v2.9.0: The Waitlist Is Open — Plus Exam Prediction on Pro & Our Biggest Reliability Release

Every product has a moment where it stops being a prototype and starts being something you'd hand to strangers. For Claritii, this release is that moment. v2.9.0 is our biggest reliability release to date — 75+ fixes across the AI pipeline, edge functions, scheduling, and the UI — and it ships with the feature we've been building toward all along: a proper waitlist for launch.

The Waitlist Is Open

Visit the landing page today and you'll meet the new waitlist experience. A full-screen overlay introduces Claritii, captures your email in seconds, and confirms your spot — with a branded welcome email to match. It appears once per session, dismisses with ESC or a click, and every "Log in" and "Get started" button across the marketing site now routes through it.

Every signup lands in a dedicated waitlist database with idempotent upserts (no duplicate entries, ever) and fires a confirmation email through a new edge function. When we open the doors, everyone on the list gets notified first. Owner access is preserved throughout the closed beta, so the app itself keeps getting tested and improved every day.

Exam Prediction Moves to Pro

Since the AI exam prediction engine shipped, it's been gated behind the Max plan. That was a mistake — the feature spec, the features page, and the pricing copy have all described it as a Pro feature from day one. v2.9.0 fixes the mismatch: Pro subscribers now get exam prediction, and the upgrade walls and messaging have been corrected across the app. If you're on Pro and your exams are coming up, your predicted topic list is waiting for you.

Your Difficulty Setting Finally Does Something

For a while now, each subject in the AI Tutor settings has had a difficulty slider. It was saved, stored, and… ignored. The roadmap planner never read it. That's fixed in v2.9.0. The AI session plan now receives your per-subject difficulty ratings and applies a clear decision rule: subjects rated 4–5 get more study time and active-retrieval methods (quizzes, blurting, practice papers); subjects rated 1–2 get faster foundation-building methods (summaries, flashcards). Set mathematics to 5 and the roadmap treats it with the respect it demands.

The Reliability Sweep

The headline fixes are user-facing, but the heart of v2.9.0 is underneath:

Rate limits that actually enforce themselves. We found a response-envelope bug in the rate-limiter's atomic counter RPC — every check read the window as empty, so abuse, load, and quota limits never triggered in production. They do now, and expired counters reset automatically so nobody stays blocked forever. This is the first release where the limits we configure are real limits.

Email infrastructure that works end-to-end. Resend webhook signature verification always failed (the signing secret was encoded wrongly and base64 parts were parsed as hex), so email open and click events were never recorded. Now fixed with timestamp freshness and replay protection. Release announcements require the true service-role token, marketing broadcasts are claimed atomically so two cron runs can't double-send, and Stripe portal plan changes now resolve tiers correctly.

Auth and account flows repaired. Deleting your account now removes the auth row first — no more half-deleted logins that break forever. Email verification and password reset now work for users beyond the first 50 (a pagination bug was silently eating the rest). The onboarding Quick Start no longer writes every subject and schedule block twice. Verify-email's "Continue" button works even without an existing session.

Nothing fails silently anymore. Over 40 handlers across the app — settings saves, onboarding inserts, dashboard dismissals, friend search, Google login, email changes, session completion, and more — previously swallowed errors and did nothing. Every one of them now logs and surfaces the failure to you, so a broken action can never masquerade as a successful one.

Study data stays intact. Failed review saves no longer reset your spaced-repetition intervals. The Review page now feeds its ratings into topic mastery, so reviewing cards genuinely moves your mastery scores. Roadmap AI plans are matched to the correct subject, regeneration no longer truncates your completed history, and deleting a subject or assessment cleans up all its dependent data.

Even the AI's LaTeX got fixed. The AI was being taught LaTeX with corrupted instructions — JavaScript string escaping ate the backslashes, so the model saw garbled text instead of TeX. The instructions now reach the model intact, so equations, chemical formulas, and scientific notation render properly in generated content.

What's Next

v2.9.0 closes the loop on quality before launch. The waitlist gives us a signal — and a channel. Between now and opening the doors, we're finishing the last mile: family-aware exam prediction, personalised learning paths, and cross-family analytics.

If you haven't already, join the waitlist. When Claritii opens, you'll be first in line — and you'll be getting a study coach that's been through its most rigorous hardening yet.