Does it generate good anagrams? Very rarely! Is it fun to use? You bet it is.
Grammy was built quickly as a way to get hands-on with AI tools and re-familiarize myself with Next.js. The MVP took roughly four hours of work to stand up, but as they say, the devil's in the details. Netlify Functions were a bonus when I decided that maybe (just maybe) I should treat my Gemini API key with some dignity.
I learned a lot building this project (maybe I'll blog about it later, who knows) and wrote plent of notes over on the GitHub repo. One major takeaway is just how much fine-tuning is needed to generate reliable outputs using LLMs. Apparently some folks are even calling anagram generation a "cognitive blind spot" in LLMs, which I totally buy given my experience on this project. Super interesting stuff!
Tools Used
- Next.js
- Netlify Functions
- Google Gemini
- HTML / CSS
- V0 / Shadcn/UI
Credits
- Ethan Morrill-Ploum
- Design & code.