TikTok | Learn | Advertise | Unsubscribe
Hey {{first name | guys}},
This week, I will be showing you a few mini AI exercises I experimented with!
But first, here’s an interesting article about why AI agent startups are becoming the “New Solo Founder Playbook”.
The Rise of the One Person Company
According to Carta’s Solo Founder’s Report, over one third of new companies are now started by just ONE person.
Starting a business used to mean finding a technical co-founder, hiring a small team, or spending months learning everything yourself.
That is beginning to change.
Forbes calls AI-agent startups the new solo-founder playbook because agents can now handle huge chunks of research, product building, marketing, support, and admin. They do not just answer questions. They can complete multi-step jobs while you act as the manager.
Apparently, a complete solo founder AI agent stack only costs $3,000 to $12,000 per year.
Compare this to the equivalent of a human team (junior engineer, marketer, designer, customer support) may cost ~ $960,000 to $1.4 million per year.
A study of more than 160,000 Product Hunt launches found solo entrepreneurship jumped after ChatGPT, but teams still dominated the highest-ranked products.
At the same time, 97% of solo launches were “one shot” products, meaning they didn’t get updated within 12 months. This means AI is making it easier to launch, BUT not automatically easier to succeed.
The real opportunity is not building a “company” full of 20 AI employees.
It is being able to stay lean for longer. Use AI to test the idea, build the first version, create the sales material, and handle repetitive work. Once it’s proven the company can make real money, hire humans to get scale it.
The new order:
Before: Idea → hire people → build → test demand
Now: Idea → use AI → build and test → hire when you find the real bottleneck
Some interesting stats:
Almost 50% of the recent increase in one-person business applications came from sectors with high AI adoption, including tech, finance, and professional services
BUT Two-person teams still win at fundraising. They represented 36% of startups that successfully raised funding and 40% of funded SaaS startups
At least four million Americans used ChatGPT to help start or operate a business in March 2026
Anthropic CEO Dario Amodei only named 3 industries he thought one person billion dollar companies would be most likely occur: Proprietary trading, developer tools, and businesses with highly automated customer service
I’m personally excited for the one person businesses (I am one myself) and I believe more people these days are building side hustles after their 9-5!
If you’re interested but don’t know where to start, Greg Isenberg has a gold mine of start up ideas here.
Speaking of one person businesses, the buzz on this new tool by Jack Dorsey (Twitter founder) may help power more solopreneurs!
New Slack/OpenClaw Killer?
What’s all the Buzz about?
Everyone is talking about this new open source “workspace” for humans and AI agents.
Think Slack but the ability to add Codex, Claude, Gemini, Hermes, OpenClaw into the one place and talk to them.
Quite interesting as the UI/UX looks very easy to navigate
Here’s why Buzz sounds like the magic pill (see my explainer video):
Bring your Claude OAuth subscription - You heard right! This is the first I’ve heard where you can do this. You can’t do it in OpenClaw or Hermes (only through API), so take advantage of it while you can!
Not being tied to 1 model - It feels like we are watching a ping pong match each week. One week ChatGPT is great, next week it’s Claude. If you’re running a business you can’t afford to be switching every time. At the same time you want to test out new features to keep ahead. Buzz allows you to talk to different models and swap the models behind each agent with 1 click
Being able to use multiple models together in one “office”- Whether as a team or in a workflow (ie. Claude negotiates deals, Perplexity populates with real time data, Codex puts together the company analysis)
Set up was super easy - I was most shocked when I clicked one button and Chatgpt + Claude were already in
Mobile access + desktop app is beautiful
Git integration - Anything to make it less technical is a tick for me
BUT does it actually work like that in real life?
I’ve only done initial testing and out of the box (other than the super easy set up), the output was kind of underwhelming. For example, don’t expect it to have great persistent memory like Hermes.
It is still early days which Buzz founder Jack Dorsey does admit but it definitely has legs so I will continue to test it out.
5 Minute AI Tutorials
A few practical AI tutorials and resources I’ve made for you this week:
Instead of setting up an automation in ManyChat everytime you have a new keyword (SO tedious to this day with a thousand clicks). Just get ChatGPT Codex to do it!
You don’t need to hire an automation specialist either, you can get it to set up your Zapier, Make, N8n automations too!
shared.image.missing_image
shared.image.missing_image
Your AI has your entire history as evidence. Why not use it to uncover 22 patterns, habits and blind spots you didn’t know about yourself and ways you can fix it.
Grab the Copy/Paste Prompt (use it in Claude, ChatGPT, Gemini or your preferred AI)
Create “Throw Away Websites”
Lately what I’ve found myself doing more is rather than reading big blobs of text from AI, I will turn them it into an interactive website. Yes, I know the concept isn’t new, making a website is “easy” with AI.
But how I’m thinking about it is more like scribbling diagrams, working things out on a sticky note. Once I’m done, I throw it out.
It has honestly been a game changer especially if you are highly visual like me. It can help you quickly:
Understand complex concepts: For example, imagining what your AI agents are actually building feels like it’s a black box.
I went back and forth for weeks trying to figure out what they were doing.. it was really confusing and felt like we were going around in circles.
Until I asked it to make me a site to explain what it was building (in ChatGPT but you can use Claude too). Here’s my explainer video.
You can see a dummy site here.
If you want to make one too, here’s a prompt for you to try:
Using @sites, create a clear, visual map of my AI agents and the system I’m building based on the information I provide. Show each agent’s role, the tools they use, how information moves between them, where human approval is needed, and any gaps or overlaps you notice. Turn it into a simple, interactive webpage that a non-technical person can easily explore and understand, and ask me questions first if important details are missing.
Prepare for consulting meeting: I had a 45min meeting with an accountant who puts out a tonne of content on youtube.
I wanted to ask specific questions that WEREN’T answered in his videos to get the most value from the call so I popped these into AI:Links to his youtube videos and asked AI to package all his strategies up into a website that was interactive with toggles I could adjust it + how he would answer my questions.
Below was an example of what it made custom to what he would recommend but I haven’t shown the exact strategy out of respect (note: numbers aren’t real!).
***Update: Ok before you go to town on those “throw away sites” I may have underestimated how many tokens it takes to make these. Maybe for now, they aren’t so cheap yet 😭
That’s it for this week! 😊
Did you Enjoy this Week's Newsletter?
Cheers,
Nat
