Every morning, before I do anything else, I check the weather. Then I open a news app. Then my calendar. Then a finance app to see what happened overnight. Sometimes I check email too, which is always a mistake.
By the time I have pieced together a picture of the day ahead, ten or fifteen minutes have gone by and I have opened half a dozen apps. None of them talk to each other. None of them know what I actually care about. I end up skimming headlines I do not need and missing things that matter.
That is the problem dawncast is trying to solve.
One briefing, every morning
The idea is straightforward. You tell dawncast what you care about: which news topics, which stocks, your calendar, your city for weather. Every morning, it pulls all of that together and gives you a single briefing. Text or audio, your choice. The whole thing takes under three minutes.
The briefing is generated by AI (Claude, specifically). It is not just a list of headlines stitched together. The model reads the source material and writes a coherent summary in a consistent voice. If something important happened overnight, it leads with that. If your day is packed with meetings, it tells you. If the market moved on something relevant to your watchlist, that is in there too.
Why the morning?
I picked the morning because it is the one part of the day that almost everyone shares. You wake up. You are slightly behind on what happened while you were asleep. You need to get oriented before the day starts pulling you in ten directions.
Most apps fight for your attention throughout the day. They want you scrolling, tapping, coming back for more. Dawncast is the opposite. Get in, get caught up, get out. The best outcome is that you close the app in three minutes and do not think about it again until tomorrow.
That is a weird thing to optimize for as a product builder. Retention usually means maximizing time spent in the app. But I think there is a version of retention that works differently: you come back every single day because the thing is genuinely useful, even though each session is short. Your alarm clock has great daily retention and you spend about two seconds with it.
What is working so far
The core loop is built. You set your preferences, the backend pulls data from several sources overnight, Claude generates the briefing, and it is ready when you wake up. The audio version uses text-to-speech, and it sounds better than I expected.
The hardest part has been editorial judgment. Not the technical side of summarization, but figuring out what deserves to be in a three-minute briefing and what does not. If you include too much, the briefing feels like drinking from a fire hose. Too little and it feels empty. I have been tuning the prompts and source selection a lot, and it is getting closer.
Now that it is live
dawncast just launched on iOS. The last few weeks before launch were all polish: onboarding, how preferences are surfaced, the audio playback experience. I wanted the first version to feel finished, not like a beta that shipped early.
Now the work shifts to listening. Watching how people use it, what they skip, what they ask for. Android is next. I will write more about the technical architecture, the AI pipeline, and the design decisions as I go. For now, I am just glad mornings got a little better.