Building a free Chrome extension in 3 days and turning it into a 5-figure-MRR ecosystem

Thanks, I really appreciate this.

I think you framed the free early access point really well: free because you need real usage data is very different from free because you are unsure whether the product has value. The first one is intentional. The second can become a trap.

On the platform risk side, I'd definitely think about redundancy early if Ridgewell depends on an undocumented API. In my case, the first version worked because I could move fast, but the cost was that every ChatGPT UI change could break something. Over time, I had to decouple as much as possible, support multiple versions of ChatGPT experiments, and keep a close eye on product changes.

The distribution loop is also something I'd try to build from day one. It does not need to be complicated, but the product should naturally create reasons to talk to users, learn from them, and bring them back.

What's the task you keep putting off the longest?

Thanks! The biggest technical bottleneck was definitely keeping up with ChatGPT changes.

I expected normal extension complexity, but I underestimated how much work it would be to build on top of a brand new product I did not control. ChatGPT's UI, DOM structure, buttons, routes, and flows kept changing, and every change could break something small in the extension.

The second bottleneck was internal tooling. Early on, I was moving fast and solving user-facing problems, which was the right tradeoff at the time. But as usage grew, I needed better tools for debugging, support, release notes, feature flags, and testing.

If I were starting again, I would invest earlier in a more robust abstraction layer around platform-specific UI changes and better automated browser-level tests for core workflows.

Absolutely. Platform risk was obvious from day one. Search, folders, export, pins, better history management, those are all things OpenAI could reasonably build natively.

The way I thought about it was: if the feature is painful enough today and the platform is growing fast enough, it can still be worth building even if it might become native later. Users needed those workflows immediately, not someday. So the early roadmap was less about pretending we had a permanent moat around every feature, and more about moving quickly, listening closely, and stacking useful workflows faster than the platform could prioritize them.

For monetization, that made us cautious. We kept a generous free version because we wanted trust and distribution first, especially without a big audience or community. Paid features had to be things power users cared about enough to support, not just basic fixes that might disappear into the native product.

Over time, the bet becomes less "OpenAI will never build folders" and more "can we stay closer to a specific group of power users than the platform itself?" That means better organization, bulk workflows, search, export, customization, and all the small edge cases that heavy users feel every day.

What's the task you keep putting off the longest?

I'd probably still start where the user already feels the pain.

For Superpower, that meant a browser extension because the problem was inside ChatGPT itself: organizing chats, searching history, reusing prompts, exporting work, etc. A standalone app would have added friction too early because users would have had to leave the place where the workflow was happening.

If I were starting today, I'd still use the extension as the fastest way to get close to the real workflow and learn from daily usage. But I'd also build the standalone/web side earlier than I did, because platform risk is real. The extension is great for distribution and workflow proximity, but the standalone product gives you more control, more room to build deeper features, and less dependence on another product's UI.

So my answer would be: start with the surface closest to the pain, but don't let that surface become the whole business forever.

What's the task you keep putting off the longest?

Exactly. I think that was a big advantage here. I wasn't trying to predict a market from the outside; I was fixing something that annoyed me every day. When you're also the user, the small workflow problems are much easier to notice.

What's the task you keep putting off the longest?

Thanks so much. I really appreciate that. Glad you liked both the product and the story :)

What's the task you keep putting off the longest?

Thanks, this is a great question.

For Superpower, the signal was not one single metric. It was a mix of repeated usage, user pull, and the kind of feedback I was getting. People were not just installing it once; they were using it as part of their daily ChatGPT workflow, asking for more advanced features, joining Discord, leaving reviews, and sending very specific requests like folders, export, prompt history, search, and higher limits.

That told me two things: the product had become useful enough to be part of their workflow, and some users were clearly power users who would pay for fewer limits and more control.

I also tried to introduce Pro in a way that did not punish existing users. The free version still had real value. Pro was more about higher limits and advanced workflows, not taking away the core reason people trusted the product.

For a product that requires sensitive business data, I would probably be more careful with "free-first." I still believe in reducing friction, but the trust path would matter more than the price. I would focus on proving security, showing exactly what data is accessed and why, offering read-only or sample-data modes, letting users get value before connecting everything, and being very clear about privacy and controls.

So I still like free as a trust/distribution strategy, but only if the product can earn trust gradually. If the product asks for sensitive access on day one, I'd design the onboarding around trust first, not just price.

What's the task you keep putting off the longest?

That was one of the harder decisions.

My rule was that the free plan should still feel like the real product, not a demo. If someone installed Superpower, they should be able to understand the value, use it in their normal ChatGPT workflow, and trust it before ever paying.

So I tried to keep the core workflow free: the features that make ChatGPT easier to organize, search, and use day to day. Pro was mostly for heavier usage, higher limits, and more advanced power-user workflows.

The line I tried to draw was: free should help a casual or regular user get real value, while Pro should make sense for people who use ChatGPT heavily enough that Superpower becomes part of their daily work.

I also paid attention to what users were asking for repeatedly. When people started asking for higher limits, bulk actions, advanced exports, more prompt workflows, etc., that was a good sign those features could belong in Pro without making the free product feel broken.

What's the task you keep putting off the longest?

Thanks, I really appreciate that.

That distinction took me a while to fully understand too. Free worked because it gave people a low-friction way to try the product, build trust, and give feedback before there was any pressure to pay.

But I think the important part is that free still needs a reason. In my case, the reason was distribution, feedback, and trust in a very new category. Without that, free can easily become just "no business model."

What's the task you keep putting off the longest?

Thank you, I'm glad that stood out. It was one of the biggest lessons for me while building Superpower.

Yeah, platform risk was probably the most stressful part of the first year, and it is still something we have to think about.

In the beginning, we did not have sophisticated monitors. A lot of it was very manual: I used ChatGPT every day myself, users reported issues quickly in Discord, and I watched support emails/reviews closely. Because the extension was used heavily, if something broke, I usually heard about it fast.

The harder part was that ChatGPT was not always the same product for every user. OpenAI often runs experiments and staged rollouts, so different users can have different versions of the UI at the same time. That meant we had to support multiple ChatGPT variants, not just “the current version.”

I also paid very close attention to ChatGPT releases and UI changes. Most people would see a new ChatGPT feature and just get excited. I would get excited and nervous at the same time, because every UI change could mean something in the extension might break. I'd sometimes look for hints in the frontend code before features fully rolled out so I could prepare.

The Chrome Web Store review process made this even harder. With a normal web app, if something breaks, you can deploy a fix immediately. With an extension, even if you fix the issue quickly, the update still has to go through review, which can take anywhere from a few hours to a few days. During that window, you can lose users even though the fix is ready.

Over time, we got better at it. The first version depended much more directly on the ChatGPT UI. After about a year, I reworked parts of the architecture to decouple Superpower from ChatGPT as much as possible, so fewer things depended on fragile DOM details. That helped a lot.

But it is still part of the job. Building on top of a fast-moving product gives you access to existing user demand, but you also inherit the platform's instability, experiments, review delays, and rollout complexity.

What's the task you keep putting off the longest?

Thank you. I think the key is to make the user conversation about their workflow, not about your business.

When I talk to users, I try not to frame it as "help me grow my product." I frame it as "what were you trying to do, where did it break down, and what would have made this easier?" That keeps the conversation useful for both sides.

A few things helped me:

1. Separate support from selling. If someone reports a problem, solve the problem first. Don't turn every support interaction into an upgrade pitch.

2. Look for repeated pain, not just loud feedback. One user request may be interesting, but when the same problem shows up across emails, reviews, Discord, Reddit, and your own usage, it usually deserves attention.

3. Be transparent when you cannot build something yet. Users usually respect "I agree this matters, but I need to prioritize X first" more than vague promises.

4. Keep giving value before asking for value. For a meal planning app, that might mean useful recipes, planning tips, saved time, or better weekly routines before pushing monetization too hard.

At the end of the day, a for-profit product is not at odds with helping users. The business works only if the product keeps solving a real problem. The danger is when the conversation becomes extractive instead of genuinely useful.

What's the task you keep putting off the longest?

Thank you! Appreciate the support!

What's the task you keep putting off the longest?

Best of luck to you!

What's the task you keep putting off the longest?

Thanks, I really appreciate that. I'm still figuring parts of it out too :)

What's the task you keep putting off the longest?

Thanks for reading and sharing your feedback!

What's the task you keep putting off the longest?

Not just Reddit, but honestly there was no big launch plan.

The first discovery came mostly from being early. I released Superpower within the first week after ChatGPT launched, and it was the first ChatGPT browser extension on the Chrome Web Store. At that moment a lot of people were searching for ways to improve ChatGPT, so timing helped a lot.

I shared it on Reddit early on, and I did a Product Hunt launch a few months later. But most of the growth came organically through browser store search, word of mouth, users recommending it to other ChatGPT power users, and people mentioning it in communities.

The lesson for me was that timing + a clear pain point can be a much bigger growth lever than a polished launch. People already had the problem, they were actively looking for a solution, and the product was simple enough to understand immediately.

What's the task you keep putting off the longest?

I agree with this. The extension usage is the deeper asset because it represents actual workflow habit. If someone opens Superpower every week inside ChatGPT, that tells me the product is part of how they work, and that is much closer to paid conversion than a newsletter subscriber.

I don't really think of the newsletter as "the real business" instead of the product. I see them as connected assets. The extension creates the daily/weekly habit, and the newsletter gives us reach, trust, sponsorship revenue, and a direct audience outside the browser stores.

The healthiest version is when they reinforce each other: the product gives people a reason to care, the newsletter keeps us connected to the AI market, and both help grow the overall Superpower ecosystem.

What's the task you keep putting off the longest?

I agree with the platform lesson. One of the biggest reasons I started Superpower Daily was exactly that: I did not want the entire business to depend only on browser store discovery or ChatGPT's UI.

The extension is still the core product and the deepest usage signal. If someone uses Superpower inside ChatGPT every week, that is real workflow habit. But the newsletter gives us something different: a direct relationship with the audience, independent distribution, and a way to stay connected even if platform dynamics change.

So I see them as two sides of the same strategy. The extension rides the platform and solves the workflow problem where users already are. The newsletter helps us own more of the customer relationship and build an audience outside the platform.

What's the task you keep putting off the longest?

Thank you! appreciate it.

What's the task you keep putting off the longest?

Exactly. Timing mattered a lot, but only because the first version solved a real problem.

If I had waited until it felt polished, I probably would have missed the moment. The first version was rough, but it was useful enough for people to immediately understand why it existed.

That early feedback was also more valuable than extra polishing would have been. Users quickly showed me which problems mattered most: folders, search, export, prompts, and reuse. The product became much better because it was already in people’s hands.

Thanks! The most useful testing has been end-to-end smoke tests around the exact ChatGPT workflows the extension depends on.

For an extension like this, unit tests are helpful for isolated logic, but most real breakage comes from the platform UI changing: selectors move, buttons change, dialogs get renamed, or a flow behaves slightly differently.

So the highest-value tests are the ones that open ChatGPT in a real browser and check core paths like: loading the sidebar, finding conversations, opening managers, exporting, using prompts, and making sure injected UI still mounts in the right places.

I also try to keep selectors and DOM assumptions centralized as much as possible, so when ChatGPT changes something, the fix is usually in one layer instead of scattered across the whole extension.

If I were starting again, I would invest in that kind of browser-level regression testing much earlier.

Thanks! Monetization was not built in from day one. For the first several months, the extension was completely free.

I wanted to first prove that people actually used it regularly and trusted it. With browser extensions, especially ones that touch a product like ChatGPT, trust is a huge part of adoption. If I had introduced paid features too early, I think it would have slowed down the feedback loop.

Once I saw organic growth, repeated usage, and clear power-user behavior, Pro made more sense. The paid plan was added around limits and advanced workflows rather than locking the basic value away.

My rough advice would be: make the free version useful enough that people understand the product, then charge when you can identify the users who are clearly getting ongoing value and need more capacity, convenience, or power features.

Thank you!

Thanks for reading!

Thank you :)

That is a good point. Low pricing can make validation look cleaner than it really is.

For Superpower, I kept the extension free first because I wanted usage, trust, and feedback before trying to optimize revenue. But I agree that once you introduce paid plans, pricing affects the quality of the signal.

A free user can tell you whether the product is useful. A paying user tells you whether the value is urgent enough to budget for. Those are different signals.

I still think free can work well for distribution, especially with extensions, but the paid tier needs to be priced high enough that users are making a real decision, not just saying yes because it is cheap.

That is exactly the right question.

The newsletter definitely started as a hedge against platform dependency. Browser stores, Chrome policies, and ChatGPT UI changes are all things I do not fully control, so having a direct audience matters a lot.

I do see overlap, but I would describe it as more than one loop rather than one perfect retention loop.

Some readers come back to the extension when I mention a new feature, a workflow, or a product update. That is measurable through traffic, installs, and upgrade activity around those sends. But not every newsletter reader is an active extension user, and I do not assume the full subscriber base behaves like a product-retention channel.

The stronger value is that both products keep me close to the same market: people trying to use AI more effectively. The extension shows me workflow pain. The newsletter shows me what people are paying attention to in the broader AI space. When those two signals overlap, it usually points to something worth building or explaining better.

Thanks, that is exactly how I think about it.

Free worked because it had a job: get the product into people's hands, build trust, collect feedback, and create distribution in a category that was still very new.

If free is just "I'm afraid to charge," it can become a trap. But if you know what it is supposed to produce, it can be a very strong strategy.

And yes, shipping early mattered a lot. The first version was not polished, but it solved a problem people already felt. That gave me real feedback much faster than building privately for months would have.

Thanks for reading :)

Thank you. The extension was completely free for about the first nine months.

I wanted the early phase to be about usage, trust, feedback, and distribution rather than revenue. In a new category, especially with a browser extension that runs inside ChatGPT, I felt trust mattered more than charging early.

When I introduced Pro, I tried to avoid taking away the core value people already depended on. The free version stayed useful, and Pro was mostly about higher limits and more advanced workflows for power users.

There was some pushback, which is expected any time a free product adds paid features. But it was not the dominant reaction. Most users understood that maintaining a fast-growing extension takes time and infrastructure, and the people who used it heavily were usually the ones most willing to support it.

The newsletter monetization also came gradually and organically. At first it was just a way to share useful AI updates with users. Sponsorships came later, once the audience was large enough and the content had its own value.

Exactly. Free needs a job.

For Superpower, the job was distribution, trust, feedback, and habit-building. I wanted people to use it enough that the value was obvious before asking them to pay.

But free without a plan can become dangerous. It can train users to expect everything forever, and it can make the business harder to support as usage grows.

The important part is knowing what free is supposed to produce and when the paid upgrade becomes natural. For me, that happened once there were clear power users who wanted fewer limits, more convenience, and more advanced workflows.

Thank you. 🙏🏼

What's the task you keep putting off the longest?

Totally fair question. We still do not have a strong automated detection setup for this, and that is one of the things I would build earlier if I were starting again today.

Right now it is still mostly reactive through user reports, reviews, our own usage, and support messages. Over time, though, we have tried to reduce how much of the extension depends directly on fragile DOM details.

A few things that help:

• avoid depending on the ChatGPT DOM when we can

• keep features decoupled so one broken selector does not break the whole page

• use fallback selectors where possible

• fail safely when something is not detected, instead of forcing behavior and risking a broken ChatGPT experience

• move more logic into extension-owned UI/state instead of relying on page structure

So the honest answer is: we are better at surviving breakage now, but still not as good as I would like at detecting it before users do. At this scale, you are right that reaction time matters a lot more. That is probably one of the biggest operational lessons of building on top of someone else's fast-moving product.

What's the task you keep putting off the longest?

Thanks, really appreciate that.

And I completely agree. The hard part is usually not the first version of the code. It's staying close enough to the people using it every day to notice what's confusing, what’s repeated, and what actually saves them time.

Even simple tools can become very valuable if they remove friction from a real workflow. That was the biggest lesson for me with Superpower too. Start with a problem you understand, ship something useful, then keep tightening it based on real usage.

What's the task you keep putting off the longest?

That makes sense, and I don’t think free-first is universal.

For Superpower, free worked well for two reasons. First, it was a browser extension in a messy/new-ish category where trust was the real bottleneck. People already had ChatGPT open, but they didn't know if they wanted an extension touching that workflow, so a generous free version reduced the decision friction.

Second, we didn't have a big social account, community, or distribution channel when we started. Free became our distribution. It gave people an easy reason to try it, talk about it, and recommend it before we had any audience of our own.

For a writing app, especially in an established category, one-time paid + trial can be totally reasonable. The user already understands the category, so the question is less "what is this?" and more "is this better for me than the other tools?" In that case, a focused trial may communicate value better than a forever-free tier.

I'd still want some low-friction way for people to experience the product before paying, but I wouldn't default to free-first for every product. The pricing model should match the trust gap, the market maturity, and your distribution starting point.

What's the task you keep putting off the longest?

What's the task you keep putting off the longest?

Thank you, I appreciate that.

For the first version, most of the core functionality was handled locally in the browser. That was intentional because I wanted the extension to be lightweight, fast, and cheap to run. Features like organizing, searching, and exporting ChatGPT conversations were much easier to scale when the user’s browser did most of the work instead of sending everything to a backend.

There was backend infrastructure later for things like accounts, subscriptions, newsletter workflows, and some product systems, but the early MVP did not depend on a heavy backend database for the core ChatGPT workflow.

That local-first approach helped a lot as a solo founder. It kept costs low, made the product feel faster, and reduced the amount of infrastructure I had to maintain while usage was growing.

What's the task you keep putting off the longest?

For me, the biggest advantage was timing. I launched within the first week of ChatGPT, so there was already a huge wave of people looking for ways to improve the experience.

Early on, Reddit was probably the most useful manual channel because people were already discussing ChatGPT workflows, prompts, and missing features. Product Hunt helped too, though that came later. But after the initial push, most growth came from organic discovery, Chrome Web Store search, reviews, and word of mouth.

So I'd say the best "channel" was being early with a useful solution in a fast-growing category. Reddit helped get the first feedback loop going, but timing plus real user pain did most of the heavy lifting.

What's the task you keep putting off the longest?

Thank you!

What's the task you keep putting off the longest?

Thanks for reading 🙏🏼

What's the task you keep putting off the longest?

Thanks, I appreciate that. I agree. Distribution works much better when it is part of the product loop early: build, share, listen, improve, and repeat. The feedback you get from that loop often shapes the product more than any private roadmap would.

What's the task you keep putting off the longest?

It was more "not yet" than a fixed 9-month plan.

In the beginning, the priority was trust, usage, reviews, and feedback. I wanted the product to become part of people's real ChatGPT workflow before asking them to pay. Once I could see repeated usage, clear power-user behavior, and enough demand for deeper limits/features, Pro started to make more sense.

If I were deciding again, I probably wouldn't pre-commit to a specific free window. I'd define the signals instead: retention, repeated usage, users asking for more, clear upgrade use cases, and enough trust that paid feels like a natural next step rather than a surprise wall.

What's the task you keep putting off the longest?

🙏🏼

What's the task you keep putting off the longest?

Exactly. A useful free layer can build trust much faster than a demo that feels locked down. The key is making sure free has a clear strategic purpose: usage, feedback, trust, and distribution. Then paid should expand the workflow for people who hit the natural limits, not punish them for trying it.

And I agree, the "solo + narrow problem + useful free layer" pattern works across a lot of categories. Starting from your own messy camera roll is probably a strong signal that the pain is real.

What's the task you keep putting off the longest?

For us it was more about repeated usage patterns than one exact threshold.

The people who converted were usually already using ChatGPT heavily and had made Superpower part of their workflow. Once someone has enough conversations, folders, prompts, exports, or saved work, the limits start to feel like friction. Pro made sense because it removed that friction.

So I didn't think of paid as "convince a free user to pay for the same thing." It was more: keep the core product useful for free, then charge the users whose workflow has grown past the free limits. The upgrade works best when it feels like the natural next step, not a wall they hit too early.

What's the task you keep putting off the longest?

What's the task you keep putting off the longest?

Thank you. I appreciate that.

For the first version, I was focused almost entirely on solving the problem and growing usage. I didn't have a polished paid plan built into the 3-day MVP. The extension was free for the first several months, and that helped build trust, reviews, feedback, and habit before monetization became the main question.

When Pro came later, I tried to make the transition feel like an upgrade for power users rather than taking away the core value people already depended on. The free plan still needed to be genuinely useful. Pro was mostly about removing limits and unlocking more advanced workflows for people who were already using ChatGPT heavily.

For your QR tool, I'd look closely at whether signup is being introduced before users feel enough saved value. If people generate and download once, they may not have a reason to create an account yet. The paid/useful moment might be around saved brand kits, reusable QR campaigns, analytics, team assets, editable destinations, or history. In other words, charge for the workflow that continues after the first QR, not just the one-time generation.

Thanks Emma, I appreciate the kind words. Listening to users and building around real workflow pain has been a big part of the journey.

What's the task you keep putting off the longest?

Superpower currently only supports ChatGPT.

Thank you!

Thank you :)

What's the task you keep putting off the longest?

Thank you. Appreciate the support!

Thank you :)

Glad you found it helpful!

Thank you!