They gave me a 2-hour take-home to build a Next.js + Tailwind comment system with CRUD, nesting, persistence in IndexedDB (not localStorage), and optional cross-tab sync via BroadcastChannel. In short: display, add, delete, support replies, and keep it stored across refresh. The challenge itself was trivial. It boils down to representing comments as a tree, and the key design decision is whether to model them as a nested tree structure or as a flat list with parent references. I chose the latter, which is the industry best practice used by Reddit, Hacker News, GitHub Discussions, and many others, precisely because it makes CRUD operations simpler and more performant. Despite this, Autarc did not invite me to an interview or even provide proper feedback on my solution. Instead, they dismissed me with a vague remark about "database design," which I can only assume referred to the data structure choice. Since that was the entire point of the exercise, their response shows a lack of understanding of their own testing criteria. If a company cannot recognize that the flat list model is the pragmatic and scalable way to approach this problem, then it suggests they are inept in evaluating technical trade-offs. My impression from the interviews also matched this: bright people, but lacking pragmatism, too quick to fall in love with reinventing the wheel, and too eager to chase shiny projects with fancy tech rather than building things that actually work.