BLang: 60,000 users, and not a business
The product worked and the company did not. This page is mostly about the second part, because that is where I learned something.
The idea
I came back from LaunchX at the University of Michigan in August 2023 having spent a summer on how companies are actually assembled, straight off a year of AP Macroeconomics, Microeconomics and Computer Science. BLang started within weeks of getting home.
The premise: language learning apps teach you a fixed syllabus, and people abandon them because the syllabus is not about anything they care about. BLang generated study modules from whatever the user was actually interested in. It produced reading passages in the target language on their chosen topic, with any span translatable back to English or Russian, plus generated grammar and vocabulary quizzes, plus an AI conversation partner for writing practice without a tutor present.
I ran the linguistics, legal, financial and marketing side. My co-founder, Mert Kaan Atan, did the technical heavy lifting.
What worked
It grew. At peak, more than 60,000 users worldwide, and these were people coming back to use it rather than sign-ups sitting idle. That came from social media marketing and a lot of direct physical outreach, including presentations like the one above to a university language centre whose students were exactly our use case.
Setting the company up taught me the parts of a business that are invisible from the outside: incorporation, the legal shape of a founding agreement, financial responsibility for something with real users attached.
What broke
We could not make a pound out of 60,000 users.
Three things went wrong at once. Ad integration never worked, a technical problem with the ad provider that we never resolved, which meant the free tier earned nothing. The premium tier converted in single digits. And AI tokens plus hosting scaled directly with usage, so every additional active user made the monthly loss slightly worse.
Meanwhile the market moved. When we started, generative AI inside a language app was a differentiator. By 2025 the incumbents had shipped the same features, and the gap we had built into was closing.
We stopped work in May 2025 and dissolved the company that summer.
What I would do differently
- Validate willingness to pay before building. We validated that people would use it, which is a much easier and much less informative question.
- Treat cost-per-active-user as a first-class metric from day one. We were measuring growth against a cost base that grew with it.
- Fix monetisation infrastructure before scaling acquisition. Marketing into a funnel with a broken bottom is just buying losses faster.
I would still do it again. Sixty thousand people used something we made, and the way it failed taught me more than a small success would have.
Tools
Meta Business Suite · Company formation & finance