The Promise That Got Me Excited

It was mid-2024 when I first saw Claude’s Projects feature launch. The timing felt like finding a door I didn’t know existed. For anyone who spends time with AI tools, Projects changed something fundamental: instead of starting fresh with context in every conversation, you could now give Claude persistent memory and custom instructions that carried across sessions. It felt like upgrading from a brilliant person with amnesia to a brilliant person who actually remembered you.

I watched the feature demos and started seeing the possibilities immediately. Long-form workflows became smoother. You could iterate on ideas without re-explaining your entire framework each time. By early 2025, Anthropic rolled out even more significant upgrades designed for exactly this kind of work. The more I used it, the more I thought: this is the tool that could finally make an AI side hustle actually work for me.

So I decided to build something. Something that would generate passive income. I was going to create digital products—templates, guides, small resources—and sell them on Etsy. My target was $500 a month. That number felt achievable. Not greedy. Just a reasonable test to see if I could build something that worked at scale.

What I Built (And How the Numbers Actually Looked)

Over three months, I created eleven digital products, each one built using Claude Projects as my primary thinking partner. I’d sketch out concepts in projects, develop content frameworks there, refine them through multiple conversations while Claude held all my previous iterations and feedback in context. It genuinely was faster than my old workflow. The feature worked exactly as advertised for that part.

My products were solid enough: budget templates in Google Sheets format, a freelance pricing guide in PDF, some Notion database templates, a few Canva template packs. Nothing revolutionary, but nothing embarrassing either. The kind of stuff people actually buy. I spent roughly 47 hours total across Claude Projects, Canva design work, and getting everything properly listed on Etsy. I even set up a simple email sequence to promote them. I did the work right, or at least I tried to.

Total revenue after three months: $23. One sale ($7), one other sale ($8), and one purchase of three templates by someone who later asked for a refund, which I gave—but it went to their credit. So effectively, $15 from two actual transactions, and then another person bought two bundle packs I’d created ($8). The math was never in question. I’d earned $0.49 per hour for my time.

Why Everyone’s Talking About This, But Nobody’s Making Money

I wasn’t alone in this. That’s the thing that surprised me most—not that I failed, but that I could see the precise shape of my failure reflected everywhere once I started paying attention. The “AI side hustle” content category pulled in over 1.2 billion YouTube views last year, with passive income tutorials dominating that space. You can’t open YouTube without someone showing you exactly how they made five figures with Claude or ChatGPT.

But then I found the actual data. According to the Influencer Marketing Hub AI side hustle report 2025, 71 percent of people who tried AI-assisted digital product creation—ebooks, templates, courses, the whole ecosystem—earned less than $100 total from their effort. Less than $100. That’s not a failure on the individual level. That’s a category-wide signal.

Etsy itself confirmed what I was seeing in real time. In their Q3 2025 earnings, they reported that AI-generated digital downloads became their most-uploaded product category. But the part that mattered: those products had the lowest average sale price at $3.40, down from $7.20 in 2023. The market is flooded. Everyone and their cousin is using Claude to make Etsy templates now. Supply is infinite. Demand is not.

The Thing I Actually Learned (Spoiler: It’s Not About the Tools)

What started bothering me wasn’t the $23. It was that Claude’s Projects feature was genuinely helpful for the creative work. That’s what made this hard to shake. The tool delivered exactly what it promised. I could maintain context across long sessions. I could refine ideas without re-explaining. The feature itself was not the problem. The problem was something entirely different, and I had to sit with that for a while before I could articulate it.

I was solving a technical problem when the real problem was a market problem. I had optimized my workflow to build products nobody needed at a price point where they might as well give them away. Claude Projects made my process smoother, but it didn’t make my products more valuable. It didn’t create customers. It didn’t answer the core question: why would someone buy my budget template instead of using a free template from a thousand other sources, or just making their own in thirty minutes?

I notice something now when I read the successful AI side hustle posts. They’re never actually about the tool. They’re always about distribution, or expertise, or an existing audience, or solving a specific niche problem that most people don’t know they have. The tool—whether it’s Claude or any other AI—is almost incidental. It’s like the difference between having great camera equipment and being a great photographer. The equipment matters less than people think.

What This Actually Opened Up For Me

Here’s what I’m sitting with now: I didn’t waste $23 and 47 hours. That’s one way to look at it, and it’s technically true. But I also got three months of real data about what doesn’t work, and I got to test a new tool in a way that actually meant something because I was trying to solve a real problem, not just play around.

Claude’s Projects feature is legitimately useful. But it’s useful for the same thing it’s always been useful for—thinking better, clearer, faster. It’s not a money-printing machine. It’s not a substitute for actual market research or customer discovery. It’s a tool that helps you execute on ideas, but it can’t generate the ideas that matter in the first place.

If you’re thinking about building something with Claude or any AI tool, go do it. But go in with your eyes open about what the tool actually solves and what it doesn’t. Don’t expect passive income from templates. Do expect to learn something real about yourself as a builder. The $23 is gone, but the experiment is still teaching me things every time I think about it.

What did you notice when you tried building something with AI tools? I’m genuinely curious whether your experience looked anything like mine, or whether you found a path through that I missed. Drop a note in the comments or shoot me a message—I want to hear what actually worked when it worked, and what you think I’m missing.