When the AI assistant updates the status of multiple cultures at once, it now batches all the changes into a single tool call and prompts for one approval — instead of issuing one approval per culture. Previously, the assistant would hang mid-task: the first approval would go through, but sibling updates were silently dropped, and subsequent approval prompts never appeared. The new batched update confirms all cultures atomically, making bulk status changes reliable and predictable.

The AI Assistant can now generate reports directly from an attached plate-well picking-list CSV. Attach the file to your message and the assistant will produce a report scoped to the wells listed — making it straightforward to summarize results for a specific selection without manually specifying each well.

The assistant also adapts its format and verbosity to the request type, so report outputs are more structured and complete compared to conversational queries.

A new History tab in the culture sidebar displays a newest-first timeline of every status change, metadata edit, and observation recorded for that culture — including who made the change and when. You can scroll back through a culture's full history without leaving the plate view, with infinite scroll loading older events on demand.

This was a frequently requested feature from customers tracking long-running cultures across multiple scans and team members. It is currently available on request — contact your account team to enable it.

Classification well filters now include a confidence threshold slider for each label. Set a minimum confidence percentage per label and only wells where the model's prediction meets that threshold will appear in the filtered view and well navigation.

This is especially useful for workflows where you want to review borderline predictions separately — for example, showing all colony-labeled wells alongside empty wells where the model had low confidence, without being flooded by high-confidence negatives.