Introduce optimizeImage (lib/functions/image-optimizer.js) and tests (test/image-optimizer.test.js). The new action sends a note's embedded image to a Tinify shrink endpoint (defaults to a Plausible proxy), downloads the optimized bytes, replaces the note image in-place, and alerts the user with byte and percent savings. Tests call Tinify directly when an API key is present. Result: smaller images, reduced storage and bandwidth.

Add a new unit test (test/proposed-agenda-next-day.test.js) and adjust behavior so the Proposed Agenda plugin will propose schedules for future dates when it's already late in the current day. One helper line was moved from lib/dashboard/proposed-agenda-service.js into the test context. Result: avoids skipping next-day proposals and improves scheduling accuracy and reminders.

Adds minute-aware clock formatting (formatClockLabel(totalMinutes, timeFormat)), exposes the proposed-agenda widget in dashboard layout profiles with adjusted ordering and sizing, and raises shared-notes MAX_SCAN to 1000 in lib/shared-notes-service.js to broaden discovery. Impact: minute-precision time labels, clearer agenda visibility in dashboards, and improved shared-note discovery.

Add ability to filter shared notes by person and paginate results; annotate and surface the time taken to build the person↔note index. Updated shared-notes styles and widget tests to support the UI changes, improving navigation of large note sets and visibility into index build performance.

Fixes an issue where the dashboard config popup could be positioned offscreen. Updates JS positioning logic in lib/dashboard/dashboard-load.jsx, lib/dashboard/dashboard.jsx and lib/dashboard/dream-task.jsx and adds/adjusts CSS in config-popup.scss and dashboard-layout-popup.scss to constrain placement. Result: config dialog no longer jumps outside the visible viewport, improving accessibility and UX.

Update ensure_optimal_seat_allocation (vendor/gems/subscribe_manager/.../seat_generation.rb) to use each committer's total commit count as a fallback when there are more seats than active committers. This preserves sensible seat distribution when recent-activity-based allocation can't fill seats, preventing unassigned seats and improving fairness. Deployed to production.

Lowered the comment-impact weight in app/lib/defines/code_lines.rb from 1.0 to 0.1 to de-emphasize comments when scoring Prompt-to-Production code. This reduces the influence of comment-heavy lines on metrics and downstream prompts, prioritizing executable code and improving metric robustness.

Implements v1 of the Maintainability Gap research feature, adding extensive integration tests (test/integration/proto_user_signup_test.rb) and route updates to support AI code-quality and write-only-mode research. Deployed to production on 2026-06-24 01:02. Enables collection and evaluation of v1 maintainability-gap metrics for the AI code-quality study.

Add a legacy_refactor_percent metric and integrate it into stat definitions and cached graph calculations. Changes in app/lib/defines/stat_segments.rb add the new segment; app/lib/presenters/concerns/cached_stat_graph_value_calculate_concern.rb wires it into interval, numerator and denominator calculations. This lets customers evaluate their refactor percentage against soon-to-be-published industry stats, improving comparability and reporting.

Add AI-Assisted Commit percentage and total token count to the default AI Usage stats view (app/views/shared/resource_actions/ai_usage_stats.html.haml). The update surfaces percent of commits assisted by AI and cumulative token usage in the UI, improving visibility into AI consumption. Deployed to production on 2026-06-22. Impact: clearer metrics for monitoring, reporting, and billing.

Adds a v2 LLM-driven Task Rescorer (lib/functions/task-rescorer.js) using gpt-5.4-mini to propose clickable score recommendations with one-sentence rationales. Adds interactive task-source selection, rescore presets, configurable max-per-pass, concurrent batching, LLM quota/rationale limits, provider logging and user API-key fallback, and a smoother apply flow. Fixes by-tag lookup bugs to improve reliability, observability and cost control.

Add grouping logic to CreateTeamMessengerNotifications (vendor/gems/resource_notifications/app/jobs/create_team_messenger_notifications.rb) to combine notifications of the same type into unified messenger blocks, pick the best image block per group, and refine channel selection. Result: reduced duplicate posts and clearer, consolidated notifications in Slack/Teams.

Record start and end dates for LLM data queries in the Data Interrogator flow. Updated app/javascript/apps/data-interrogator.js (submitQuestion, runQuestion, submitFollowUp) and transcript-inline-chart.js to capture and surface timestamps for each query. Deployed Jun 1, 2026. Improves observability, auditing and timeline/charting of LLM interactions.

Added two cohort-level metrics—'Issues resolved by cohort' and 'Defect originated by cohort'—to the AI usage cohort stats view (app/views/shared/resource_actions/ai_cohorts_stats.html.haml). This change restores missing analytics signals, addressing a bug that hid these metrics and improving cohort-level visibility for resolution tracking and defect attribution.

Introduce a new JS module (app/javascript/bundles/data-interrogator/transcript-inline-chart.js) to render charts inside the Data Deep Dive popup. Made small adjustments in transcript-event.js (renderToolBlock/renderToolActions) to integrate the chart component. Impact: charts now display inline in the popup, improving on-demand data visualization during deep dives.

Add resume bar and follow-up question support to the Data Deep Dive drawer. Implemented new renderResumeBar and follow-up logic in app/javascript/apps/data-interrogator.js (renderDrawer → runQuestion → renderResumeBar), adding UI elements and question handling; released to production 2026-06-01. Enables users to resume previous deep-dive sessions and ask guided follow-ups, improving investigation continuity and UX.

Rework pricing page to modernize visuals and copy: add dark/light hero PNGs, adjust pricing.html.haml to surface “Inspired AI ROI stats” and link to AI research, remove annual-contract blurb, and update introduction SCSS gradients, alignment, and margins. Changes deployed to production to improve theming, clarity, and positioning of product benefits.

Implements v3 LlmUsageMetric to start recording all provider token count types aggregated over the past year and removes a few non-essential columns. Adds provider_usage JSONB builders and extends platform/team LLM usage stats in external_usage_client/anthropic_concerns.rb, and removes an obsolete fetch_cursor_ai_usage_stats call in populate_llm_metrics.rb. Deployed May 25 — improves telemetry for reporting, billing, and analytics.

Add a migration (db/migrate/20260525152531_add_llm_usage_cached_token_count.rb) and update the LlmUsageMetric pipeline (vendor/gems/llm_metrics/app/lib/external_usage_client/external_cursor_ai_usage_client.rb) and tests to record all provider token count types over the past year while removing a few non-essential columns. Business context: closes a gap in provider token telemetry for usage/billing. Impact: more accurate historical token metrics and a simplified schema.