Introduce a DaySketch dashboard widget (lib/dashboard/day-sketch.js) with unit tests (test/day-sketch.test.js): a 6am–9pm notebook-style planner that preloads agenda items and debounced-saves entries into a per-day Day Sketch note. Also prevent duplicate Recent Notes entries, fix Dream Task reseed flow and logging, and simplify embed handling for improved mobile UX — resulting in a tested planner, fewer duplicate notes, clearer reseed logs, and smoother mobile embeds.

Add app/javascript/apps/starred-report-picker.js — a new Starred Report Picker component that implements autosuggest and input filtering (filterSuggestions with autoExpandTextarea, buildSuggestionList, and key handlers), saveStarredReport flows (renderAutosuggestItem, handleSearchInputChange) and AI-driven suggestion cards (requestAiSuggestions/renderAiSuggestionCard). Improves report selection and description entry UX and surfaces AI suggestions to speed saving starred reports.

* Build out all_time_defect_finder scope, alongside period_defect_finder* Reuse the concerns_team scope for a second use case* Update concerns_team scope to accept a resource to combine with team for deriving repo_ids

Adds a new API endpoint to provision a user using a Personal Access Token (PAT) and introduces an integration test (vendor/gems/customer_api/test/integration/api_provision_user_test.rb) to validate provisioning flows and error cases. Enables programmatic user provisioning with PAT and improves API test coverage and reliability.

BackfillCommitRelease and CommitConcerns updated so the backfill job prefers a commit's pull request's repo_release when deciding release attribution. Added assign_pull_request_commits and adjusted ensure_repo_release_recorded/ensure_commit_repo_issue_released to consult PR data first, with existing heuristics as a fallback. Produces more accurate repo_release backfills and fewer incorrect release mappings.

Implemented per-month quarterly planning in lib/data-service.js (getMonthlyPlanContent, createOrAppendMonthlyPlan) and added _readMonthContentWithRetry to handle API eventual-consistency after inserts. Overhauled dashboard mood graphs (date handling, tooltips, responsiveness, popover placement) and corrected agenda timezone rendering and chronological mood ordering. Result: more reliable monthly planning and improved dashboard UX and accurate agenda/mood display.

Extract shared inspiration-card styles into app/assets/stylesheets/help_pages/_inspiration_shared.scss and update the index inspiration section. Convert category chips to clickable links, refresh the mood‑meter UI, and add a roomy 'spacious-form' admin style. Addresses design inconsistencies and poor tag discoverability; improves tag navigation, mood feedback, and visual consistency in production.

Updated mood-ratings.js to refresh MOOD_RATING_EMOJIS and MOOD_RATING_FLARE_EMOJIS and to completely rewrite MOOD_RATING_AFFIRMATIONS with livelier, more varied copy. Also adjusted descriptor labels (added “Ok” and reordered labels). Pure content/UX change (no logic), improving tone and user engagement when recording moods.

Update renderColumnDelta in app/javascript/bundles/directory-browser/browser-column-renders.js to display LLM-generated delta data alongside existing standard per-directory deltas, with accompanying style tweaks in repo-file-browser.scss to support the new indicators. Result: directory browser UI now shows both standard and LLM deltas for clearer, side-by-side change context.

Add segments for commit_ai_lines_committed_count and delta_from_llms. Updated llm_resource_stat_creation_calculation_concern to create/emit the new segments and extended SegmentDisplay plus ApplicationResourceHelperConcerns to include display entries and chart icons. Deployed to production — enables aggregation and visualization of these new LLM-derived metrics in dashboards.

Adds a new HAML page (app/views/pages/mathematics_of_diff_delta.html.haml) implementing the Claude v1 "Mathematics of Diff Delta" documentation. The page contains structured explanatory content, formulas and markup to integrate with site styles and tooling. Deployed to production to provide a canonical, developer-facing reference for the diff/delta mathematics and improve onboarding and troubleshooting.

Add Escape-key handling in editing-comment-content.js by implementing _onKeyDown logic to cancel/dismiss the comment input/edit UI when ESC is pressed (with related key-handling branches). Improves keyboard accessibility and lets users quickly close the comment editor for a better UX.

Fix parameter parsing in CustomerApiBaseController::derive_api_request_committer to correctly accept a committer_identifier when it's provided as a remote_id (used by AI API usage stats). Updated integration tests (api_ai_usage_test.rb) to exercise the case. Result: correct committer attribution and accurate AI usage reporting.

Add logic in CommitConcerns::ValueEvaluationConcerns to detect high-velocity commits associated with an LLM model, auto-approve them as valid Diff Delta, and deduce state/value that splits reward between the developer and the LLM. Added test coverage in llm_metrics assign_commit_llm_model tests. Impact: LLM-generated high-throughput commits are auto-validated and compensation is allocated between human and model.

Updated config/appsignal.yml to stop sending background job metadata/telemetry to AppSignal. The change prevents background job traces from being reported, reducing AppSignal usage and expected monthly costs so we can evaluate AppSignal for performance tracking without incurring ~$1k/month.

Add a backfill job and Rake task to populate LLM model/classification on commit-related subscriber entities. Implements a new ai.rake task and updates AssignCommitLlmModelTest to cover the behavior; schedule entry ensures the job runs as configured. Benefit: populates existing commits with model assignments so downstream features relying on LLM classification behave correctly.