096 - Stash Update

Stash is a local-first text-to-speech app. Paste any article, get a podcast-style audio file, listen on the go. No account, no server, no subscription.

Our first TestFlight testers hit a wall we should have seen coming: the app required a model download on first launch, and one tester on bad 5G simply couldn't use it. That's a build-to-last failure. It also means a model provider sits between the user and the app, and if that host goes away, so does Stash.

So we're cutting every dependency we can. The voice model ships inside the app binary. Web scraping happens on-device instead of through a Cloudflare-style service. Sync runs through iCloud, and the app still works fully offline if iCloud doesn't.

The new model also moved our workload from the GPU (which iOS won't let you run in the background) to the Neural Engine. Result: background processing, cooler phone, less battery, and roughly 10x faster audio generation. A six-minute article converts in a couple of seconds.

Join the waitlist: https://juststash.app