I Almost Built Another Chrome Extension. Then I Looked at the Numbers.

At 1:40 in the morning I was staring at a Chrome extension called Reddit Comment Exporter. The free plan stopped at 20 comments. The paid plans were $4.99 and $12.99 a month. The interface was tidy, the problem was easy to understand, and for a moment the conclusion felt obvious: this is the kind of small product I should build.

There is a reassuring rule in independent software: when you do not know what to make, find someone who is doing well and study the pattern. It is better than inventing a market in your head. But there is a trap hidden inside that rule. A polished pricing page proves that someone can publish a pricing page. It does not prove that customers are paying.

So before opening the code editor, I spent a day trying to disprove the idea.

The question was not “Can I build this?”

Of course I could build a comment exporter. With AI, so could thousands of other developers. The more useful question was: after building it, would I own anything scarce?

I looked at 20 Chrome extensions across four rough groups: established specialist tools, general AI assistants, free extensions used as lead generators, and small products that looked easy to imitate. This was a purposive sample, not a scientific census. Chrome Web Store user counts are rounded, they do not reveal active users, and they definitely do not reveal paying customers. The sample cannot tell us that extensions are dead. It can tell us whether writing code right now is a sensible next move.

The distribution was brutal

At the top were products such as Sider with 5,000,000 users, Monica with 3,000,000, and Merlin with 900,000. These are not weekend extensions. They are funded products with teams, models, distribution, and years of accumulated surface area.

The durable specialist tools were more interesting. Keywords Everywhere showed 1,000,000 users. Text Blaze showed 700,000. Bardeen showed 200,000. They looked like browser tools, but the extension was only the doorway. What customers were really buying was data, saved history, integrations, automation, collaboration, or a workflow that had become expensive to replace.

Then came the imitation-shaped products: similar feature, similar store description, little evidence of distribution. In one small cluster the visible user counts were 3, 7, 21, 30, and 6,000. The median was 21. Four of the five had fewer than 100 users.

The Chrome extension was not the business. In the winners, it was the last inch between an existing business and the user’s current tab.

What the successful products had that I did not

• A proprietary or costly data source.
• A repeated workflow tied to a profession, not a vague productivity wish.
• History, saved state, or team collaboration that made switching inconvenient.
• Distribution outside the Chrome Web Store: an audience, an SEO engine, a larger SaaS product, or an existing customer base.
• A reason to keep paying after the first successful export.

My proposed Reddit exporter had none of those. It would save a few clicks, perhaps once or twice. A developer could reproduce it. A non-developer could ask an AI tool to help. Reddit could change its page. A free competitor could appear. Even if people installed it, the path from install to recurring revenue was mostly hope.

The strange part is that stopping felt like failure. Nothing had been shipped. There was no launch screenshot and no tiny line of revenue to celebrate. But this may have been the cheapest successful product decision I made that week: one day of research prevented several weeks of polishing a commodity.

The narrower experiment I would still run

I have not crossed browser extensions off the list. I have crossed off generic extraction tools. I would reconsider only if I found one profession that repeatedly moves, checks, or reformats information between a specific B2B portal and a spreadsheet, CRM, accounting system, or case file.

The next test would last three days and contain no production code:

1. Talk to five people who perform the same browser workflow.
2. Confirm that at least three repeat it every week.
3. Show a manual or clickable version and get two people to ask to install it.
4. Ask for a price. Continue only if one person accepts an explicit amount.

If those conditions are not met, the track closes. No logo, no subscription screen, no week spent arranging pixels. The evidence threshold is deliberately small, but it is real.

Why publish a product that was never built?

Because research should not disappear just because the product did. I am trying to build a small independent studio with AI, and most of the honest story is not a sequence of clever launches. It is a pile of plausible ideas that did not survive contact with evidence.

This is the first entry in Roads Not Shipped: notes on the things I nearly built, what stopped me, and which small bet remained after the excitement wore off. If one of them eventually works, the earlier wrong turns will be part of the result—not an embarrassing preface to delete.

Sources checked

• Reddit Comment Exporter — Chrome Web Store
• Sider — Chrome Web Store
• Keywords Everywhere — Chrome Web Store
• Text Blaze — Chrome Web Store
• Bardeen — Chrome Web Store