Skip to content
← Back to Rove

THE LAB NOTEBOOK / Rove

Little updates.
Big plot twists.

New tricks. Fewer hiccups. The occasional “how did we live without this?” Here’s what’s happening with Rove.

4 releases21 improvements & fixesSTILL TINKERING. ALWAYS.
CORTEX LABS
FIELD NOTES

LATEST CHAPTER

v0.1.3November 5, 2026
Rove
a little better, one version at a time.

More data, clearer choices.

7 things to get excited about

  • Open more NEXUS state-label exports and inspect labels, constant characters and uncertainty before running. Complete state labels can define the character alphabet when you explicitly choose them. True polymorphism remains unsupported for inference.
  • Analyze aligned DNA from FASTA or NEXUS with JC69, or combine one morphological matrix with one DNA alignment. Review taxon matching and choose a fixed or estimated morphology-to-DNA rate ratio. The combined model shares its tree, clock and gamma shape; HKY and GTR are not available yet.
  • Inspect the rooting split and the clock-sigma prior curves. Yule is now an optional pure-birth prior for contemporaneous species, with a configurable birth-rate bound. Birth–death remains the default, and fossil dating keeps FBD.
  • Review sampling, agreement, trees and predictive evidence together in History. Paired posterior predictive plots help you inspect discrepancies. Export supporting files with a manifest and choose whether to include the original matrix.
  • An optional joint topology/rate proposal is available for undated relaxed-clock analyses, with weight zero by default. This is partial ORC coverage; the full BEAST operator suite and comparative performance validation remain pending.
  • Reopen the original thank-you letter from Help → A thank-you to Thiago, just as you can revisit the tour and What’s New.
  • Keep the previous app to resume checkpoints created by its engine. Completed results remain in History. Existing 0.1.2 installations can receive the signed update; 0.1.1 needs one manual DMG installation.

A little more room to explore.

5 things to get excited about

  • Choose intermediate sampling presets or enter your own step count and sampling interval. Run up to 16 independent chains and review every choice with the analysis.
  • Set comparison burn-in, execution options and the relaxed-clock sigma prior in the app. An estimate of time remaining follows the engine's recent progress; it does not predict convergence.
  • Open external traces, import completed CLI runs with matrix verification, and estimate model evidence directly from Review. Trace diagnostics also accept burn-in as a number of samples.
  • Check for signed updates from the Rove menu or Settings. Installation waits for active calculations to finish, then for your confirmation to relaunch.
  • Upgrading from 0.1.1 needs one manual DMG installation to add the updater. Keep the previous app to resume checkpoints created by its engine; completed results remain available in History.

Trees from anywhere.

3 things to get excited about

  • Open a NEXUS tree sample from BEAST 2 or another program from the welcome screen, the sidebar, File → Summarize Tree File… or the matrix import, which recognizes it instead of reporting an error. Summaries live in a folder named after the file, with its path and SHA-256 on record, and History lists it as a tree file with the explorer.
  • The summary sheet fits the window: no truncated text, whole summary cards, the file named at the top, and a footer that stays reachable.
  • The engine's rove summarize reads tree files written by BEAST 2, which carry a zero-length root stem and a rate on the root; the root rate is dropped.

Your first field notes.

6 things to get excited about

  • First notarized beta of the native app, for Apple silicon Macs on macOS 14 or newer. Open the disk image, drag Rove to Applications and launch it; access is by invitation.
  • Import a NEXUS matrix, choose an outgroup and monophyly constraints, and run the bundled engine with the strict or relaxed clock, Mk, Mkv or Mkparsinf, fossil tip dating with a fixed or estimated FBD process, and one, two or four chains that can stop on diagnostic thresholds, pause and resume.
  • The advanced sampling options are in the app: priors only, stopping thresholds, checkpoint interval, proposal weights and starting step sizes with the two evaluated schedules as presets, and tuning off.
  • Summarize posterior trees into an annotated MCC tree and explore it; read trace diagnostics with ESS, R-hat and clade frequencies.
  • Compare the chains of a group or independent runs from History, run posterior predictive checks, and estimate marginal likelihoods with log Bayes factors between runs of the same matrix. Every number shown is the engine's own; the app only subtracts two estimates and combines their errors.
  • History keeps each run with its setup, record and derived files. English and Brazilian Portuguese.

You’ve reached the beginning of the story.

We’re still writing the good bits.Back to the latest ↑