kuper

500 million smart speakers have been sold. (WILD!) NONE can run AI agents. NONE let you choose your model. NONE are open. BROKEN! We’re building the open alternative FROM SCRATCH! - A new OS - Open source abilities. - Open hardware. Jailbroken. 🤫 Build anything.

Introducing GPT-Realtime-2 in the API: our most intelligent voice model yet, bringing GPT-5-class reasoning to voice agents. Voice agents are now real-time collaborators that can listen, reason, and solve complex problems as conversations unfold. Now available in the API alongside streaming models GPT-Realtime-Translate and GPT-Realtime-Whisper — a new set of audio capabilities for the next generation of voice interfaces.

we crossed $20k/mo in 83 days instead of everything that went right, here are the mistakes we made (so you don't make them!) : 1. During the early days, we tried too many marketing channels due to fomo. - we literally tried SEO, pSEO, meta ads, google keyword ads, tiktok video, and cold email all within 4 weeks lol - even if the channel turns out to not work, we think it's worth committing and testing for at least a month to get a good understanding of it - this is especially true because there are things you start to see the more you try and bear through it. trying is different from reading about it. 2. Don't spend more than a few days on a new feature or experiment - this pertains to both product decisions and marketing decisions - honestly i get easily excited about a new idea, and as a founder you probably do too - but it's always the same pattern again and again: i get super excited about the "next big thing", spend 2 weeks on it, and totally flunk - no matter how promising the idea seems, spend max 5 days on it to release and test. if it takes more than that, it's probably too big for experimenting. 3. Stop copying and listen - with 100% accuracy, all of the features that failed were 1) ideas that just magically came from imagination and 2) ideas that were inspired from other products - ofc inspiration is good, but you HAVE TO verify it with your users. do your users really need that feature? - keep in mind the product is for your user, not for you. if you're trying to build a serious product for a real business and not a play project, this is important - on the surface, every idea seems like a good idea. but adding a new feature = more complexity. be very strict about why that feature needs to be in your product - i think this is especially true for consumer products. simplicity is key. 4. Once you start making revenue, spend more where it's needed - 2 weeks ago, our app just flat out stopped working because our servers maxed out. CPU usage went over 100%. - it was literally 1am, users emailing us nonstop, requesting refunds. i panicked. - this was a stupid mistake. a small buffer would have saved us from a bunch of users having a bad experience these are the 4 mistakes on the top of my head but as I remember more, I'll add it to the thread

Vibecoding an app to talk with my Vietnamese Grab driver

Today we’re introducing Scribe v2: the most accurate transcription model ever released. While Scribe v2 Realtime is optimized for ultra low latency and agents use cases, Scribe v2 is built for batch transcription, subtitling, and captioning at scale.