How we shipped a journaling app in five weeks
Jot Journal is a journaling app for people who want a private place to think, with an AI that learns their voice over time and surfaces patterns they didn't notice. The founder came to us with a single paragraph and a deadline: thirty days to a TestFlight build, six weeks to App Store and Play Store. We hit both. This is the week-by-week of how, including the parts we'd do differently.
Week 1 — spec and design in parallel
Most teams treat spec and design as sequential phases. We don't. Our co-founder Spruha started high-fidelity Figma screens on day one while our engineering team scoped the data model and built the first React Native + Expo skeleton in parallel. By the end of week one we had clickable Figma flows for the top five screens AND a deployed Expo Go build with empty pages but real navigation. Two artifacts, one week.
The reason this works is that with Cursor and Claude pairing, the engineering work in week one is mostly mechanical — auth, navigation, base components, the data model. Things we've built dozens of times. We don't need final designs to scaffold them.
Week 2 — the screens go live
By Monday of week 2, the top five Figma screens were finalised. We translated them into React Native components in two and a half days — a number that would have taken the same team a full week without Claude pairing. Cursor wrote 70% of the layout code from screenshots. Our engineers spent their time on the parts that needed taste: spacing, motion, the parts users actually notice.
By Friday of week 2, the founder could install the app on his own phone via Expo Go and use a working version of every primary flow. No backend yet — everything was local. But it felt like a real app. And feeling like a real app to the founder in week 2 changes everything about the next four weeks of feedback.
Week 3 — backend and AI
Jot has two AI features: voice journaling that transcribes and summarises, and a chat that knows the user's history (a RAG system over their journal entries). We built both in week 3, in parallel with the backend stand-up. The RAG system was the part most teams would budget two weeks for. We did it in three days because it's a pattern we've shipped before, and Claude can help us scaffold the embedding pipeline, the vector store integration, and the chat interface in hours.
Week 4 — polish and the parts we missed
Week 4 is where we do the work nobody talks about: the empty states, the offline modes, the loading skeletons, the push notification scheduling, the rate limiting, the friend who tries to sign in with the wrong email. These are the things that make an app feel finished. They take a week. They take a week even with AI helping. Don't believe anyone who tells you otherwise.
We also got the first thing materially wrong in week 4: we built an onboarding flow that was too long. The founder pushed back hard and we cut it from 7 screens to 3. We should have shown him a working prototype of the onboarding in week 2, not week 4. Lesson: the boring screens deserve early prototypes too.
Week 5 — submission
Week 5 is store submission. EAS Build for both platforms. App Store screenshots generated from Figma. Privacy nutrition labels filled out (this is genuinely tedious). Listing copy written by Claude and edited by the founder. Submitted to TestFlight on Tuesday, App Store Review on Thursday, Play Console on Friday. Approved on the following Monday and Tuesday respectively.
What it cost
$32k flat fee for a team of three engineers and one designer for five weeks, including the App Store submission grind. The same engagement at a traditional agency would have been quoted $80–120k and four months. Same outcome. Different tools, different pace, different price.
What we'd do differently
- [01]Show the founder a working prototype of the boring screens (onboarding, settings, account) in week 2, not week 4.
- [02]Spec the empty states and error messages with the same rigor as the happy path.
- [03]Set up crash reporting on day one, not week 4. We caught bugs in week 5 we could have caught earlier.
- [04]Run the RAG evaluation harness from week 3, not week 4. We had two embedding-quality issues we should have caught a week earlier.