Frontend Architecture
Putting it together: free-recall review
Retrieval beats re-reading. Each prompt spans several units at once — answer it fully from memory before revealing the model answer, because reconstructing the connection between layers is exactly the senior skill this capstone tests.
Reconstruct the whole track as one cascade — why state shape sits at the bottom, how the fetch graph owns LCP, what a semantic token layer buys, why boundaries are the build-time multiplier, and how to review it all in failure-cost order.
- 01Why does a senior review a frontend bottom-up by failure cost rather than top-down by file structure, and which layers are most vs least expensive to get wrong?
- 02A dashboard janks on every keystroke and the team wants to code-split the charts. Walk through finding the real layer and why their fix fails.
- 03How does the data-fetching layer decide LCP, and what turns a slow page into a fast one?
- 04Explain the three-layer token structure and what each layer buys when a rebrand or dark mode lands.
- 05Why are monorepo boundaries the build-time multiplier, and what two levers cut CI time the most?
- 06When is code-splitting the right tool versus the wrong one, and how do splitting and the build pipeline relate to the layers below them?
If you could reconstruct each answer from memory, you hold the track’s spine: state shape sets the re-render blast radius, the fetch graph owns LCP, a primitive-to-semantic token layer makes a rebrand a value swap, clean monorepo boundaries plus affected-only execution and a remote cache are the build-time multiplier, and code splitting and the pipeline are the cheapest layers to fix last. The recurring move is to read a symptom, drop to the lowest layer that owns it, and fix there — never patch a layer above the cause.