Silvia Casagrande
Walkie
/26
The Problem
In Italy, 67.3% of women say they are afraid walking home in the evening, and 38% have given up going out because of the journey back (Censis, 2025). The existing category reacts after the fact: virtual escort apps offer company over video, SOS apps fire an alarm, navigation apps route you down better-lit streets. None of them solve the thing people actually want, which is someone to walk with.
Constrains & Product Architecture
Three findings governed every decision that followed:
The behavior already exists, people look for company on the way home.
Trust is built on identity: verified ID document, reviews from other users , a real profile photo , the possibilty to cancel a match at any moment.
The obstacle is density, not willingness, lack of company was the leading barrier.
The strongest qualitative concern was the risk of predatory use: someone downloading the app specifically to locate women travelling alone.
Execution & Key Decisions
Active Matching, Not a Map: Competitors in the category (PinkRoad in Italy, Juno in Canada) publish trips on a map the user has to search alone. Walkie compares trips and notifies. Three rules define it: a ±30-minute compatibility window, proactive push notifications, and partial matches on a shared stretch of route.
Progressive Disclosure as the Privacy Layer: Before both people confirm, each sees only a neighborhood-level zone, a roughly 30 minutes band and the transport mode. Exact meeting point, departure time and route unlock on mutual consent, so browsing the app cannot reconstruct anyone's travel habits. Location is used in real time and deleted at trip end.
Moderation That Actually Holds: Mutual reviews with a dispute path, three levels of moderation, and a re-registration block tied to ID verification which is what makes an expulsion effective.
AI Across the Whole Pipeline: Research, personas, flows, logo and brand palette in Claude Chat; low-fidelity wireframes in Google Stitch; interface screens in Claude Design, handed to Figma via MCP for manual refinement; a working Flutter prototype generated through the Figma-Claude Code integration.
Conclusion
The prototype runs end to end, built single-handed.
An empty quadrant, evidenced: 12+ apps assessed across virtual escort, route intelligence, SOS and companion matching, plotted on two axes. No Italian competitor offers real-time algorithmic route matching.
Open decisions, stated rather than hidden: which city to launch in, since density is what makes matching work; how to fund the app without pricing out the people who need it; and whether the verification selfie counts as biometric data under GDPR, which would reshape the requirements.
Full process notes & deep-dive wireframes available on Notion