Changelog - What we shipped

Honest side-by-side comparisons: Xaiku vs ChatGPT, Anyword, Jasper, and Optimizely.

Six ready-to-use templates: homepage hero, email subject line, pricing CTA, ad copy, landing page headline, product description.

Structured product page at /ai for search engines and language models. Includes SoftwareApplication schema.

Hero, problem, how-it-works, features, and differentiator sections rewritten. Shorter sentences, clearer benefits.

Free-first positioning, tier descriptions, per-CTA friction reducers, and "How impressions work" explainer.

First-person founder letter. New values section: Built by a maker, Proudly independent, Generous by default.

Paying users now see contextual upgrade suggestions — not hard gates, just helpful prompts when a higher tier unlocks something relevant.

New grid layout with winner card, projected impact, computed insights engine, and guided setup for empty states.

Fixed sidebar animation, chart sizing, and dark mode theming on the analysis page.

Every variant gets a stable letter (A, B, C…) and color. Consistent across cards, charts, integration table, and analysis.

Floating glass panel on the analysis page with spring animations. Quick reference for which variant is which.

Single cookie-based preference system for sidebar state and other UI settings. No more localStorage drift.

Eliminated the old "challenger" terminology. Every UI label, tooltip, and API response now says "variant."

Analytics events are now aggregated as they arrive — no more cron jobs. Faster dashboards, simpler architecture.

Connection check now returns server status instead of blindly mutating it.

Automatic insight generation from experiment data. Winner card, projected impact with extrapolation, guided setup checklist.

Observed power calculation using erf-based CDF. Bonferroni correction for multi-variant comparisons.

All editing tools — casing, normalize, find & replace, append/prepend, checks — removed from the subscription gate.