A high bar on code quality and product judgement, with a smaller loop that leaves nowhere to hide.
CRED runs a deliberately small loop with a high bar per round, which means a weak round is harder to offset than at companies running seven interviews. The code quality expectation is unusually explicit: naming, structure, error handling and testability are scored, not just whether the program works.
Product judgement shows up in technical rounds rather than in a separate product interview. Expect to be asked why a feature should exist, what you would cut to ship it sooner, and how you would tell whether it worked. Answering purely as an implementer, without engaging the product question, reads as a gap here in a way it would not at a larger company.
90 min
Working code with a visible quality bar: structure, naming, error paths, tests.
60 min
Defend and extend your implementation under new requirements.
60 min
Depth over breadth - expect sustained follow-up on one component.
45 min
Craft, ownership, and judgement about what is worth building.
Explicit and scored. Meaningful names, clear separation of concerns, handled error paths, and code an interviewer can extend without a rewrite. Finishing more features with tangled code is a worse outcome than finishing fewer cleanly.
Fewer than most big tech loops, with a higher bar per round. There is less room to recover from one weak interview, so the build round in particular deserves disproportionate preparation.
Engineers are expected to have an opinion on what is worth building. Being able to say why a feature matters, what you would cut, and how you would measure success is part of the technical signal rather than a separate interview.
Practise the rounds CRED actually runs, and get scored on where you stand.