Inspiration
In 2012 I crashed.. pretty hard. After a six-month burnout I really had to stop everything. After some months of searching, I discovered meditation through the Headspace app. The streak counter that was meant to keep me “accountable” soon created fresh anxiety; if I missed a day I felt guilty, not calm.
Fast-forward: I created the official Plum Village App (now 1.5M users) so others could practice without anxiety.
And when I started learning French, every language app and especially Duolingo, pushed my anxiety buttons with the streaks and tons of notifications.
Tjat was the spark for Zenglish: a calm, AI-native language-learning app grounded in meditation principles and the 80/20 rule of vocab acquisition.
What it does
- Personalised AI micro-chats with Zen our capybara mascot, tuned to your interests and level.
- 80/20 Curriculum Engine: surfaces the 20 % of words & structures that unlock 80 % of daily conversation for your specific context.
- Guided mindful pauses (short breaths ) between exercises to keep anxiety low.
- Adaptive spaced repetition via flash-cards that grow organically from your chat history.
- Zero streaks, zero guilt: you practise because it feels good, not because you’re scared of losing a flame.
Live demo: https://zenglish.io
How we built it
I started in Bolt and for the more complex stuff I asked my friends Theo and Rishi to help. We use a mix of NextJS, OpenAI, Two devs, one designer, 18 days.
Challenges we ran into
Balancing calm with engagement: because we stripped away gamification, we had to invent some subtler feedback loops (like serene confetti & bells sounds instead of XP).
Contextual AI prompts: getting the OAI models to stay concise, friendly and error-free inside a 30-second chat that required dozens of guard-rails and semantic filters, not easy..
Speech latency: browser Speech-to-Text APIs add jitter, especially Safari; we built a fallback text-only flow for low-bandwidth users.
Bolt + different repos: we've been collaborating in a... messy way over different repos which meant adding an extra level of complexity within the Bolt framework.
Accomplishments that we're proud of
- Shipped a usable web app within the hackathon window that's teh main thing YAY!
- We have used lots of things I learned over the previous 9 years working on the Plum Village app!
- First cohort of real learners finished a 15-minute French session and reported a 4.8 rating (“felt relaxed while learning”)
What we learned
- Calmer UX is not so obvious to build, especially in the current dopamine-fueled landscape
- Early users love our cusomization of content: letting them type (“I want to learn bakery vocab”) before a session boosted retention of new words!
- AI guard-rails are never “done": Theo's iterative prompt tuning is pretty hardcore
What's next for Zenglish.io - Calm Language Learning
Mobile App: we are in the process of getting it approved by the App Store Language packs: Spanish & French first, co-created with native speakers who have mindfulness experience. Research partnership: with a cognitive-science lab to study stress biomarkers during calm vs. gamified learning sessions :)
Much love from Cata, Rishi and Theo!
Built With
- bolt
- entri
- fastapi
- firebase
- langchain
- netlify
- nextjs
- openai
- prisma
- revenuecat
- supabase
Log in or sign up for Devpost to join the conversation.