Lecture 7

AI in the Stack

You've been using AI to build all semester. Now: AI as an ingredient inside the products you design.

Hour 1 — Models as materials

  • LLMs demystified: prediction machines, not databases — why they hallucinate
  • The AI API call: it's just another API (prompt in, text out, tokens as billing)
  • What models can do well (language, classification, generation) and badly (facts, math, consistency)
  • Prompts as design artifacts: system prompts are product decisions
  • The current landscape: models, wrappers, agents — telling substance from hype
  • Designing AI features responsibly: failure states, trust, disclosure

Hour 2 — Studio

  • Exercise: Exercise - Add an AI FeatureExercise - Add an AI Feature
    Put a model inside your product

    Add one genuinely useful AI feature to something you've already built this semester. The bar: it must be better because it's AI, not AI for its own sake.

    Ideas


    ...
    — add one genuinely useful AI feature to something you've built this semester
  • Crit: is this feature better because it's AI, or is it AI for AI's sake?

Homework

  • Start thinking about final project ideas — teams and topics get decided in Lecture 10Lecture 10
    How Software Gets Built

    The people part: how ideas become products, and how teams avoid chaos.

    Hour 1 — The lifecycle

    The SDLC: idea → requirements → design → build → test → ship → maintain
    ...