Superpower ChatGPT: Built in 3 Days, Free for 9 Months — 420K Downloads and 350K Newsletter Subscribers

Dollar figures are paired with rough yen equivalents at 1 USD = 150 JPY.

In the very first week after ChatGPT launched, Saeed Ezzati shipped an extension he’d written in 2-3 days to the Chrome Web Store. Local search and sync for chat history, folders, conversation export, message pinning, prompt management, tone and style controls — a tool that added, from outside, the organizational features ChatGPT lacked at the time. And then, for nine months, he charged nothing at all. Today it has over 420,000 cumulative downloads, over 150,000 weekly active users, five-figure MRR (over $10,000/month, or over ¥1.5M), and a newsletter with 350,000 subscribers. The developer is still just one person.

The numbers, stacked up

Note: this article describes the revenue as “five-figure MRR” because the source only discloses the order of magnitude. The lower bound of over $10,000/month is confirmed, but the exact figure isn’t publicly disclosed. We’re not filling that gap with a guess.

The turning point IS the nine months he chose not to charge

The pivot in this case is delaying monetization for nine months, not a feature addition or going viral. His own words put it plainly.

Free wasn’t charity. It was distribution.

In a brand-new category, nobody yet has a benchmark for “would I pay for a ChatGPT extension?” Put a price tag on it at that stage, and comparison friction appears, halting download growth. He bet instead on driving friction to zero and maximizing user count. Nine months later, with a base of 420,000 downloads and the trust of thousands of reviews behind him, he launched a Pro tier, keeping core value in the free plan while Pro unlocks higher limits and advanced workflows for power users.

There’s a second, unplanned turning point. Inside the product, he’d been sharing AI news and useful tool recommendations. Users liked that so much that a single in-product feature grew into an independent newsletter. That’s “Superpower Daily”, 350,000 subscribers, reaching a segment of people who actually use AI tools, and it’s become a second revenue pillar through sponsorships.

Breaking down the mechanics that worked

Start with an asymmetry in timing. Right after ChatGPT’s release, the pace at which users accumulated frustration outpaced the speed at which the official product added features. He slotted a small product, buildable in 2-3 days, into that window. His statement, “ship before it feels comfortable; your first version doesn’t need to represent your entire vision,” assumes just how narrow that window was.

Add to that watching recurring pain rather than requests. On how he set his roadmap, he says he tracked patterns across reviews, Reddit, Discord, and support emails, looking for “recurring pain” rather than individual feature requests. It’s a system built by layering multiple channels together, designed not to be pulled along by one loud voice.

The acquisition path itself never relied on advertising at all: a Product Hunt launch, Reddit (“especially effective early on”), update announcements on X, discovery within the browser store, a Discord community, and word of mouth. All of these are “places where people who already want to try AI tools are gathered,” and being free directly compounds with shareability. A paid product placed in the same spots would have had a lower viral coefficient.

And running through it all, product and distribution weren’t treated as separate jobs. As he puts it, “product and distribution shouldn’t be separate jobs.” In practice, the newsletter grew out of the product, and the grown newsletter now sends people back to the extension. For a business dependent on someone else’s turf, the Chrome Web Store and ChatGPT itself, a 350,000-person subscriber list is also the one distribution channel he owns outright. What he collected during the free period was more than a user count. It was this “self-owned pipeline.”

The hardest part was that the ground kept shifting

What he describes as the hardest part isn’t competition or monetization. It’s “surviving the platform underneath.”

ChatGPT kept changing. Buttons moved. The DOM structure changed.

An extension works by hooking into someone else’s webpage structure. Every time the official UI updates, features break. He wrote it in plain JavaScript, HTML, and CSS without a framework specifically to keep direct control over that instability, and later invested in refactoring to decouple from dependence on the UI structure. On top of that, Chrome Web Store review queues doubly delayed shipping fixes for broken features.

Looking back, what he says he lacked was internal tooling and automated testing from the start. His reflection: if you’re running a product this fragile solo, a system that automatically detects breakage should come before anything else.

How far does this generalize

What’s replicable is clear. Pick a domain where you’re genuinely a heavy user yourself. Keep the first version small. Delay monetization until proof of value is already established. And build only what shows up repeatedly across multiple complaint channels. None of these require capital or headcount. The free/Pro split (keeping core value free, and pushing paid toward higher limits and advanced workflows) is also a highly generalizable design. It’s precisely because the free version wasn’t a “crippled version” but stood complete on its own that it grew frictionlessly to 420,000 downloads.

What can’t be replicated deserves an equally direct look. The window of ChatGPT’s launch week only opened once. Run the same playbook today in the AI-extension market, and hundreds of competitors are already on the same shelf. Running free for nine months was also possible because he could absorb the operating cost of that period from a position of stability. And the foundation of the business is still held in the hands of OpenAI’s and Google’s decisions, if the official product ships folder functionality natively, part of the core value disappears. The 350,000-subscriber newsletter asset is best read as insurance built in preparation for that day.

Related reading

• Photo AI — a solo developer running an AI product
• Carrd — one person, $2M ARR, no-code