- Startup Spells 🪄
- Posts
- The AI Agents Powering John Rush’s 26 Startups
The AI Agents Powering John Rush’s 26 Startups
PLUS: 1,000,000 US-based Facebook followers for $1,000!?
The AI Agents Powering John Rush’s 26 Startups
Over the last 3 years, John Rush has launched 26 startups, most of them AI-powered, and serves nearly a million users without relying on VC funding or a traditional team.
His secret is a system of highly specific, no-interface AI agents that handle everything from SEO content to outbound sales.
These agents are not general-purpose tools or assistants. Each one replaces an entire workflow, sometimes even a full-time role, without requiring complex configuration or human oversight. The result is a stack of single-purpose, autonomous systems that let him move fast, validate ideas, and stay lean.
Listing Bot
Purpose: Automated product distribution across all relevant platforms where user acquisition can happen—forums, aggregators, directories, and any place where submissions are allowed.
Functionality:
Identifies every site where a product should be listed
Navigates through submission forms using browser control
Automatically registers the product
Ensures presence across platforms with zero manual outreach
“Listing Bot is an example of an agent that finds all places on the internet where your product can be relevant, and it lists you there... anywhere where it’s allowed to submit a product.”
This replaces hours (or days) of repetitive outreach. Instead of hiring a VA or manually filling forms, Listing Bot handles everything through browser automation—submitting on behalf of the founder, quietly scaling presence.
Listing on popular sites like Reddit, LinkedIn, and directories is how Google finds you especially if you don't install any Google products like Google Analytics or submit your site to Google Search Console for indexing purposes.
SEO Bot
Purpose: Full automation of long-form SEO content creation, indexing, and publishing—triggered by a single input.
Functionality:
Accepts a single URL
Uses AI to determine what topics are worth writing about
Generates blog articles in bulk
Publishes them directly to a CMS
Requires no human involvement after activation
“It does everything for you. It creates all the articles, sends them to my CMS. I just never touched it again.”
There is no interface, no content dashboard, and no editing step. SEO Bot turns content into a one-button operation, bypassing typical writer workflows completely. It’s designed for founders who want growth without recurring effort.
AI Scraper
Purpose: Autonomous data extraction to populate directories or build datasets from public websites—without writing scrapers or hiring contractors.
Functionality:
Takes a list of URLs
Visits each site and analyzes the content
Extracts structured fields like price, features, pros/cons
Compiles the data into CSV for direct CMS import
“AI can look at the websites and find the information it has to find... and creates a CSV file that I bring back to my CMS.”
This removes the need to contract scraping jobs or manage scraping logic. Instead of dealing with breakable XPath code, John points the agent at URLs and gets back enriched product data, organized for public consumption.
Tiny Ads
Purpose: Monetization agent that handles sponsor discovery, outreach, and ad placement—especially for small or niche products that aren’t yet big enough for inbound advertiser interest.
Functionality:
Understands the theme or vertical of the product
Finds businesses interested in reaching that audience
Extracts contact details from the web
Reaches out via cold email or social media
Lets sponsors place ads directly on the platform
Operates entirely without founder interaction
“The system finds them and lets them place their ads on your platform, and you just get the money.”
Tiny Ads converts a directory’s passive audience into revenue automatically. It doesn’t wait for ad inquiries—it finds relevant advertisers, initiates the deal, and closes it, all while the founder works on something else.
Purpose: Turns weekly memory fragments into polished, well-researched AI news content, ready for social sharing.
Functionality:
Begins with a list of AI events John remembers from the week
Researches each bullet point to gather exact details, stats, and context
Returns an enriched summary
Final write-up is manually rewritten for tone and virality
“I make a list of everything I remember from the last seven days. Then I take that list and I give it to my tool called Social Bot... I ask it to bring details because I don’t want to go and, you know, find exact numbers.”
This hybrid approach allows him to keep the human voice while automating the labor-intensive parts of research. It’s not designed to write the final draft—but it dramatically cuts the time it takes to go from rough idea to polished post.
A System That Replaces Work, Not Just Tasks
What makes John’s agents unique is their level of autonomy. They don’t just assist—they operate independently. Most don’t even have a user interface. His end goal is to reduce his workload to the point where he only handles creativity and strategic input. Everything else, from content to monetization to outreach, is managed by a system.
He doesn’t build tools with dashboards or feature sets. He builds tools with “just one button”—and in many cases, no buttons at all. These agents function like micro-employees: invisible, tireless, and designed to solve a single pain point from end to end.
Credits to Angel Poon for the insight.
Top Tweets of the day
1/
The problem with Bluesky (and Mastodon as well) is they mostly have one kind of asshole. They operate unchallenged, leading to extreme asshole saturation.
Meanwhile, X has *all* kinds of assholes working against each other, so they kinda cancel each other out.
— Andreas Kling (@awesomekling)
9:56 AM • Jun 4, 2025
Phenomenal product insight. Bluesky and Truth Social have a similar eco-chamber.
2/
So many enterprise SaaS companies are f*cked.
They are lazy, fat, and drunk on years of quasi-monopolized access to multi-year, 7-figure SaaS contracts, pre-AI.
What they previously considered a moat is now a mirage & the racket is coming to an end.
As they complacently
— Alex Lieberman (@businessbarista)
3:22 PM • Jun 4, 2025
For 2 decades, enterprise SaaS companies could get away with charging lots of money. See Salesforce and Slack's pricing. Now they have to innovate due to competition.
AI has reduced the time it takes to make a product in 1 month to making it in 1 day. Countless SaaS were never attempted because they took 1 or 2 years to get off the ground. Now with a cracked team, you can build a similarly scoped product in <3-6 months.
3/
Products with extensive/rich UIs lots of sliders, switches, menus, with no scripting support, and built on opaque, custom, binary formats are ngmi in the era of heavy human+AI collaboration.
If an LLM can't read the underlying representations and manipulate them and all of the
— Andrej Karpathy (@karpathy)
8:02 PM • Jun 4, 2025
The future of all apps is text-based so that AI can automate it.
Website builders like Webflow basically store their layout values in JSON/XML/YAML. This helps once you can make the AI translate values to JSON/XML/YAML.
Such products will lost longer than alternatives. Figma has its data in such text format whereas Adobe probably doesn't. Any product that has stored its data in binary product have to rewrite it un.
Rabbit Holes
How Starbucks Lost $100 Million in Australia by Paddle Studios
ClaudeLog - Experiments, insights & mechanics about Claude & Claude Code by InventorBlack
What’d ya think of today’s newsletter? Hit ‘reply’ and let me know.
Do me a favor and share it in your company's Slack #marketing channel.
First time? Subscribe.
Follow me on X.
More Startup Spells 🪄
Typeform’s One-Field Onboarding UX: Conversion Lessons from Gas, Snapchat, Duolingo, and Spotify To Boost Signup (LINK)
Riverside FM's Dwell Time Using Free Transcription Service (LINK)
Unordinary Objects To Win Social Media Game (LINK)
Chess.com SEO Case Study: Scaling to 30M+ Monthly Visitors and $150M in Revenue (LINK)
Reply