I’ve spent years making money online with affiliate marketing. Now I’m seeing what happens when you throw AI into the mix. I test tools, marketing strategies, traffic sources, and AI business ideas — and share the useful stuff here. Consider this my public experiment.

Here's a simple way to use ChatGPT to find out why people aren't buying your product. And you don't need to be a marketing expert. Let's say 100 people visit your sales page, but nobody buys. Instead of guessing what's wrong, try this. Step 1: Copy your sales page into ChatGPT. Step 2: Give it this prompt: "Act like a customer who is interested in this product but isn't sure about buying it. Tell me 5 reasons you might leave this page without buying. Point out anything that's confusing, missing, or hard to believe. Then suggest simple ways to fix each problem." Step 3: Look at the feedback and pick the 2 biggest issues. Step 4: Make changes and test the new page with real visitors. I've spent years in affiliate marketing, and one thing I've learned is that small problems on a sales page can cost you money. ChatGPT won't know for sure why people aren't buying. But it can help you find things worth testing before you spend more money on ads.

Most people use Claude like it's just another chatbot. Big mistake. It can analyze PDFs, build interactive apps, create visuals, and automate parts of your workflow. Here are 8 powerful things you can do with Claude that you might be missing. Bookmark this for later 👇

Canva just made an interesting move in the AI image race. Instead of only focusing on making prettier AI pictures, the company launched an image model called Craft. It's trained using designs made by humans. And I think that tells us something important. People don't always need a beautiful picture. Sometimes they need a useful design. A flyer for their business. A social media graphic. An ad for a new product. A presentation for a client. Canva is trying to make AI better at creating those kinds of things. Now, we'll have to see how well it works in practice. But I like the direction. The best AI products aren't always the ones with the most impressive demos. Sometimes they're the ones that make an everyday job easier.

I think AI is going to make fake experts easier to spot. Anyone can use ChatGPT to write a post that sounds smart. They can share 10 business tips. Explain how to make money online. Or create a guide about something they've never done. And the writing might look great. But ask them one question about what actually happened when they tried it. That's where things get interesting. I've spent years running online ads, and I can tell you that real experience usually comes with mistakes, failed tests, and lessons you won't find in a basic tutorial. AI can help you explain what you know. But it can't give you experience you never had. I think real stories are going to become more valuable as AI content fills everyone's feed.

I think this is where AI starts getting really interesting. Claude Duet is designed to keep working even when you're not using it. Imagine waking up and finding out your AI assistant already checked your emails, found important updates, and prepared your daily tasks. That's a lot different from opening ChatGPT and typing another prompt. For someone running an online business, this could save hours every week. Of course, I'd still want to approve anything involving money or customers. But the idea of AI doing useful work while I'm away from my computer? That's something I'd definitely want to try.

You can only keep ONE AI superpower. A) Turn any idea into a working app B) Research any topic in minutes C) Create a week's worth of content in an hour D) Automate your repetitive work The other 3 are gone forever. Which one are you choosing?

Here's a simple ChatGPT trick that could save you hours of research. Let's say you want to start a business selling AI tools to real estate agents. But you have no idea what problems they deal with. Here's what I'd do. Step 1: Go to Reddit and find 10 posts where real estate agents talk about problems at work. Step 2: Copy those posts into ChatGPT. Step 3: Use this prompt: "Read these 10 posts. Find the 5 biggest problems people keep talking about. Explain each problem in simple words. Then show me how AI might help solve it. Don't make up problems that aren't in the posts." Step 4: Pick one problem and talk to 5 real estate agents about it. Ask how much time it wastes. Ask what they do to fix it now. And ask if they'd pay for a better way. That's it. You don't need to spend months building an AI tool before finding out if people want it. I've learned from affiliate marketing that a good idea means very little until someone is willing to pay for it.

650 experiments in just 2 days. That's what caught my eye here. Karpathy used AI to help test hundreds of ideas without doing every step by hand. Now think about what this could mean for a small business. You could use AI to help create different ads. Try new sales page ideas. Compare offers. Or find problems with your website. Of course, you still need real customers to know what works. But here's what I find exciting. A small business owner can now do some things that once took a whole team. And I think we're just getting started. x.com/karpathy/status/203077…

Shopify's CEO made a pretty bold move. Before asking to hire more people, teams have to show why AI can't do the work. Think about that. For years, when companies got too busy, they hired more workers. Now they're asking a different question. Can AI handle some of this first? I think we're going to see a lot more companies doing this. And that creates a big chance for people who know how to set up AI tools for businesses. You don't need to build the next ChatGPT. You just need to help a business save time on work it already does. x.com/tobi/status/1909251946…

You have 30 days to make your first $1,000 using AI. You can only choose ONE tool. A) ChatGPT B) Claude C) Canva D) Cursor Which one are you betting on? And what would you build?

Shopify's CEO introduced a pretty unusual rule for hiring. Before asking for more employees, teams had to demonstrate why AI couldn't handle the work. In April 2025, CEO Tobi Lütke shared an internal memo explaining that using AI effectively was becoming a basic expectation at Shopify. One part stood out. Teams requesting additional resources were expected to explain what the work would look like if AI agents were already part of the team. That's a big change from how most companies traditionally operate. Usually, when a department gets overwhelmed, the first question is whether they need more employees. Shopify wanted teams to consider AI before making that decision. Now, that doesn't mean AI can replace every employee. And it certainly doesn't mean every business should adopt the same policy. But I think it tells us something about where companies are heading. Business owners are starting to evaluate work differently. Not just asking how many people a task requires. But asking which parts genuinely need human judgment. If I were starting an AI automation business today, I'd be studying the tasks companies are still hiring people to perform. Some of those job descriptions might reveal opportunities worth investigating.

AI makes mistakes. I think the next big opportunity in AI might be fixing those mistakes. Everyone is focused on generating more content, writing more code, and automating more tasks. But who's checking the work? AI can write a convincing article with incorrect facts. Generate code with security problems. Summarize a contract while missing an important clause. Or produce a financial report with the wrong numbers. And as businesses use AI for more important work, those mistakes become expensive. I wouldn't be surprised if we see an entire category of businesses built around AI quality control. Testing AI-generated code. Verifying research. Checking automated workflows. Reviewing AI customer service conversations. The more companies depend on AI, the more valuable it becomes to know when AI gets something wrong. Sometimes the opportunity isn't building another AI tool. It's making sure the existing ones actually work.

People are starting to realize they don't always need to buy another subscription. Sometimes they can just build the tool themselves. Need a simple expense tracker? Build one. Need a dashboard that combines data from different platforms? Build one. Need a tool that organizes leads in a specific way? Build one. Obviously, building reliable software still takes testing, maintenance, and some technical judgment. But the economics are changing. For years, businesses had to choose between buying software that almost fit their needs or paying a developer to build something custom. AI is starting to create a third option. And I think that's going to put pressure on a lot of smaller software companies. x.com/karpathy/status/188619…

You can automate ONE part of your work forever. 1. Writing emails 2. Creating content 3. Researching information 4. Organizing your daily tasks AI handles that task for you. Which one are you choosing?

I think this is where AI starts getting really useful for business owners. Most people don't need another chatbot that tells them how to do something. They need one that actually does the work. Think about running an online business. Checking campaign results. Updating spreadsheets. Finding new leads. Preparing reports. Following up on unfinished tasks. That's hours of work every week. If AI agents like Grok Bot can handle those jobs reliably, that's a pretty big deal. Of course, I'd still want a human checking anything involving money or important decisions. But I'd much rather spend my day growing a business than clicking through 20 different websites.

Here's a simple way to turn one customer question into 10 useful pieces of content with ChatGPT. Let's say you sell an AI tool that helps small businesses manage customer inquiries. A prospect asks: "Will this work with my existing CRM?" Instead of answering that question once and forgetting about it, paste it into ChatGPT. Use this prompt: "Turn this customer question into 10 content ideas for X. Include: 3 common mistakes people make. 2 step-by-step tutorials. 2 comparisons. 2 frequently asked questions. 1 opinion that challenges a common assumption. Keep every idea focused on helping someone understand the problem. Don't invent product features or customer results." Now you've got a list of potential posts based on something a real customer actually wanted to know. I'd pick the best 3 and write those first. Then watch which ones generate questions, saves, and profile visits. Coming from affiliate marketing, I've learned that customers often tell you exactly what content you should create. You just have to pay attention to what they're asking.

This is the kind of AI business idea I'd pay attention to. A startup built an AI assistant to help tradespeople prepare customer quotes. Not exactly the kind of product that's going to dominate tech headlines. But think about the problem. A plumber finishes a job. An electrician visits a potential customer. A contractor needs to prepare an estimate. Every hour spent on paperwork is an hour they aren't doing paid work. Tradestack reportedly built its assistant in just 6 weeks. And the interesting part is how specific the problem is. They're not trying to create an AI assistant for everyone. They're helping a particular group of workers complete a particular task. I think there are still plenty of opportunities like this hiding in industries most AI entrepreneurs never think about. x.com/LangChainAI/status/183…

Most people are trying to write the perfect prompt. But what if the real opportunity is building a process that doesn't need you to keep prompting it? Think about a business that receives 100 customer inquiries every day. Instead of asking ChatGPT to answer each one, you could build a workflow that: Reads the inquiry. Classifies the problem. Drafts an appropriate response. Flags anything complicated for human review. Tracks the outcome. That's a completely different way of using AI. And for entrepreneurs, I think learning how to design these repeatable processes could be more valuable than memorizing another 50 prompts. x.com/rohanpaul_ai/status/20…

Someone gives you $100/month to spend on AI tools. But there's a catch. You can only subscribe to ONE. A) ChatGPT B) Claude C) Midjourney D) Cursor Which one gets your $100 and why?

Klarna thought AI could handle most of its customer service. Then the company had to rethink its approach. In 2024, Klarna announced that its AI assistant was handling about two-thirds of customer service chats. The company said the assistant was doing work equivalent to hundreds of full-time agents. Pretty impressive numbers. But the story didn't end there. In 2025, Klarna acknowledged that focusing too heavily on reducing costs had affected service quality. The company began emphasizing human support again. I think there's an important business lesson here. Just because AI can automate a task doesn't mean customers will always prefer the automated version. Sometimes people want a quick answer from a chatbot. Other times, they want to explain a complicated problem to a real person. If I were building an AI customer service business, I'd pay close attention to that distinction. The goal shouldn't be to eliminate humans at any cost. It should be to make the entire customer experience better.