The data export issue is a critical operational question that often gets overlooked in these discussions. From a platform management perspective, I've observed that services built around AI-assisted analysis tend to prioritize data ingress over egress. The value is in retaining the organized knowledge base.
I haven't performed a bulk export from SciSpace personally, but typical patterns for similar platforms include offering CSV exports of citation metadata while retaining summaries, notes, and AI-generated content within the ecosystem. Sometimes this is a manual, per-item process. This creates a significant data preservation risk for long-term projects. Your question about bulk export is the right one to ask directly of their support before any long-term commitment.
It underscores a broader principle: any tool that becomes a primary workspace should be evaluated not just on its features, but on its ability to let you leave with your work intact.
Let's keep it constructive
You're right about the technical overhead being a blocker, but calling a local model a "lightweight script" is a bit optimistic. Setting up Ollama, keeping it running, and managing prompts for decent summaries is a part-time job. The monthly subscription isn't just buying queries, it's buying back the weekend you'd spend debugging. That said, if you're already comfortable in a terminal, the lock-in avoidance is a pretty compelling reason to eat the setup cost.
But what about the edge case?