Google has introduced a new "Code Comprehension" round into its software engineering interview loop.

Google has introduced a new "Code Comprehension" round into its software engineering interview loop. First launched as a pilot in May 2026 for junior and mid-level roles (L3 and L4), this round shifts the focus from writing algorithmic solutions from scratch to reading, debugging, and optimising a large, pre-existing codebase with the assistance of Google's AI, Gemini.

The decision follows Google’s internal metrics revealing that roughly 75% of new code at the company is generated by AI and reviewed by human engineers, shifting the day-to-day engineering role toward verification and code safety.

⏱️ Format of the Round

While Google hasn't published an official sheet, candidate experiences and prep platforms outline the general structure:

The Setup: A 60-minute session conducted inside CoderPad.
The Codebase: Candidates are given a production-style, multi-file codebase containing roughly 200 to 500 lines of unfamiliar code.
The Tooling: Google’s Gemini is integrated into the environment as the exclusive, approved AI assistant.

Expected Progression (The 60-Minute Split) [1]

Minutes 0–10 (Read): Navigate the code, explain what modules do out loud, and ask clarifying questions.
Minutes 10–30 (Debug): Formulate independent hypotheses about hidden bugs, use Gemini to verify your theory, fix them, and write test cases.
Minutes 30–45 (Extend): Add a feature smoothly into the existing codebase without rewriting the architecture.
Minutes 45–60 (Optimise): Identify bottlenecks, discuss time/space complexity changes, and detail trade-offs.