By
@torren
socle is a discovery platform for finding and sharing places worth remembering, from restaurants and cafés to hidden gems. instead of endless reviews and search results, socle focuses on curation, personal taste, and recommendations from people you trust.
socle is built around a simple idea: finding somewhere good should feel personal. it brings together places, curated lists, recommendations from people you trust, and a taste profile that develops through your activity. i wanted a better way to discover places that fit you and keep the good ones somewhere worth coming back to.
the hardest part has been making those systems work together reliably. repeated activity shouldn’t let one place overwhelm your taste profile, and opening a place page shouldn’t count as physically visiting it. i worked on those distinctions alongside search, missing photos, chat navigation, loading states, and account isolation. testing kept revealing small problems that made a bigger difference to the experience than i expected.
i’m proud of how the two sides of socle are coming together. discovery helps people explore and share places, while the business side connects guest feedback with insights about their experience. the design took plenty of iteration too, from discovery’s warmer colors to the interactive hardware animation on the business homepage. the native iphone and mac versions are also progressing, though location tracking, notifications, and widgets still need further real-device testing.
how to test: click testers at the bottom-left of the sign-in page, or open socle.pages.dev/testers. enter stardance when asked where you’re from; capitalization doesn’t matter. a wrong answer gets one final question asking who you’re testing on behalf of.
you’ll receive a 24-hour testing account. explore discovery, then open the profile menu and look under your workplaces for stardance testers. there are owner and employee entries, and each tester gets a separate workspace with sample reviews so you can experiment without changing a real business’s data.
the linked public repository contains a limited discovery homepage demo; the full application and backend remain private. the devlogs cover the wider project, and ai’s substantial contribution to implementation and documentation is explained in the ai usage declaration.
socle’s pretty much ready to be shipped in a few minutes, just made a last minute testing flow for you stardance testers:)
socle was looking ready for testing, but using it exposed things i couldn’t leave broken. i fixed search, added sliding transitions between chats, and built the native foundations for notifications, four widgets, and location tracking. the biggest change was separating browsing from actual visits: opening a place page shouldn’t count as going there, and a visit now needs five minutes of reliable location evidence. i also rebuilt the business homepage with a more interactive design. the simulator and mac builds passed alongside 23 automated checks, though location accuracy and the apple integrations still need real-device testing.
i spent this session testing whether socle intelligence could give a business something useful beyond a summary of its reviews. i worked on how it uses customer feedback and business context, and added support for graphs and tables so answers are easier to understand. some responses still missed the point, so comparing answers and finding those gaps became a big part of the work. i also fixed cover-image bugs across the site. there’s progress, but i still want more evidence before trusting the answers.
one review can contain a complaint about slow food and a compliment about the server. getting socle intelligence to recognize both became a useful test of whether it actually understood the feedback. i experimented with models, settings, and context, while building the connection between the model and a business’s data. running it locally also introduced speed and reliability problems to work through. the hardest part is still knowing when several reviews reveal a pattern and when there just isn’t enough evidence to say.
this was where socle’s discovery and business sides started connecting. i worked on turning feedback, saves, visits, and employee recognition into insights a business could use. the difficult part was deciding when those interactions supported a conclusion, so i built around showing the evidence and admitting when there wasn’t enough. i also separated the broader patterns managers need from the feedback an individual employee should see. the first engine uses rules i can inspect and test, giving me a clearer foundation to build on.
i kept fixing one small issue and finding three more. this session turned into a sweep through socle’s navigation, spacing, loading states, and awkward interactions, trying to use it like someone who didn’t already know where everything was. there wasn’t one big feature, but the fixes removed a lot of little interruptions. we were hoping to ship afterward, then ran into repo issues. that pushed the release back while we worked out a cleaner source setup.
i’d gotten used to some of socle’s rough edges, which made them harder to notice. i went back through the main screens with a first-time user in mind and cleaned up confusing buttons, inconsistent layouts, navigation, and loading states. most fixes were small, but together they made the app easier to move through. i’m getting closer to the experience i want: people exploring places without having to figure out the interface first.
taste needed a practical way to test visits before full location tracking was ready. i replaced the old flow with a temporary visit button, which let me check how visiting a place changes someone’s profile. i also adjusted how actions are weighted and repeated activity is counted, so one place doesn’t overwhelm everything else someone likes. the screenshot keeps the exact scoring weights private, but the goal is straightforward: taste should develop from someone’s activity without becoming too easy to distort.
socle needed more places to explore as well as a cleaner interface. i added locations and filled out existing pages, then worked through inconsistent layouts, spacing, small bugs, and interactions that felt awkward. having more content also made the gaps in the experience easier to spot. this update brings the app closer to a wider test, where people can spend time discovering things instead of reaching the end of the available content.
before bringing in more testers, i wanted to clear out the obvious things that would get in their way. i went through the main screens fixing layout issues, inconsistent components, and navigation that took too much effort to understand. these were mostly small changes rather than new features, but they made the experience more consistent. the next round of feedback should help me find the problems i haven’t already learned to work around.
i used codex to help with implementation while i worked through socle’s screens and tested the main flows. that testing uncovered bugs in a feature i thought was already finished, so i went back, fixed them, and checked it again. i also cleaned up the surrounding ui and interactions. it was a reminder that finishing the code isn’t the same as finishing the experience, and there’s still more to catch before opening it to more testers.
