
Daytrip is a local-discovery app for the person who already lives here and has run out of ideas for their own city. It is live on the App Store as a soft launch. Not a trip planner for tourists, and not an event finder for a specific night out. I designed it and built it solo, directing AI coding agents, from a few sketches to a running product with a real backend.
Introduction
The idea came after returning from my first solo trip to Portugal. Why can't I make everyday life feel like travel? I wanted a cohesive, done-for-you day trip experience that someone could just pick up, go, and discover their neighbourhood, proving that I could bring an idea from my head to something real.
My role
I worked on everything, with the help of Claude. Strategy, audience, product spec, voice, brand, the visual system, the motion system, the mockups, the component library, the API design, the data model, the launch plan, and the operations tooling.
I decided what gets built, and the strategy to take. Claude helps me find any spots I may have missed, particularly around implementation. Having some front-end background helped me understand the trade-offs better. For example, Claude suggested a live server instead of simpler API functions, which would have added unneeded costs for this app.
V1 in Replit
In late March 2026, during a Replit buildathon, I wrote a build plan and gave it to Replit Agent as the prompt. Within about a week it ran on my phone: a map, trips, stops, and a plan builder.
Then I restarted from scratch because Replit felt like a black box. I couldn't see inside what the agent had made, so I couldn't reason about it or trust it. However the buildathon showed me people liked the pitch. The project drew support from over 30 community members. My takeaway that week: building is quicker than ever, but getting something in front of people is as hard as it was before AI.
## Overview Daytrip is a hyper-local, community-driven daytrip discovery and planning app. Locals submit, browse, and build itineraries from real neighbourhood stops. The tone is human, low-key, and curious; not touristy. ## Stack (let Replit decide) Mobile app. use Google Maps. ## Visual Theme 1980s/90s paper map aesthetic - Base map: custom Google Maps JSON style, desaturated, cream/tan base, reduced POI clutter, warm road tones - Typography: slab serif or condensed retro font (Arvo, Zilla Slab, Playfair Display) - Cards: off-white, subtle grain, colored top stripe per daytrip - Icons: simple retro line-drawn category icons - Route lines: thick polylines, color-coded per trip ## Data Models User, Daytrip (stops[], localness_score, saves_count, color, status: pending | approved | rejected), Stop (name, category, coords, notes), Plan (stops[], source_daytrip_id, generated_prompt, synced), Review (rating, localness_rating) ## App Structure Bottom nav: [ Explore ] [ Plans ] ---- [ + ] - Explore: map base with browse sheet overlay - Plans: saved, submitted, and local draft plans - +: always visible, opens Create Plan ## Explore Tab - Each approved daytrip = thick color-coded polyline connecting its stops in order - Tapping a polyline or pin highlights its card in the browse sheet and snaps to it - Browse sheet peeks ~30% up with a horizontal row of daytrip cards; pull up for the full list - Card: colored top stripe (matches polyline), title, theme tag, stop count, localness score ## Daytrip Detail (in-sheet) - Stop-by-stop list: category icon, name, short note - Crowd localness score, 1-5, "Tourist spot" to "Hidden gem" - "I tried this" leaves a rating + localness score - "Build a plan with this" clones the stops into a new plan draft ## Create Plan [ Edit manually ] [ Generate ] - Manual: drag to reorder, swap, delete, add note; free-form, no enforced template - Generate (subscription): prompt field with starters like "a slow rainy afternoon"; pulls from submitted stops only; result is an editable stop list - [ Save locally ] [ Save & Submit ] ## Submit a Daytrip Name it, pick a theme tag, add stops with a short note each, self-tag a vibe. Goes pending, then live on the Explore map after admin approval. First submission unlocks 1 free week of premium. ## Auth Gate Contextual, not aggressive. Inline "Sign in to save". Full-screen flow only on Submit or Generate. ## Local Draft Sync Not logged in: saved to localStorage. Log in later: drafts sync to Supabase. Never signs up: draft stays on device. ## Monetization Browse: always free. Save, review: login. Submit: login + 1 free week. AI plan generation: $7/mo or $49/yr. ## Build Order 1 Supabase schema, 2 seed data (5-10 Vancouver daytrips), 3 Google Maps base, 4 Explore tab, 5 Daytrip Detail, 6 Auth, 7 Create Plan, 8 Submit flow, 9 Plans tab, 10 AI Generate, 11 Stripe, 12 Admin route
Decisions
An added benefit of building with Claude is that I kept a running log of every decision, built into the spec documents, code comments, and commit history. These four shaped the product most.
- A specific person over every kind of traveller. I set anti-personas, the tourist and the event seeker, before drawing any screens, because I wanted to solve one specific problem first.
- A vibe filter over tags. I iterated from 8 moods and 12 scenes down to three dials: energy, scope, and social. Fewer, broader choices reduce decision fatigue, and stop the filter from leaving too few trips to pick from.
- A paper map brand over a modern travel look. It evokes how it felt to discover an area before reviews and the internet.
- An AI workflow that kept me as the decision maker over handing it all to an agent. Hand sketches, ASCII wireframes, mockups, then a component library. Each stage was something I could review and reason about before the next, which is what the Replit version lacked.
Outcome
I learned a lot of what it takes to bring an app to life. It showed me that when I fully own an idea I'm excited about, the words come easily - something I struggle with when an idea isn't genuinely interesting to me.
Despite not using its outputs, trying an AI building tool helped me get a sense of how an app should be structured. Intentionally asking it to teach me while it builds helped me broaden my understanding. In the end, the design is still owned by me as the creator of this app.
Team
Solo, with AI coding agents
Tools
Penpot/Figma/Paper, Claude, Storybook, Replit, Supabase, Vercel, PostHog
Skills
Product strategy, UX, brand, component systems, API and data modeling, AI-directed build
