← All Updates

Plate Imaging & AI Assistant Classification Insights

July 12, 2026
Image Classification Insights in AI Assistant

For organizations running image classification, the AI Assistant can now pull up predicted labels and per-well confidence scores directly in conversation — no more digging through raw data to find what needs a second look.

Details

  • Ask for classification results on a specific plate to see labels and confidence per well.
  • Ask for the lowest-confidence hits across your most recent plates to quickly triage what needs review.
  • The feature stays out of the way for orgs without classification data, so it won't clutter the assistant's toolset unless it's relevant.
Row & Column Labels for Plate View

Row letters and column numbers are now shown on the default plate view for all plate formats larger than 6 wells (12, 24, 96, 384, and beyond), regardless of whether the "collapse to occupied wells" view is active. This makes it much faster to identify a specific well's position without counting across the grid.

Behind-the-Scenes Fixes & Tweaks

Large plate reports (like 96-well summaries) now complete reliably instead of failing silently, and tool approvals in the AI Assistant respond faster and more consistently.

Well montage images no longer show doubled features at inconsistent interior seams, thanks to a new global tile-alignment solver.

Well and montage images now preserve the microscope's native aspect ratio instead of being forced into a square, so round wells render as circles rather than stretched ovals.

Image processing pipeline now supports 16-bit source images with proper contrast scaling, so images from compatible instruments display with correct brightness and detail instead of appearing flat or clipped.

Import now reliably detects plate holder types and well counts from nested XML in packaged plate configuration files.

Eliminated intermittent "database is locked" errors that could occur while the packaged desktop app was starting up.

Workcell file sync now removes files that were deleted from the source, keeping the local directory clean and up to date.

Routine template lookups now return clear, actionable messages instead of opaque failures when a routine or template file is missing or malformed.