Dexter

An open-source autonomous financial research agent built with TypeScript + Bun. 22.8k Stars. Excellent Skills framework and self-validation mechanism, but strictly finance-only. Good for investors automating earnings analysis and DCF valuations, and for developers studying autonomous agent architecture. Single-maintainer risk is real.

R verification layer launched on six statistics tools

• Every result on T-Test, ANOVA, Correlation, Descriptive Statistics, Regression, and PCA can now be re-computed by a real R runtime running in your browser — one-click consent, no server upload
• Distribution shape (skewness + excess kurtosis) added to Descriptive Statistics
• Loading overlay now follows the dark theme; cancel button truly aborts in-flight R calls; storage-quota warning before opt-in

ANOVA

• R verification: every result can be independently re-computed by R running in your browser (one-click opt-in, no server upload)
• Group comparison chart with ±SD error bars and stacked significance brackets across 3–6 groups
• Dunn pairwise z formula realigned to R's per-pair Mann-Whitney form; all 7 prior parity gaps closed

Correlation

• R verification: every result can be independently re-computed by R running in your browser (one-click opt-in, no server upload)
• Publication-style scatter plot with least-squares regression line and inline `r · p · n` annotation
• Min-sample-size guidance when X / Y have fewer than 3 paired values

Descriptive Statistics

• R verification: every result can be independently re-computed by R running in your browser (one-click opt-in, no server upload)
• Distribution Shape section added: skewness and excess kurtosis (matches R `moments::skewness` / `e1071::kurtosis`)
• Min-sample-size guidance when only one value is entered
• Documentation note added clarifying Q1 / Q3 use linear interpolation (R `quantile(type=7)` / NumPy linear)