What I Built
Daily Briefing is a personalized AI-powered morning dashboard that gives you everything you need to start your day, in one place. Instead of opening five different apps and tabs every morning, you open Daily Briefing and get your local weather with a season-aware tip, live stock prices with AI market commentary, top global news rewritten into clean summaries, a word of the day, a joke, a dog of the day with a breed-specific fun fact, and three historical events that happened on this exact date in history.
The entire content layer of the app is generated by Gemma 4. Raw data feeds (weather, stocks, news headlines, a dog image) are fetched first, then passed to Gemma in a single prompt that produces all the human-readable content at once. Nothing the user actually reads is hardcoded or pulled from a static source.
The app is also timezone-aware. It detects the user's location coordinates, maps them to their local timezone, and uses that timezone for the greeting, the date used in the AI prompt, and the historical events. A user in Lagos and a user in Tokyo get a briefing that is correct for their local date and time of day.
Demo
Live site: https://ifeanyi-brief.netlify.app
Code
Github Repo: https://github.com/MasterIfeanyi/daily-briefs
How I Used Gemma 4
We used the Gemma 4 26B MoE model (google/gemma-4-26b-a4b-it:free) accessed via OpenRouter.
We chose the 26B Mixture-of-Experts variant specifically because of the nature of our task. Every morning, a single prompt asks the model to do several different things simultaneously: rewrite news headlines, analyze market sentiment, generate a vocabulary word with etymology, write a joke, produce breed-specific trivia about a dog, write a weather tip that accounts for the user's local season and climate, and recall three verified historical events for today's date. That is a high-breadth reasoning task, not a deep single-topic task. The MoE architecture is designed exactly for this: routing different parts of a complex prompt to specialized experts efficiently, which means we get broad, high-quality output without sacrificing speed or burning through tokens unnecessarily.
The 128K context window also gave us confidence to pass the full raw news data array and detailed stock information directly into the prompt without worrying about truncation.
Gemma 4 powers the following features directly:
- News summaries - raw headlines are rewritten into clean, readable stories
- Stock commentary - one sentence of market context based on live price and change data
- Word of the Day - a real, interesting English word with definition, part of speech, example sentence, and language origin
- Joke of the Day - a clean, family-friendly joke generated fresh each day
- Dog fun fact - a specific, surprising fact about whatever breed the dog API returned that day, not a generic placeholder
- Weather tip - a practical, friendly sentence about what to expect based on the user's actual location, current weather data, and the typical season for that region
- On This Day - three verified historical events that happened on the user's local calendar date, across different centuries and topics
All seven of these replace what would otherwise have been either separate third-party API calls or static hardcoded content. The result is a website where the core of what you read every morning is genuinely AI-generated.
Team:
Built by @ifeanyichima and @rae_codes