1500 万次浏览、1 万用户、1000 美元 MRR：我的应用教会了我代码从未教会的东西

Most of the year I spend building AI voice agents and software for startups across the US, UK and Europe. The engineering is the part I am comfortable with. But the product that taught me the most was one of my own, and almost none of what it taught me was about code.

LectureNotes AI is a note-taking app I co-founded with Benjamin Rhodes. A student records a lecture, and gets back a summary, the key takeaways, and a clean outline they can actually study from. In two months it hit around 15 million organic Instagram views, grew to nearly 10,000 users, and settled around $1k MRR, with zero ad spend.

This is the honest retrospective, written for other builders. Not a growth-hack listicle. What actually drove the reach, what the reach did and did not turn into, and the lessons I now apply to everything I ship. The engineering was the easy part to be proud of

The core loop is simple: point your phone at the lecture, get usable notes back. No setup, no manual note-taking, no fighting a transcription tool that spits out a wall of unpunctuated text.

As an engineer I wanted to be proud of the model pipeline and the app. And I was. But the honest truth is that the single most important technical decision had nothing to do with accuracy or latency. It was making the value land in one sentence. You cannot grow a product organically if a stranger has to think for ten seconds about what it does. Everything that happened later was downstream of that clarity, not downstream of the code quality.

That reframed how I think about building. The shareability of your core moment is a design decision you make while building, not a marketing phase you bolt on after. How the 15 million views actually happened

The reach came from short-form organic video. Not paid ads, not a clever launch stunt. The pattern behind it was not magic, it was matching the content to the exact moment the product mattered. Show the pain, not the product

The videos that worked were not feature tours. Nobody shares a screen recording of a settings menu. The ones that traveled showed the student's actual problem: the 8am lecture you cannot keep up with, the exam you are cramming for, the notebook that makes no sense the night before a test. Then the app quietly resolves it. The product was the punchline, not the premise. When the hook is a feeling every student recognizes, the algorithm does the distribution for you. Volume and iteration beat any single viral idea

No one on the team could predict which post would take off. Some did nothing. Some did numbers we did not expect. The only reliable move was to keep posting, watch what resonated, and make more of that. Reach at this scale is not one perfect video. It is a lot of attempts, honest attention to which ones stuck, and the discipline to repeat the winners instead of chasing a new format every week.

That is the same principle I lean on shipping software: ship, watch, iterate, do not fall in love with the first idea. The audience was already gathered

Students are one of the most concentrated, self-identifying audiences on short-form video. They follow study accounts, they share things with classmates in group chats, and they are online at exactly the hours the product is most useful. We did not have to manufacture an audience. We had to speak clearly to one that was already there and already had the problem. The gap between views and revenue

Here is the part most growth stories skip, and the most useful lesson in this whole piece.

Fifteen million views is a huge number. It became about 10,000 users and roughly $1k MRR. Those are real numbers I am proud of, but the ratio is the education.

Views are not users, and users are not revenue. Every step down that funnel loses people, and the losses compound. A view is a second of attention. A user is someone who downloaded, opened, and got value at least once. A paying user is someone for whom the value was worth a recurring charge. Each of those conversions is its own product problem, and no amount of top-of-funnel reach fixes a leak further down.

If I were doing it again, I would spend as much energy on the path from view to paid as we spent on the path from stranger to view. The reach proved the demand existed. Converting that reach is a separate, unglamorous discipline: onboarding that gets a student to their first good set of notes in under a minute, a paywall placed at the moment value is felt rather than before it, and retention that keeps them through the next exam season. The virality was the easy part to be surprised by. The funnel is where the real work sits, and it is engineering work as much as marketing work. What I carry into every product now

A few lessons from LectureNotes AI have stuck, and now shape how I build, including for clients.

Distribution is a product feature, not a phase. How easily a happy user can show a friend is something you design while building. Bake it in early.

A sharp niche outperforms a broad one. "Notes for students" spread precisely because it was narrow. It knew exactly who it was for and what week of their life it fixed. A vaguer "productivity app for everyone" would have had no natural audience to catch fire in. Specific beats broad, every time.

Speed compounds. We could post, read the response, and adjust in days, not months, because the team was small and the loop was tight. That same bias toward shipping fast is why I can take a client's idea and put something real in front of users quickly.

Honest numbers teach more than vanity ones. The 15M is the headline, but the 10k users and $1k MRR are where I actually learned. I would rather tell a founder the real ratio than a round number that flatters the story, because the ratio is the thing they can act on. Would I chase 15 million views again?

Yes, but with clearer eyes. Organic reach at that scale is a genuine advantage, and for a bootstrapped product with no ad budget it can be the difference between existing and not. It validated that students wanted this, cheaply and fast, and that validation is worth a lot when you are deciding what to build next.

But I would not mistake the reach for the business. The views told us the demand was real. Turning that demand into a durable product, one people keep paying for after the exam is over, is the harder and more valuable problem, and it is the one I would put first next time.

If you are an engineer sitting on a product with real pull and a funnel leaking underneath it, that gap is usually where the growth actually is. And most of the time, closing it is a build problem, not a marketing one.

I build products for founders and AI voice agents for startups across the US, UK and Europe through Null Studio. LectureNotes AI lives at lecturenotesai.org.