1
My wife is a huge word game nerd (Wordle, Spelling Bee, you name it!), and I wanted to create something fun for us to play daily that combined a couple of mechanics we enjoy: finding words in a grid and solving anagrams. It started as a personal project but evolved into something I thought others might enjoy too.
You're presented with a 4x4 grid of letters. Find words by connecting adjacent letters (like Boggle).
Finding words scores points and contributes to revealing letters in a hidden "Mystery Word" (an anagram of 8-10 letters).
The more words you find, and the faster you find them (streaks!), the more points you get. Longer words give bonus points.
Score milestones help reveal and eventually order the Mystery Word letters.
Once you think you know the Mystery Word, you can try to solve it for a final bonus score.
There's a new puzzle generated each day.
I built this using React, TypeScript, Vite for the frontend, and Firebase Hosting for deployment. The grid generation and dictionary logic are handled client-side.
This is still very much a work in progress, and I'd absolutely love to hear any feedback you have – gameplay balance, UI suggestions, bugs, anything!
You can play it here: https://playgridagram.com/
My wife and I are heading to Thailand soon, and while we're learning some basics, I was thinking about those quick, everyday interactions where a language barrier can be awkward – asking for directions, ordering street food, chatting with a vendor, etc.
I didn't want the friction of firing up a dedicated translation app each time or dealing with account signups for something temporary.
So, I decided to build Kodo Chat as a little side project to scratch my own itch. The idea is a super simple, temporary chat room focused purely on real-time translation between two people.
How it works:
One person ("host") visits the site, selects their language, and gets a QR code and a shareable link. The other person ("guest") scans the code or opens the link on their phone, selects their language, and they're connected in a temporary room.
Messages sent by either person are translated (currently using OpenAI - gpt-4.1 by default, but configurable) and displayed along with the original text on the other person's device.
Rooms and user links expire automatically (via Redis TTLs), so there's no persistent data or accounts needed.
It's built with Expo (React Native Web) for the frontend, Node.js/Express/Socket.IO on the backend, and Redis for session/room management. The UI is pretty basic right now (using React Native Paper), and the translations are only as good as the underlying AI model, but it seems to handle simple conversational stuff reasonably well in testing.
It solved my immediate need, and I thought others might find it useful, so I've open-sourced it.
Live Demo: https://kodo-frontend.onrender.com/ GitHub: https://github.com/jonathanleane/kodo
Would love to hear any feedback, suggestions, or criticisms you might have! Especially curious if anyone else has faced similar communication hurdles while traveling and what solutions they've found.
You could do the standard n=1 thing where you conduct a self-experiment and do a little write up, which is still cool, but is of limited value to the world... Not only because the sample size is too small, but because you probably didn't try to control for any confounding variables.
What I'm proposing to build is a platform that lets basically anyone create an experiment, with a proper scientifically sound method, and then other people on the platform can anonymously volunteer themselves as test subjects...
The idea would be to take away as much of the drudge work as possible from the citizen scientist except for coming up with the hypothesis (and designing a good experiment - but the platform could help with this too).
People who join the platform as willing study participants do so anonymously, but in return need to provide information on things like their age, gender, current health issues, current medication, etc. so that they can be auto-excluded from a study if necessary, or (maybe even better) still included, but analysed at the end of the study in some kind of multivariate test.
It would automatically assign participants to control and different experiment groups, and every week (or whatever) the participants would log in, and submit data... In other words, all data collection happens within the app. In my example above, maybe their exercise routine and a mood rating.
At the end of the study, the platform then figures out things like statistical significance for you automatically and even partially generates the results and conclusion sections of your paper (maybe using GPT3), and publishes it. Studies where no correlations were found would be automatically published too.
The 'hook' (aside from doing a nice thing to benefit humanity) is that everyone who participates in the study has the option to get mentioned as a contributor if they want to. All research findings are available for everyone to see, for free in perpetuity, with people strongly encouraged to discuss the results and suggest ideas for future study. The raw data is available for everyone to download.
Is this a crazy idea? I feel like all you'd need is a dedicated and honest core of 10,000 users or so from a fairly wide cross section of the population and you could probably discover a LOT of really amazing things, and the nice thing is so many people have various fitness trackers now, you could get in some really interesting quantitative data.
tl;dr: If you have even a hint of 'quantified self' type curiousity (like me), rather than posting half-assed n=1 type stuff on Reddit, why don't we join forces and created a free and open source platform to transform a lot of anecdotes into actual statistically significant findings?
If anyone's interested, I'll start up a discord.