Roast My Day

What is Roast My Day?

Most productivity apps lie to you. They tell you every day is a chance to crush your goals and live your best life. Roast My Day tells you the truth - but makes you laugh about it.

You log what you actually did today. The app roasts you for it. Then it finds the hidden win buried in your chaos and gives you one real thing to do tomorrow.

It's part comedy, part coaching, part mirror.

Inspiration

I wanted to build something that felt genuinely human - not another task manager or wellness tracker. We all have days where we negotiate with ourselves for four hours before doing anything, eat takeout three days in a row, and somehow still rate ourselves a 5/10. That deserves to be celebrated, not optimized away.

The roast format felt perfect because humor is disarming. It lowers your defenses so the real insight can land.

How I Built It

I built the entire app using MeDo through natural language conversations - no manual coding.

The core of the app is a carefully structured prompt engine with four distinct roast personalities:

• Sarcastic Bestie - loving but ruthless
• Disappointed Coach - baffled by your choices but believes in you
• Corporate HR Gone Rogue - office speak completely unhinged
• Late Night Host - big energy, audience-ready punchlines

Each roast follows a three-part structure: the ROAST (sharp and specific), THE TWIST (reframing a failure as an accidental win), and TOMORROW'S EDGE (one cheeky action to take the next day).

I used MeDo's multi-turn chat to iterate the tone across dozens of test inputs until the roasts felt genuinely funny rather than generic. The share card feature - which auto-generates a screenshot-ready image of your best roast line - was built entirely through natural language instructions to MeDo, with no manual design work.

Challenges

Getting the tone right was the hardest part. Too mean and it feels cruel. Too soft and it's not a roast - it's just a compliment. The line between "painfully accurate" and "actually upsetting" is thin, and I spent a lot of prompt iterations finding it.

The other challenge was the share card. Getting MeDo to auto-select the single funniest line from a multi-part roast - rather than just the first sentence - required some creative prompt structuring.

What I Learned

MeDo's strength is in how fast you can go from concept to working app. The bottleneck is not building - it's deciding. Every feature I was unsure about, I could test in minutes rather than days. That completely changes how you make product decisions.

I also learned that the prompt IS the product. The difference between a good roast and a great one is entirely in how precisely you define tone, structure, and constraints for the AI.

Built With

• medo