Loop
A personal assistant that suggests what to do and who to meet up with, so you stay in the loop with the people and places you care about most.
Screens from loopaiapp.com, Loop's launch site.
Plans with friends, without the group chat stall
Loop combines your calendar, your location, your taste and the people you like spending time with, then suggests the right place or event at the right time. The north star says what it's for: group plans completed each week where at least two people actually showed up.
Ask Loop
An AI concierge you can ask in plain language, like "somewhere quiet for brunch with three friends."
For You feed
Personalized picks, each with a "why this fits" card that explains the recommendation.
Today's Loop
Your day drawn as a route on a map, from home to each stop and back.
Group plans
Invite friends, find a time that works, and check in when you arrive.
Not just what's good, but when you can realistically go
For every pick, Loop works out the soonest time you could actually arrive. It combines your free time, the venue's real hours (including places open past midnight) and how long it takes to get there, and it keeps that time current as the day goes on. Early builds suggested a venue 32 miles away for 22 minutes from now. Fixing cases like that became the basis of Loop's patent-pending method.
Each tool solves one piece. Loop plans the whole thing.
The biggest real competitor is the group chat. Discovery apps suggest places but ignore your friends and your schedule, and invite apps only help once someone has already done the planning. Loop combines availability, interests and location in one plan.
| Product | Knows who’s free | Suggests what to do | Plans it with friends | Nudges at the right time | Finds the most convenient time and place |
|---|---|---|---|---|---|
| Loop | Yes | Yes | Yes | Yes | Yes |
| Group chat | No | No | By hand | No | No |
| Google Maps + Gemini | No | Yes | No | Minimal | Distance only |
| Yelp | No | Yes | No | No | Distance only |
| Partiful, Apple Invites | No | No | Invites only | No | No |
| Howbout | Shared calendars | No | Partial | No | Time only |
| Snap Map | Location only | No | No | No | No |
Swipe the table sideways to see every column.
What the combination unlocks
The decisions behind the product
Keeping AI and API costs flat as usage grows
Loop builds and maintains its own index of places and events, so the app reads its own data instead of paying per tap. Per-user AI quotas and spending guards keep costs flat as usage grows.
Recommendations that explain themselves honestly
A review caught "why this fits" reasons that claimed personalization without evidence, and a small friend signal outranking a much stronger match. Now a reason only appears when real evidence backs it, reasons sort by actual weight, and a test pins that behavior.
Measuring AI quality instead of eyeballing it
Evaluation scripts score the feed: a quality gauntlet across categories, an LLM judge that rates the relevance of the top five picks, and a readiness check per city. Engine changes have to hold those scores.
Fixing what users see first
Photo coverage was the first thing users noticed, so a reproducible probe traced the failures and the fixes were measured, not guessed.
Keeping keys off the phone
Paid services run behind server-side guards with per-user and global caps.
Measured, not estimated
Expo · React Native · TypeScript · Postgres + PostGIS · Jest
A product team of one
Loop was built like a small product team: one person owning product, design, architecture and final review, with AI coding tools handling much of the implementation. A few habits kept quality high as the codebase grew.
- Independent review. No change merged without a separate check against written acceptance criteria.
- Living documentation. Design docs, a launch tracker and a findings log kept everything current. The rule: every number in a doc comes with the command that reproduces it.
- Real-device testing. Every build tested on a real phone before it moved forward.
The real app, running today
The illustrations above show the idea. These are screens from the current build, captured from the app itself.
Group planning first, then a personal AI assistant
Loop launches in Dallas as a group planning app, with Los Angeles, Austin and New York City next. Once the user base is established, it evolves into a personal AI assistant that knows the places, events and people you love, and makes sure you don't miss them.
The end state: opt in, and every morning you wake up to a personally curated list of things to do, from AI that knows where you've been, what you like and who you want to see.
Want to try it or talk?
Loop is in public beta on iPhone: install it with TestFlight at testflight.apple.com/join/nhfZxpSc. On Android, request a download link.
Loop's places data covers Dallas–Fort Worth today. Outside DFW, set your location in Profile to a Dallas address to see real recommendations.
Walkthroughs of how Loop was built are available on request, too.
nick_casey@icloud.com