Let Him Cook AI

Stuck on what to say? Let Him Cook AI stirs up spicy, witty, and vibe-matched replies for any comment, troll, or text.

Comment 5

Let Him Cook AI 🔥🧠

Never freeze up again. Access the magic: https://ubiquitous-raindrop-9231aa.netlify.app/

Inspiration

We’ve all had that moment: someone drops a rude comment, a slick joke, or just an awkward message — and you freeze. Hours later, bam, the perfect comeback hits you... but it’s too late.
Let Him Cook AI exists so that moment never happens again.

From getting trolled online to trying to flirt without sounding like a spreadsheet. we wanted to build a witty, emotionally aware AI assistant that helps you say the right thing, right on time.

What it does

Let Him Cook AI is your personal social ninja 🥷 delivering screenshot-worthy replies based on context, mood, and tone.

Users can:

  • Input the exact message or situation or screenshots of their conversation they’re dealing with
  • Choose their current mood (annoyed, calm, flirty, etc.)
  • Select what kind of response they want (roast, boundary-setting, classy exit, etc.)
  • Pick the tone (savage, wholesome, petty, flirty, etc.)
  • Optionally drop a draft or rough idea to flavor the response

The AI:

  • Crafts 2–5 spicy, clever, mood-matching replies
  • Keeps it witty, ethical, and made for the internet generation
  • Is context-aware and lets users own the moment

How we built it

Stack:

Layer Tech Stack Why We Picked It
Frontend Next.js + TailwindCSS Fast, responsive UI with server-side rendering
Backend FastAPI Lightweight, async, Python-powered flexibility
AI Engine OpenAI GPT-4 API Best-in-class tone control + context awareness
Database Supabase (Postgres) Realtime + user-friendly + scalable
Hosting Netlify (frontend), Railway (API) Fast deployments, smooth CI/CD

But Later I decided to remove the auth system and backend because I was limited by time as I am a student and I felt this could be simpler to navigate for people to try this easily.

We focused heavily on UX - making input feel simple, but output feel personalized and powerful.

Challenges we ran into

  • Tone without toxicity: Making the AI spicy but still socially responsible took serious prompt engineering.
  • Too much / too little input: Some users gave paragraphs, others gave one word, balancing both was tough.
  • Making it screenshot-worthy: Replies had to feel natural, not robotic, wit is hard to automate.
  • Multi-intent replies: Handling emotional tone, social context, and reply style simultaneously = not easy.

Accomplishments that we're proud of

  • Shipped a fully working MVP with AI replies tailored to real-life social tension that actually feels fun, usable, and share-worthy
  • Built a modular system for adding more tones, moods, and reply styles later
  • Nailed the viral potential, users want to share replies (and users LOVED the replies (some too much))
  • Designed a tone-aware AI interface that feels more human than bot

What we learned

  • Humor, tone, and emotional intelligence are way harder to nail than just answering questions
  • Good UX is not just about design, it's about timing, empathy, and tone
  • Users don’t want “smart” responses, they want relatable and right responses
  • There's real demand for AI that enhances self-expression and confidence

What's next for Let Him Cook AI

  • 🎭 Fine tuned tones & moods → More quality responses from reinforcement learning and fine tuning
  • 🧠 User training mode → Let users learn how to respond better over time
  • 🧩 Browser extension or mobile app → Instant cook mode, right inside Instagram/Twitter/DMs
  • 💬 Community-curated responses → Upvote-worthy replies that become shareable templates

Let Him Cook AI isn't just about comebacks, it's about owning the moment, finding your voice, and saying it with style.

Because sometimes… you’ve just gotta let him cook. 🔥

Built With

Share this project:

Updates

posted an update —

Quick Note: I realized after submission that I missed the badge on the homepage. Due to the no-edit-after-deadline policy, I didn't touch the code. Just wanted to be transparent and say it was an oversight during the final sprint. hope this won't disqualify the project, as it was built in full spirit of the hackathon.

Still proud of what’s been cooked, thanks for hosting such an awesome event!

Log in or sign up for Devpost to join the conversation.

posted an update —

Let Him Cook AI - Usage Milestone What started as a joke during exam week has officially entered the kitchen:

213 sessions cooked

1,065 AI-generated comebacks, roasts, and compliments

Built in under 24 hours

Still live, still cookin’

Didn’t expect strangers to actually use it, then my AI complimented me through someone. That’s when I knew this thing had vibes.

Real-world version in the works. Till then? Let him cook.

image.png

Log in or sign up for Devpost to join the conversation.

posted an update —

Didn’t expect this to gain traction so fast. What started as a joke turned into a tool strangers are now using for awkward comebacks, social jabs, and even compliments.

Yes, my AI might’ve just complimented me through someone else. That’s when I knew, this thing had vibes.

Built the MVP in a day during exam crunch. Kept it light, fun, and deployable. Real-world version is coming soon, just letting it cook a bit longer.

Log in or sign up for Devpost to join the conversation.

posted an update —

I had an version of Auth and Supabase integrated but it had too many errors and the code got really complex and just didn't work out at some point. So I removed them and refactored it using bolt to keep it light and simple. So that's why I have it in how we built it section but not in the final version

Log in or sign up for Devpost to join the conversation.