Changelog

Changed

• Free plan moved from a weekly to a monthly billing cycle. Each Free account now receives 500 credits per month, resetting on the 1st.
• Pro plan stays at 5,000 credits per month with all existing features.
• Ultra plan increased from 14,000 to 20,000 credits per month at the same price.
• Web search infrastructure for Deep Research migrated to a faster, lower-cost provider. Same source coverage, less credit consumption per report.

Improved

• Credit hints, mode-upgrade prompts, and the pricing page now consistently reflect the monthly cycle.
• Quota error messages are clearer when a credit limit is reached.

Added

• Cross-device research resume. A Deep Research run started on one device can be followed or finished on another while signed into the same account.
• Research lifetime decoupled from the active browser tab. Closing the tab, switching networks, or reloading no longer interrupts an in-progress run.
• Clarification questions are persisted with the conversation so reloading the page restores your in-progress interview.

Improved

• Deep Research pipeline rebuilt around a single authoritative run state. Activity timelines no longer stall at “connecting” or display phases out of order.
• Cancelling a research run now reliably stops backend work within seconds; previously it could continue silently for several minutes.
• Faster Deep Research finalisation by removing an unnecessary multi-minute citation rebalance pass on every run.

Added

• Editorial typography pass across the product: serif body text in reading panes, monospaced uppercase kickers for section headers, and calmer spacing throughout.
• Activity stream redesigned. Live web-search responses now show a per-source fan-out grid with reading progress for each source, not just per-stage summaries.
• RAG pull-quote evidence: when an answer is drawn from your own existing reports, the original passage is shown inline as a quoted source.
• Free plan limits enforced consistently across all chat modes, with a clearer upgrade prompt when limits are reached.

Improved

• Diagram repair quality on partial or invalid canvases.
• Book Pane displays a writing skeleton while a Deep Research run is mid-flight.
• Mode-upgrade prompts now run after retrieval, so questions answerable from existing reports no longer trigger an unnecessary search-upgrade card.

Added

• First-time user tour: an 8-step guided walkthrough of the workspace, replayable from Settings.
• Two-factor authentication via TOTP (Google Authenticator, 1Password, Authy, and other standard authenticator apps).
• Email verification with auto-login once the verification link is clicked.
• Welcome email sent on account creation with a brief product orientation.
• In-app feedback widget supporting multiple screenshots per submission.
• PWA install flow and a public Download landing page covering Mac, Windows, and PWA install paths.

Improved

• Settings panel internationalisation completed across all nine supported languages (Data & Privacy, Sessions, 2FA, and other panels).
• Marketing UTM parameters forwarded into the app for accurate signup attribution.
• Admin console fixes for mobile layout and conversation drop-off triage.

Added

• Version-aware research flow where deep research creates a new report page version instead of overwriting current work.
• Page overview graph for navigating version ancestry and previewing report content.
• Improved report-to-canvas handoff for node hydration and interactive branch exploration.

Improved

• Section persistence behavior tied to active canvas version.
• Project loading reliability and duplicate-project bootstrap handling in sidebar logic.

Added

• Rich notebook editor with slash commands, image upload support, and autosave conflict handling.
• Report text selection to note handoff flow for fast extraction into notebook context.

Improved

• Upload resilience and draft state handling in notebook autosave pipeline.
• Export readiness for report-derived sections.

Added

• Live research activity stream with stage-level progress and reconnection behavior.
• Run cancellation controls for both chat and deep research requests.

Improved

• Error normalization and user-facing request-state messaging.
• Debounced save guards to prevent cross-project stale writes.

Added

• Initial public marketing site for product, workflows, gallery, docs, and pricing.
• Foundational policy pages for terms, privacy, and refunds.