Release Notes - May 2026

Below are summarized product changes released during May 2026.

Highlights by Theme

Project attachments now appear in the discussion feed when they are added, and they are also available in a searchable panel.

Kohai update

Kohai can now search PubMed for papers and add them to your project as References, including PDFs when available. The list also shows citation counts to help you quickly identify the most referenced papers.

Categorical parameters in the simple solve panel now use a drop-down list instead of a free-text field, which makes overrides easier and less error-prone.

Computational Model update

Solver handling for infinite values in root conditions is more robust, which should allow more models to run successfully.
You can now export a model to a standalone R script for use with nlmixr and RxSolve.
The export supports most model equations, but not events or DDE/DAE models.
Discontinuities are now handled more smoothly when AddAuxiliaryEvents is set to discontinuityEvents.
More models can now be solved, and many models run faster.
Some models that previously worked with discontinuityEvents=WarnOnly may now fail. Users should migrate to AddAuxiliaryEvents.
Discontinuities in delayed differential equations, such as those using the delay function, are now handled more robustly.
The maximum number of time points in simple solve has been increased from 100 to 1000.
Julia exports now support time-based and root-based events, and exported plots include names for the different time series.
Errors related to compartment volume are now shown directly under the volume field in the model editor.

Computational Model update

Protocols with invalid tmax or tstep overrides now fail sanity checks instead of causing a crash.
If all data tables have null weights and there are no active objectives in the advanced outputs, the system now shows that a fitness function cannot be computed.
The default trial duration has been shortened from 2 years to 1 month, and the default trial step size has been changed from 1 week to 1 day.
Additional sanity checks now apply to protocol time overrides:
- tmax must be greater than the base period tmin.
- tstep must be greater than (overridden_tmax - tmin) of the base period.
- tmax overrides now apply only to the base period when the new value is greater than the base period tmax; otherwise, they apply to affected periods and discard periods that no longer fit.
- tstep overrides now apply only to the base period.

Computational Model update

The timeseries visualization now includes a “Number of time points in timeseries” option in the right panel. It defaults to 100, is capped at 1000, and is saved with the view.

Trial Viz update

Solver handling for infinite values in root conditions is more robust, which should allow more models to run successfully.
The bound scaling descriptor modal no longer duplicates obsId values when the same observation is defined on multiple protocol arms.

Solver handling for infinite values in root conditions is more robust, which should allow more models to run successfully.
Calibration now shows an error when all data tables have null weights.
Calibration now requires at least one data table with a non-null weight or an active objective.

Datatable update

Observations outside the simulation range are now ignored with a warning instead of causing a scoring evaluation error.

The bound scaling descriptor modal no longer duplicates obsId values when the same observation is defined on multiple protocol arms.

Datatable update

Data table wrangling has been refactored with improved code generation and execution.

Datatable update

Data tables with null weights can now be used for visualization in calibration.
The data wrangling generation workflow now has more focused code generation, clearer progress updates, better retry and self-fixing behavior, and reduced data exposure in prompts.