Decision frameworks

Interactive decision trees for the architecture questions you'll face repeatedly. Each branch explains why the question matters — the goal is to internalize the reasoning, not memorize verdicts.

Should I use an LLM?

Decide between deterministic code, classical ML, a fine-tuned small model, or an LLM.

Can the logic be written as deterministic rules with high accuracy (lookup, regex, arithmetic)?

Why this question matters: Deterministic code is cheaper, faster, testable and never hallucinates. If a rule gets you 99% on a labelled sample, an LLM adds cost and variance.