Unpopular opinion: The social 'following' features are a distraction. Keep it a tool.

I've been using SciSpace (formerly Typeset) for several months now to navigate dense CS and systems papers. The core value proposition is excellent: parsing PDFs, summarizing complex methodologies, and answering specific technical questions saves immense time.

However, I've noticed a significant latency issue, and it's not in their API. It's in the UI/UX. The recent push into social features—following researchers, a feed of "activity," and network building—introduces cognitive overhead. My workflow is interrupted by notifications unrelated to my immediate research query. It feels like the product team is optimizing for "engagement metrics" (user sessions, time-on-platform) rather than "utility metrics" (query-to-insight speed, accuracy).

From a systems design perspective, every feature has a cost:

* **Database load:** Social graphs require complex joins (`User` > `Follows` > `Activity`), increasing query latency for what should be simple document Q&A.
* **Cache invalidation:** A user's feed is highly personalized and volatile, making it difficult to cache effectively compared to static, shared research content.
* **API complexity:** The backend must now service two distinct patterns: low-latency, read-heavy document interactions and write-heavy social interactions. These are often at odds.

I'd prefer a focused tool. The ideal would be a clean interface with powerful querying, perhaps even a CLI option for scriptable interactions. The social layer should be optional and minimal—a simple bookmarking system for papers or authors would suffice, stored locally if needed.

Are others experiencing this drift? I chose SciSpace as a performance tool for my research, not a social network. Keeping these concerns separate seems architecturally sound and user-centric.

-- latency