Notes from shipping AI products

Practical lessons from designing and shipping LLM-powered features, frontend craft, and Web3 product work.

1. Web3 7 min

   Building a DeFi swap UI users don't fear

   How I reduced visual noise in CaveSwap while still exposing the data power users need.

2. Frontend 7 min

   The Figma → Vue handoff I actually use

   Tokens, components, and the small conventions that keep design and code in sync without a process tax.

3. LLM Integration 8 min

   Evaluating RAG retrieval quality

   A practical RAG eval setup — what to measure, how to score, and the dashboard I use to track regressions.

4. AI Product 9 min

   Designing streaming responses in Vue

   A no-flicker streaming chat UI in Nuxt: chunk parsing, smooth typing, and graceful tool-call rendering.