Sign in

Week 1Foundations

5 lessons · project on day 6

Week progress0 / 5
  1. Day 1

    AI Engineer

    What an AI engineer actually does, and how this 30-day path gets you there.

    Open
  2. Day 2

    Programming

    The fundamentals that survive contact with non-deterministic systems — typed errors, idempotency, and reasoning about failure.

    Open
  3. Day 3

    JavaScript

    The JavaScript an AI engineer actually needs — async iteration for streaming tokens, AbortController, and backpressure.

    Open
  4. Day 4

    Node.js

    Streams, edge vs. Node runtimes, and timeouts on long generations — the server-side plumbing that makes AI features reliable.

    Open
  5. Day 5

    React

    Rendering partial model output, optimistic state, and streaming UI — the React patterns every AI product interface needs.

    Open

Day 6 project

Prompt Playground

A single-page playground that sends the same prompt to a model at three different temperatures and renders the outputs side by side, with the full request logged underneath. Building it forces you to hold the whole request/response shape in your head — messages, system prompt, sampling parameters, token usage — instead of treating the model as a text box.

Stack

  • TypeScript
  • Node.js
  • React
  • Vercel AI SDK
  • localStorage

The hard parts

  • Streaming three responses concurrently without the UI blocking or interleaving tokens into the wrong column.
  • Keeping the API key server-side — the first instinct is to call the provider from the browser, which leaks the key to anyone who opens devtools.
  • Making the comparison fair: same seed-independent prompt, same max tokens, only temperature varying.

Why it belongs in your portfolio

Shows you understand that model output is a function of parameters you control, not luck. The logged request panel is the part reviewers notice — it proves you can debug an LLM call rather than re-roll it.