How it started
I was going through YC's Requests for Startups list one night and got stuck on "AI hasn't had its multiplayer moment yet." I already had this collaborative app idea sitting around that I hadn't seen anyone actually build. Had a free weekend and a Claude Pro subscription, so instead of thinking about it more I just built it. About 4.5 days of actual work. Also, being the only software engineer in the family, the "tech wiz" everyone hits up when they want something built, I've basically got a low effort, high output, repeatable response now.
The hill i died on (but actually lived)
Everyone's shipping "prompt to code" right now, that part's not interesting by itself. What I wasn't willing to cut, even on a weekend deadline, was building the whole thing around sharing instead of exporting. Every app gets a link the second it exists, and anyone with that link can hop in live with you. Basically the AI-app version of Google Docs, whoever's in there is looking at and messing with the exact same live version. I looked into what else is out there after the fact, turns out Lovable and Replit both do real-time collaboration too, but theirs is fork-and-merge, everyone gets their own draft or task and it gets folded in once someone accepts it. Mine skips that entirely, whoever prompts next, everyone just watches it land, no accept, no merge. There's version history if you want to go back, but what you're actually looking at day to day is the shared thing, not your own copy of it.
Corners i happily cut
Each app runs in its own sandbox with its own storage. Under time pressure I just went with defaults (React, Vite, SQLite, Node) instead of trying to support every stack on day one, though you can prompt your way into swapping any of that out later. Also skipped building any kind of onboarding, the app's supposed to explain itself from the prompt box, sink or swim.
What actually surprised me
I threw in some of my own API keys and told it to pull from Unsplash, and that made things way more interesting than I expected. Wasn't part of the plan, just fell out of wanting the sandbox to actually be open instead of locked to some template.
Not spilling this part yet
Keeping the actual sync/collaboration internals out of this post for now, partly competitive, partly because I'd rather talk about it once it's held up to more use. Happy to answer questions in the comments at a high level though.
Receipts
A product page with a real Stripe checkout: https://sharefreeform.com/a/d0e55aff-8aec-45dc-9874-5dc1b943afa8?utm_source=devto&utm_medium=content&utm_campaign=launch-2026-09
A record player that plays MIDI and iTunes music: https://sharefreeform.com/a/661d308e-1751-49c3-bd68-f7a58c051a03?utm_source=devto&utm_medium=content&utm_campaign=launch-2026-09
My next goal: total chaos
I want to pool a bunch of agents onto one build and let them scramble to build your app as fast as possible without stepping on each other, kind of like the Hunger Games but for code. No idea yet if that's actually a good idea or just chaos for chaos's sake, but that's where my brain's at next.
Where things stand
Small right now, a handful of early users (mostly friends), just started sharing this more broadly. Try it at https://sharefreeform.com/?utm_source=devto&utm_medium=content&utm_campaign=launch-2026-09, and if you hit a bug or build something worth showing off, email me at hello@sharefreeform.com, and I read every one.
Top comments (0)