Case study · P/01 Live on iOS & Android

Prismo

An AI learning journal that turns what you read, watch and hear into organized notes and spaced-repetition quizzes.

Role: Solo founder & engineer: product, design, mobile, backend, AI and release
Timeline: 2026 – present
Platforms: iOS · iPadOS · Android · Web
Status: Live on iOS & Android

Fig. 1 Prismo

870+

installs on iOS & Android since launch
68

countries with active users
Mar 2026

launched on the App Store & Google Play

§ 1 · Overview

The product

Most of what we read, watch or listen to is forgotten within days. Prismo is a learning journal built to fix that: you write, dictate, paste or photograph what you learned, and an AI pipeline extracts the individual ideas, files them into a knowledge library it organizes for you, then quizzes you on them with spaced repetition so they stick.

The user stays in control: the AI only ever proposes changes (create, update, move, merge), and nothing touches the library until it's approved on a review screen. I designed and built all of it: the mobile app, the AWS backend and the LLM pipeline.

Key features

Capture by typing, dictation, photos, PDFs or links
AI proposes notes and folders; nothing changes until you approve it
Quizzes on your own notes, scheduled with spaced repetition
A knowledge graph of how your ideas connect

§ 2 · Engineering

How it's built

01 Capture
- Text & voice
- Photos (OCR)
- PDFs & links
- Share sheet
02 Expo app
- Journal entry
- Review screen
- Local library
03 AWS (CDK)
- API Gateway + JWT
- Lambda pipeline
- Cognito
- S3 sync · DynamoDB
04 LLM
- Understand
- Write
- Validate (deterministic)
- Repair ×1

Fig. 2 How an entry becomes knowledge. Proposed changes return to the app and are applied only after the user approves them.

01

A multi-stage LLM pipeline with deterministic guardrails

Entries go through understand → write → validate → repair. The model works from a compact summary of the user's library rather than the whole thing, and a deterministic, unit-tested validator rejects unsafe changes (duplicate or cyclic folders, deleting notes it hasn't read) before anything reaches the user, with at most one repair call. An offline evaluation set tracks quality against a cost target of under $0.50 per user per month.
02

Offline-first sync across devices

Libraries sync through S3 with vector clocks for conflict detection and tombstones for deletions, so edits made offline on two devices merge without losing data.