A Mac Mini sitting on a desk made $19,522 in 90 days.
Nobody touched it.
Three times a week at 6am, the pipeline triggers. Script. Voice. Video. Upload. Forty minutes later, three videos are scheduled.
No face. No editor. No team. Cost per video: $1.50.
n8n orchestrates every workflow. Google Trends finds topics. Claude writes scripts and SEO. ElevenLabs voices it. Pictory renders video. YouTube API uploads on schedule.
One Mac Mini. Four workflows. Pays for itself in four months. Costs $10/month after that.
Nine videos a week. $8,000/month from ads alone. Sponsorships from month three. Affiliate links passive in the background.
Two channels: $16,000/month.
Three: $24,000/month.
Copy the workflows. Change the voice ID. One hour of work.
I'm on pace to hit over 1,000,000 impressions on X this year.
All of my best performing posts took 3+ months to write.
The physical act of writing didn't take 3+ months.
Creating new knowledge to write about did.
The big spikes you see in my analytics chart below are from when I wrote about new knowledge that I created.
I spent a few months building agents for GTM, then gave away all the sauce on how to build your own, for free.
Building agents is a new field. You can't learn it from LLM training data. You can only learn it by doing it.
Here's the thing. AI killed repackaging existing knowledge. ChatGPT can do that for anyone.
Writing posts about knowledge people can get from AI doesn't work.
I've gained the most impressions from creating new knowledge (about AI) and sharing it with people.
The act of writing my top posts was done by AI in just a few minutes. I provided the codebase I created, and creative direction.
The creative direction took me 15 years to learn.
So basically when you read my posts, you're getting a million years of knowledge in a few minutes ;)
(ht Christopher Lochhead for writing about the concept of knowledge creation vs packaging yesterday)
$23,048 landed in 5 days on a YouTube channel that has never once shown a face
The face belongs to a 17 year old who dropped out of school at 15, nobody believed him, and this is his smallest channel, he runs 12 on YouTube
6 silver play buttons sit on a garage floor, propped against a black car still wet from rain. The kid who earned them is upstairs, face down on the bed, asleep
The browser is still open behind him, estimated YouTube revenue $23,048.59, the line climbing while he sleeps
The strategy is almost insultingly simple, he rips trending Roblox clips off TikTok, an AI recuts them, recaptions them, and reposts them as YouTube Shorts to a US audience that never saw the original, 12 channels, on a schedule, no face, no camera
Here's the part nobody runs the math on, YouTube Shorts pay about $0.04 per 1,000 views, so $23,048 is more than half a billion views in 5 days. One kid can't watch that much, he doesn't have to
The TikTok creators made the videos, he built the YouTube machine that resells them
He was asleep when the $23,048 came in
Facing defeat in the Machine War, the humans launched high-altitude bombers to disperse nanoparticles into the upper atmosphere, blanketing the planet in a permanent layer of black clouds
I never understood the 'billions required' AI narrative. A student built a $15 offline ChatGPT that works when the cloud doesn't. Distributed > centralized, always.
Even Gen Z aren't safe from AI anymore lol
This video is completely AI and I bet if I showed it to majority of AI guys even they wouldn't be able to tell
Best part is this video took me 10 minutes to make and edit + only cost me $0.30
@0xkerazcity $3/month to run what exactly? A quantized 7B model? The second you need reasoning depth or multi-step agents you're back on API calls. Like saying you saved on gas by walking to a different city
How to use Claude effectively and earn $10,000 a month?
If you’re using AI for business, there are many details you need to understand. You also need to know how to write prompts correctly and communicate effectively.
If your company is constantly hitting its limits, it’s likely that your staff don’t know how to communicate properly with the agent.
There are repositories that completely change the approach
→ Gstack is a free repository that adds 23 ready-made skills to Claude Code.
→ According to many, this saves around 5 hours every day.
What’s inside?
/office-hours - idea validation before development begins
/plan-ceo-review - finding the best product version
/plan-eng-review - architecture, databases, APIs and security
/design-shotgun - 4–6 design options at once
/qa - automated testing via browser
/ship - pre-release check
/document-release -automatic documentation updates
→ Its best feature is /office-hours
→ It asks the tough questions and helps you understand whether you should be pursuing what you’re creating.
If you run a company, your staff often simply don’t know how to interact with AI. Because of this, the desired results are often not achieved.
If a company knows how to interact with AI, it will achieve the desired results. The goal and the task are important for everyone carrying out the assignment.
BUSINESSES PAY $1,200/MONTH FOR RECEPTIONISTS WHO CAN BE REPLACED BY AN $8/MONTH LOCAL HERMES AGENT
a local insurance agency owner in california went from $40,000 to $120,000 a month in revenue by replacing manual customer calls with a custom on-premise receptionist
you deploy the stack using a local mac mini, configure the routing skills, and secure a $5,000 setup fee plus recurring monthly maintenance retainers
the setup is straightforward:
- configure claude or any local models to output structured skill rules for greeting callers and scheduling bookings
- link the local engine to a free booking service for automated calendar updates
- record a screen capture of a live call to build instant proof for outreach
- deploy the node under the client's desk to handle customer inquiries 24/7
this allows local agencies to cut overhead costs to zero while developers build recurring cash flow
full breakdown on how to package the service and close B2B clients ↓
The new ChatGPT Image gen is sooo friggin good! 🤯
Today should be the LAST day you use standard icon-sets for your marketing and websites.
Just crafted 10 *insanely* good on-brand icons for a new AI community we're launching this month (shameless plug: https://t.co/7FNyCyQKDv).
Here's the how-to (full prompt below):
- Go to ChatGPT
- Upload your existing logo
- Explain or list the categories you need icons for
- Press Submit
Full prompt:
I need to create icons for each of these categories to replace the current emojis. Do not feel the need to copy the same symbol or shape of an emoji. Design an icon image that you think is absolutely best for each category. Use the logo supplied simply as a brand style reference guide. Create each file individually please with a white background, or, within a light cream circle surrounded by a white background. whatever makes the most sense to make these amazing!
been consuming a lot of @matt_gray_ content lately.
one thing he talked about recently that stuck with me - get clear on your big three before the weekend hits.
i commonly mistake this to be my to do list, my inbox; but instead it's three things that actually move my life forward over the next two weeks.
here's mine this week. have a great weekend!
Only a tiny fraction of people are talking about this
She says the easiest business she would rebuild from zero is kids content
Just simple toddler videos that get replayed over and over
She finds a big channel, studies the format and asks ChatGPT to help shape the idea into a video prompt
Then she drops that into an AI image and animation workflow and builds the whole thing on mobile
Kids content is one of the strongest loops on YouTube because parents do not search for something once and move on
They replay the same video again and again
That means one good clip can keep earning attention long after it is posted
While everyone else chases saturated niches and weaker monetization, she picked the audience that does not even know what a skip button is