Back to all features
AuraNote V1 marketing site hero
Product design

AuraNote V1

Built inside AuraNote2025 — 2026

AI journaling web app built on the bet that a journal should give something back. Contextual prompts, streaming reflections, and a unique generated image for every entry. Tested across a 9-month private beta with 100 invited users.

Context

AuraNote is an AI journaling app I co-founded. My partner and I both had long-running mental wellness practices, and both had bounced off journaling. The experience felt one-dimensional, all output and nothing returned. Each of us had started using AI to process feelings, him in chat asking for images back, me out loud. The image idea clicked, and the pitch crystallized: a journaling practice that gives something back, makes you feel something, gives you something to hold.

Challenge

Most journaling apps absorb writing and offer nothing in return. We wanted to test whether returning something visual, reflective, or conversational could turn a practice that most people abandon into one they came back to. I had under four months to design, build, and ship a complete AI-native product end to end before the private beta opened in April 2025.

Decisions

  • Built a context-aware prompt engine that pulled from each user's recent trajectory to generate writing prompts and intra-entry provocations, nudging the writer deeper without breaking the flow.
  • Shipped a streaming reflection pipeline that delivered the post-entry response token-by-token, tuned to reference specific phrases from the entry so it read as a reply rather than a summary.
  • Wired automatic mood classification into an image pipeline that generated a unique visual header for every entry, cached in S3, so the feed became a scrollable visual memory of the practice.
  • Layered an open chat surface over the entry, the reflection, and the image so users could explore further. Scoped as non-clinical, not a therapist substitute.
  • Designed and built the whole product solo, front end, back end, infra, and brand. Shipped as an installable PWA scoped as a serious demo testbed for the thesis.

Result

The 9-month private beta ran across 100 invited testers, peaking at 45 DAU. Testers praised the writing infrastructure, the speed, and the prompt engine, with reflections close behind. Image generation was the original pitch, but it landed for only a small enthusiastic minority and didn't carry the rest of the cohort. Stickiness eroded as the beta matured. V1 was a confident swing that proved the harder question wasn't whether we could give something back. It was understanding why people who say they want a journaling practice still don't keep one. That question is the foundation of V2.

100 testersClosed beta
Stack
Next.jsOpenAIPostgreSQLRedisS3ClerkTanStack QueryTailwindshadcn