Track A: the coding round

Staff and TL loops still have coding rounds. Google, Meta, Stripe, Databricks, Uber and most AI labs keep at least one, and the candidates who fail them are usually the ones who assumed seniority made them optional. The bar is not harder problems than a senior loop gets. It is cleaner code, better tests, and no flailing.

That makes this chapter two different things at once. The first page is about behaviour: the five things a staff candidate does that a senior candidate does not, which are worth more marks than any single algorithm and are trainable in a way that inventing algorithms under pressure is not. The remaining six pages are the pattern checklist, twelve patterns that between them cover the overwhelming majority of what gets asked, grouped so that patterns which share an idea are taught together rather than as a list to memorise.

Read the first page even if you skip the rest. The five behaviours (contract before algorithm, tests unprompted, complexity stated before coding, an answer ready for "make it production", and restraint about abstraction) are where experienced candidates actually lose these rounds.

Every page here carries runnable code, included verbatim from the section's code/ directory, and every output block on these pages was produced by running it. The volume target from the source is about 70 to 90 problems total at roughly 20 percent easy, 65 percent medium and 15 percent hard, not the 400-problem grind. The drill format is 35 minutes, timed, out loud, in a plain editor, with one session a week recorded and watched back.

What this chapter covers

Source: §4.