Argument imports now allow more headroom for long-text segmentation based on your account tier (Tier 2 gets a higher cap than Tier 1). Long-text imports that previously hit an internal cap mid-generation now have room to finish cleanly.

Several import-stability fixes also land in this release:

• Imports of arguments containing self-evident / axiomatic premises (e.g. "by definition" statements) now persist correctly instead of failing with a database error mid-run.
• Refreshing a task-status page while an import is still running no longer surfaces database errors in the server log.
• The task-status page no longer logs a React hydration warning on first paint.
• Imports no longer get stuck on the "Building your argument…" spinner when an unexpected error occurs mid-pipeline — the task transitions to a clear failed state so the UI can react.
• Imports of documents containing embedded NUL bytes (common in PDF-extracted text) no longer cause silent observability write failures.
• The pipeline-status view now shows all upcoming stages from the start, not just the ones that have already begun, so you can see the full pipeline shape at a glance.
• The argument view now reliably clears the "Building your argument…" spinner after a completed import's data refresh finishes — even if the refresh itself fails — so you can never get permanently stuck on the spinner.