just fascinating. I want you to show me that. Number five, audit my company for agent opportunities.
Look at how this business works and find the tasks we should give to agents before hiring another person. This is just how, by the way, side note, this is how I'm thinking. Before I hire anyone, I'm like, should I just get an agent?
to do this? Sometimes yes, sometimes no, but that's been the default of how I've been thinking as I'm building my company. For each task, estimate the current human time, the cost of mistakes, the tools involved, the difficulty of automating it, and the first safe version we could deploy.
Prioritize things that save money or create revenue within 30 days. And then we have, I'll do... And we're almost done.
I'll do them all. Be my browser operator. So you can actually use Astra to do this, right?
Use the browser to complete this workflow. You give it the workflow. As you go, click through the actual sites, collect the data, fill the forms where appropriate, and keep notes on what broke or slowed down.
When you're done, give me the output, the repeatable SOP, and the automation plan so this can become an agent. And it will do this well. I'm so bullish on computer use.
And a lot of people might think it's insane or it doesn't work or it breaks. You'll be surprised how good Astra is. If anyone's in Canada, then you know how terrible the CRA website is, how atrocious that monstrosity is.
If the government's watching, I'm just joking. But I literally got Astra with Codex to comb through. Tons of pages to find me the exact thing I want that I couldn't find on Google search, right?
So it really is that good. 100%. Number seven, you can literally build a QA team.
So every night, open my app on a real phone, go through signup, checkout, and the main flows, and screenshot anything that's broken or confusing. So keyword confusing too. Like when I say QA, I don't necessarily mean there's like...
Of course, it could be like a bug, but it can also be like a UX bug. People are just confused what happens here. It's too much information.
Competitor spy. Sign up to my top three competitors, sit inside their product and their emails, and send me a monthly report on every new feature, price change, and the thing they do better than I do. That is a good one.
I think what's interesting is there's two types of founders. Either the founders who are obsessed with the competition and are spending too much time looking at competition. And then the founder who ignores it completely and doesn't know what they're doing at all.
And I think what's good about this is it's sort of like the happy medium. It's just monthly. Every month I want to know what I need to know and that's it.
No, that makes sense. And again, it's competition for a reason, right? Like there's sort of this element of like, oh, you know, looking around and stuff.
But obviously, innovation always wins. And this one, yes, it is make a video. The last one is make a video game.
But it's making a video game with the purpose of getting you email signups and phone numbers and data and helping you sell whatever you're selling. Build a browser game around this mechanic. You give it a mechanic.
Don't just make a cute demo. Add progression, tension, scoring, failure, polish, and one reason someone would send it to a friend. That's extremely key because you want to create something that's a viral game.
I saw a lot of games that are really cool, but people wouldn't actually send them to their friends. Then add lead capture, like email and SMS, and it needs to tie into my core product, which sells XYZ. yeah that's actually pretty good like a like a really like something cute and fun that i'd want to share because to your point right like people are making these first person shooters which is great it shows the model capability but i have call of duty at home exactly all right i showed you mine you show me yours yes so um Please don't make fun of me, people watching.
But I've had this idea of a hardware startup for a while. Obviously, we've seen the rise of personal agents. I have one in the works that I plan on launching soon that I think is a different take and pretty cool.
But the one aspect that the reason why I wanted a personal agent and like this is, again, like a dream of mine as a child, like the whole Jarvis thing is I wanted there to be physical hardware. Right. And it's such a miss for companies like, you know.
Amazon and like Apple, like they have like these beautiful products, but like their agent or the AI sucks. In particular, I'm talking about the HomePod. If anyone has the HomePod, Siri is terrible, but this is the best looking speaker, in my opinion.
And I just got to think, I'm like, what if I could put my agent on like a HomePod? So first, my thought process is buy the HomePod and jailbreak it. I did a little bit of research, not possible to do it.
I'm like, what if I built it, right? So I literally, told the agent and i'll zoom in so we can see the convo i asked i told the agent i have blender install can you use it it's i know that it can use it i just wanted it to set up a connection it did that and i asked it to create a home pod like speaker and it did and i asked him i want to build an ai home speaker i want to load my agent ruth from and i gave it my repo how would i diy and then it basically said that I would need a Raspberry Pi speaker.
I could load in my agent and it was possible. Now this got me excited because I didn't think this was possible. I didn't think that I had the capability to do it.
And it basically created me a diagram. And it said, it even went as far as to say, this is what I would need to buy to have a first physical prototype. Now I got excited, right?
And it even gave me a budget. It's like 350 to 450 before tax. I live in Canada.
So this is like 700 bucks, 600 bucks. I was like, cool, give me amazon .ca links, right? And it got me the links.
And it was $5 .61 before tax. I was like, I just bought. I just automatically bought.
But then I panicked. I'm like, what do I do? I don't know nothing about hardware.
So I asked the agent, Greg, can you use Blender to show me the setup of this project? And will you believe it? It showed exactly how I would have each thing laid up, right?
This is the Raspberry Pi. This is me plugging it in. This is some sort of speak.
This is the speaker. This is some sort of connector. I don't even know what this is, but this helps.
And I think this is the voice recognizer. But it basically laid out on Blender how I would have this set up. Now I've got to the point where I'm like, oh, I can build this.
And then it was telling me once we've proven the proof of concept, then this is what we would need to build out the HomePod. uh version and then it even in a separate thread i have emails there so i don't want to share but in a separate thread it was even giving telling me uh chinese suppliers to accept to reach out to to build the home pod shell so i went from idea to setup i bought all the products right i told it i bought all the products and then i was like okay like i bought this stuff but i have my agent repo what code would i need to add or what changes would i need to make to set this up guess what it did greg It wrote the code and it made a PR.
I literally had code. It made a PR where like once I have the thing set up, I can easily load my agent. Mind you, this is a couple of minutes conversation, but I dared to think a little more bigger out of my box.
So it gave me the PR. It explained to me the code. And then I asked for the links again because I didn't purchase at that time.
I asked for the PR. I can click and I can show you. I merged the PR.
It made a PR. So I know like once I have my once I have all the hardware stuff ready, I'm going to build it and I'm going to have it set up. And I even asked it just to be sure.
Do I get one each of one one of each, by the way? And it said yes. So we went from idea from exploration, Greg, to set up to initial prototype.
Cost you 500 bucks. It gave me it showed me exactly how to set everything up. Right.
And if I have further questions, I'll continue to work with the agent. It also wrote the code that I'm going to load into the Raspberry Pi to make my speaker agent work. This was half an hour work.
This, in my opinion, is the true power of Astra, where like we have models like Fable, like, OK, not Fable, but we have like 5 .6 Sol and all these different smaller models that are good at a lot of things. What makes Astra amazing? is for these insane hardware projects, this next level software.
So again, if you're using Astra to build a landing page, it's a sad use of intelligence. But I kind of wanted to use this moment to encourage you all, if you've had that hardware idea, if you've had that next level software idea, if you want another, I don't know if I have the tab here. I have a PS4 lying around.
And I was wondering, can I boot Linux on a PS4? and run agents on it. And I literally said, I have PS4 based on what you know about me.
I don't plan on gaming with it. What can I do? That was the first thing I asked.
And it's like, you could turn it to a media server, this, this, that. But then it said, turn it into a Linux learning project. I was like, oh, I didn't know you could do that.
So I said, low -key nuking it and running Linux would be cool. And then it says, yeah, a PS4 you can SSH into and use as a Linux lab would be a fun project. And then people have been talking about Omachi, which is a type of Linux distro.
And I was like, could I run it on PS4? Yes. I could run Linux on my dust collecting PS4.
I would have never imagined, thought, or even tried to do this. But with this model, I'm starting to think way outside the box. And mind you, you might think, oh, you're a developer, so this is easier.
This is networking OS stuff. i'm more of a web dev mobile dev type of development this is this is like a little above like my pay grade and but like the last couple of months i'm doing stuff like this i have a home lab i'm buying i bought a couple dj exports like i am thinking way above what my normal means would have been if i didn't have access to this intelligence and there's no reason you shouldn't i'll stop yapping greg so you can say some things as well i remember when to 2024 when Lovable and Bolt and Replit and V0 came out, all of a sudden anyone could be a vibe coder.
And we started seeing all these vibe coding projects, websites, landing pages, apps. And the effects of that are still playing out to this day. It feels like we're having a similar shift right now, but for physical projects.
So what you showed me... with your hardware product, you know, that would have required you become a full -time 3D designer or slash engineer. I would have never known how to do any of this.
And to be honest, I wouldn't have tried because it's intimidating. And like, I think, you know, most humans, like I myself, like I like knowing the stuff that I'm doing. And if I don't know like about hardware, for example, it would feel intimidating for me to even try.
Now... I have a friend for $200 a month or $50 a month or $20 a month that can actually walk me through all of this. And it's super possible.
I even have a desk coming up that's going to be right here. And it's going to be my hardware station. And I'm going to just start building random stuff.
And let's say nothing turns to a product. The enjoyment of learning and developing, and maybe it leads to a job opportunity, a business, whatever it is, you owe it to yourself to do this stuff. Yeah, I mean, make a printable object, a CAD file, a blender model, a circuit board, like something.
Because I think that there's opportunity in terms of niche products that you might be like, oh, I don't know how big of a market there is for this sort of thing. it doesn't need to be huge. And if it doesn't cost a lot to create, there's a business here.
I actually personally believe that there's going to be thousands of $1 million a year plus businesses that get created in this whole shift that OpenAI has spearheaded with Astra around vibe manufacturing. Will the other models catch up? Absolutely.
But I think that it's fair to say that Astra was the first model that ushered in this era. And I think this is just really, like for me, where my brain is going is like, okay, what are some small little products that seem niche but solve a real pain point that I can go and create in a short amount of time and then create meta ads to drive traffic to those products.
That's what's going through my mind. yeah and when i build my home pod i'm gonna hit up greg to teach me all that stuff so make sure you're subscribed because that's all the stuff he teaches here too i appreciate you mickey thanks for coming on astra is awesome um and i'm curious to see how the other models uh take note of it sure anthropic is sweat for sure for the first time i would like if i if someone asked me pick a subscription I would cancel my Anthropix subscription.
I would cancel Claude. Astra's that good. So please take advantage of the subsidies before they eventually start charging us the real price of these things.
Yeah, because the value, I mean, think about it. Like the value is obviously more than what we're paying now. Yep, yep.
Like I asked my Astra agent to calculate how much I've been using, like the subscription price. but also like in terms of API price, how much like I've been using in terms of model costs. And in the last week, I've spent over $24 ,000.
Mind you, I just have a couple of $200 subscriptions, right? So there's an insane level of subsidization that's happening. Take advantage of it now.
Dream big, act big, and build big. I appreciate you. I'll include links where you can follow Ross Mike as he's known on the internet.
youtube x and all those places go go and subscribe go and check them out thanks again for coming on i think this is probably your what is it six seventh eighth yeah i'm a regular this on this pod and as always i appreciate you greg it really means a lot for you bringing me on and thank you everybody for always watching and showing love as well see you later have a good one
The Hook
The bait, then the rug-pull.
Every other GPT-6 Astra video is a 3D game demo. This one is two builders testing whether the model can actually make or save money, and by the 20 minute mark a hardware idea has become an ordered parts list, a Blender wiring diagram, and a merged pull request.
Frameworks
Named ideas worth stealing.
04:14list
Nine Astra Prompts
The bill renegotiator: go through internet, phone, and software bills and negotiate or cancel them
Service-to-software converter: reverse-engineer a niche service business into a $500-5,000/month product
Marketplace watcher: text me when Facebook Marketplace or Craigslist has a mispriced item
One-person company dashboard: read docs, Stripe exports, and calls, then name the top 3 actions
Agent-opportunity audit: find tasks to hand to agents before hiring another person
Browser operator: complete a defined workflow and hand back an SOP and automation plan
Nightly QA agent: open the app on a real phone and screenshot anything broken or confusing
Competitor spy: sign up for the top 3 competitors and send a monthly report on what changed
Lead-capture browser game: build a game tied to the core product that collects email and phone
The list Greg posted publicly, later reposted by Greg Brockman: nine ways to point an agentic model at money-making or time-saving tasks instead of another demo.
Steal forany solo operator or small team looking for concrete agent tasks instead of another landing page
CTA Breakdown
How they asked for the click.
VERBAL ASK
22:32next-video
“youtube x and all those places go go and subscribe go and check them out”
Greg plugs Ras Mic's YouTube and X at the very end, right before signing off, rather than mid-roll.
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
A solo walkthrough of five open-source GitHub repos getting traction right now, an AI writing editor, an agent-run CRM, a video-editing agent, a skill security scanner, and a phone-controlling harness, each with the exact install command and the first small workflow to try.
Greg Isenberg and developer Vinny break down WebMCP, the experimental browser feature that lets any AI agent search, compare, and buy on a website without scraping the page, then price two startup ideas built on it.
Cloudflare quietly built the payment rail for AI agents to pay tiny fees for web resources, and that rail is about to spawn a whole generation of businesses selling clean data, agent-readiness, and packaged expertise.
Greg Isenberg gets a live, screen-shared tour of Jack Dorsey's new agent-native chat app from an early user — and presses him on whether it actually beats Slack.
Cody Schneider maps the exact infrastructure — pipeline, warehouse, agent — behind a Facebook ads system that researches, creates, publishes, and kills its own losing ads.
Nick Vasilescu spins up an AI co-founder running on Grok 4.5 and, in one live session, takes it from idea to landing page, thumbnail, and cold-email sequence.
00:25
00:56
01:07
01:25
01:50
02:19
02:37
02:54
03:12
03:38
03:47
03:59
04:40
04:57
05:15
05:32
05:50
06:07
06:24
06:42
06:59
07:17
07:34
07:52
08:09
08:27
08:44
09:02
09:19
09:37
09:54
10:11
10:29
10:46
11:04
11:21
11:34
12:13
12:26
13:04
13:21
13:38
14:04
14:34
14:41
15:04
15:21
15:36
15:55
16:06
16:48
17:06
17:25
17:43
18:01
18:26
18:54
19:10
19:27
19:44
20:00
20:17
20:34
20:50
21:08
21:34
22:01
22:12
22:23
24:43
28:58
34:10
38:44
Video of the Day37:47
Video of the Day56:00