Instacart AI coding interview: 2 candidate reports
Latest report: June 2026 · Updated
What candidates say the round looked like, summarised in our words, each linked to where it came from. Instacart hasn’t confirmed these details unless the source says “Official page”.
read the files, write code, run tests (at least half of 1 reports)
The reports
Online assessment · June 2026
CodeSignal · Python
Full-stack assessment on CodeSignal with two weeks to complete, in 5 rounds. 1: chat with an AI product manager to gather requirements for a search and filter feature in a library app. 2: implement it with Claude Code in the given project, with full test coverage expected; Claude Code was very slow and CodeSignal felt able to lose work, so commit after each feature. 3: a debugging task, a wrong value on a metrics page that turned out to be a FastAPI backend bug in the order of operations. 4: a harder requirements chat for a notification and holds feature. 5: build it with Claude Code. Short on time, the candidate ran two agents at once, they overwrote each other and broke the project.
AI: Built-in AI chat that could read the files, write code, run tests · Task: Mixed · Source: Reddit thread
Online assessment · April 2026
Didn’t pass
CodeSignal
This candidate scored full marks on Instacart’s CodeSignal assessment without using the AI, and was rejected the next day. Replies pointed out that coding isn’t the only signal assessments look at, and other reports describe Instacart’s version as built around using Claude Code, so not using it may itself have counted.