A software engineer's responsibilities have shifted significantly in the era of LLM coding agents. We're spending much more time reviewing code than writing it. Judging software quality is more important than ever!
This repository is the output of GitHub Copilot circa December 2025. It was given an programming problem we, Path Robotics, used to give candidates. We're asking you, the candidate, to review the code and provide feedback. Specifically, we're asking ⬇️
- Pretend this repository is a pull request made by someone you work with.
- It's your job to give constructive feedback to improve the solution.
- Everything is fair game! Consider things like:
- Use of data structures
- Adherence to software design principles
- Flow control
- Testing
- Important: There is no "right answer" we're expecting.
- The goal here is to frame a conversation about software quality, not to find some hidden bug or "gotcha".
The agent was asked to write a maze solver that could find a path from a start point (S) to an end point (E) in a 2 dimensional maze. Here's an example ⬇️
#########################
S # # # #
# ### # # ##### ####### #
# # # # #
### # ############# # # #
# # # # # # # # #
# # ### # # ##### # # # #
# # # # # # # #
# ########### ####### # #
# # # # # # #
# # # # # ##### # ##### #
# # # # # # #
##### # ##### ##### #####
# # # # # # #
# ### # ### # # # # # # #
# # # # # # # # # #
# ##### # ##### ####### #
# # E
#########################
It was given a single constraint: use python. Everything else was left up as a judgement call