They are very proud of their fully automated process for the first step of the interview and even separately blogged about it. They were sure their approach was less stressful and would avoid false negatives. They talked about making API calls and in between processing some information and that I should understand JSON and basic data structures.
After 2 days of preparing myself for what I thought could be involved in the actual assessment, I felt confident that I could handle whatever the problem with speed and grace. I've also been doing dozens of HackerRank problems for weeks and I was excited about the prospects of working at HubSpot in the future.
When I started the assessment, I realized pretty quickly that the entire thing was designed to be a false negative to prevent the need for actual human interaction with candidates due to the 3 hour time limit and the complexity of the actual problem to solve. Instead of determining engineering skill or problem solving skills, they had essentially created a problem to solve that only identifies people that have literally bumped into this exact problem, Googled their interview ahead of time, and had knowledge of the obscure data structures required to solve it with an implementation already prepared.