What's New
The latest updates and improvements to Curia.
New
Structured Analysis Reports
Transitioned from memo-based output to structured evidence analysis reports. All output types now build from consistent, well-organized evidence analysis data rather than prior memo prose.
- Analysis output is now built from structured evidence analysis data, not unstructured memo text
- Quality evaluation judge now correctly assesses analysis reports instead of outdated memo format
- UI labels and status messages updated to reflect structured report terminology
- Removed memo-era language from analysis prompts to prevent downstream re-appending
- All output types now use consistent structured data foundation
Improved
Improved Email Communications
Enhanced onboarding, upgrade, and re-engagement email templates with clearer messaging and more relevant content.
- Rewrote onboarding emails for new users with clearer value proposition
- Upgraded re-engagement email prompts with better tone and call-to-action
- New case alert prompts now highlight specific holdings rather than vague filler content
Fix
Search and Data Quality Fixes
Fixed critical bugs affecting search reliability and data consistency across the platform.
- Permanently resolved case_id ambiguity issues in case search results
- Fixed nullable newCasesCount handling in engagement checks
- Improved UUID validation and consistent credit check logic
- Enhanced error handling, design system consistency, and security practices throughout
Research Pipeline Stability and Accuracy
Resolved multiple bugs affecting research quality and reliability. The platform now correctly handles search timeouts, question detection, and provides more reliable case recommendations.
- Fixed hybrid search timeout by removing expensive operations from keyword search branch
- Corrected cache scoping and question detection logic for consistent results
- Improved escalation query handling and reflection accounting across pipeline stages
- Softened prior-citation reliability claims to prevent overstating case relevance
- Added word boundary constraints to refinement prefix matching
Expert Report Analysis Pipeline Improvements
Fixed multiple issues in the expert report analysis workflow to improve accuracy and prevent hallucinations. The system now correctly identifies expert reports, validates document types, and provides better-grounded research recommendations.
- Clarified that expert attack analysis is specifically for defence medical expert reports
- Added document type validation to ensure only valid reports enter the analysis pipeline
- Fixed unwired prompt issue and removed prosecutorial bias from analysis instructions
- Improved follow-up system prompt with dedicated research guidance and provenance discipline
- Resolved prompt extraction speculations and grounding issues in follow-up recommendations
Improved
Enhanced quality of generated legal research content
Upgraded the prompts that generate research memos, analysis memos, research answers, counter-arguments, and outreach emails to produce higher-quality, more professional legal content.
- Analysis memo prompt now produces professional-grade output
- Research memo prompt upgraded to match analysis memo quality standards
- Research answer and counter-argument generation improved
Fix
More accurate document analysis quality checks
Fixed false positives in the quality assessment process and improved JSON parsing from analysis results, reducing incorrect quality alerts.
- Reduced false positive quality alerts on document analysis
- Improved JSON extraction from analysis responses
- Quality judge now correctly interprets analysis output
Improved memo and analysis generation reliability
Fixed timeout issues that occurred during memo and document analysis generation by implementing streaming responses, ensuring these operations complete successfully regardless of processing time.
- Memo generation now uses streaming to prevent timeout failures
- Document analysis processing now uses streaming to handle longer operations
New
Enhanced Navigation and Sidebar
A redesigned sidebar navigation now makes it easier to organize and access your matters, with improved layout and quicker access to frequently used files.
- Recently accessed matters appear in the sidebar for quick navigation
- Expandable matter folders with subcategories (documents, research, analysis, outputs)
- New command center for centralized matter management
- Matter overview page with dedicated sub-pages for different work types
Self-Reflecting Research Pipeline
The platform now automatically evaluates and improves research answers before delivering them to you, ensuring higher quality case law analysis and recommendations.
- Research answers are automatically reviewed for accuracy and completeness
- Answers are refined before presentation to ensure better legal insights
Improved
Improved Layout and Display
Content areas on research and analysis pages now use the full available width, and interactive elements respond better to user actions.
- Removed width constraints from matter analysis, research, and outputs pages
- Better visibility and workspace for case law research
- Improved hover interactions and marquee controls
New
Citation Verification in Analysis
The platform now checks that all citations in your analysis actually exist in the case law database and surface citation counts as analysis progresses.
- Real-time citation count displayed during analysis
- Citations are verified against the case law database
- Invalid or missing citations are flagged for review
Practice Area Context in Research
The platform now uses your matter's practice area to scope and refine searches, ensuring results are more relevant to your specific legal domain.
- Matter practice area automatically scopes case law searches
- Search results are reranked to prioritize cases relevant to your practice area
Follow-Up Conversation Limits
Conversation limits help keep research sessions focused and prevent excessive back-and-forth. You'll receive a gentle reminder at 5 follow-ups and the session will pause for review at 10.
- Soft nudge after 5 follow-up questions encouraging session review
- Hard stop at 10 follow-ups with server-side enforcement
- Limits are enforced across all research sessions
Save Research to Matters
You can now save your research sessions and expert attack analyses directly to a specific matter, making it easier to organize and reference your work.
- Save-to-matter option available during standalone research sessions
- Expert attacks can be saved directly to your matters
- Quick access to save from research banners
Clarifying Question Triage
The platform now asks clarifying questions when needed to better understand your research goals before diving into full analysis. This helps ensure you get more relevant results from the start.
- Automatic clarifying questions appear at the beginning of research sessions
- Follow-up research now includes a triage step to refine your inquiry
- Your clarification answers are remembered throughout your session
Improved
Search Accuracy Enhancements
Multiple improvements to search relevance and result quality, including better handling of practice area filtering and improved citation matching.
- Relevance ranking improved for citation matching
- Better handling of practice area scope in search results
- More accurate similarity scoring in hybrid search
Improved Search Deduplication
Hybrid search results are now deduplicated by case ID to prevent duplicate cases from appearing in your results.
- Duplicate case results removed from hybrid search
- Search performance optimized to prevent timeouts
- More concise, relevant result sets
Fix
Bug Fixes and Stability
Various fixes to improve platform reliability, including corrections to follow-up limits, citation checking, quality alerts, and feedback handling.
- Quality alerts now read correct case counts
- Feedback display time increased for better visibility
- General experience feedback category now properly supported
- Fixed edge cases in citation verification
New
Export Research Results as PDF and Word Documents
You can now export your research findings directly from the platform in both PDF and Word formats, making it easy to share results with colleagues or include in client materials.
- Export button available on research, expert attack, and matter research pages
- Support for both PDF and DOCX formats with professional formatting
- Proper table rendering, bullet points, and blockquote formatting in exports
- Consistent fonts and layout across all document types
User Feedback and Quality Monitoring
New feedback collection system helps us continuously improve search results and research quality. Users can now easily report issues, and our team receives real-time alerts about quality concerns.
- Feedback widget automatically appears during research sessions
- Quality alerts sent to our team when issues are detected or when users provide negative feedback
- Branded feedback interface aligned with Curia's design
Improved
Enhanced Research Context in Follow-Up Conversations
Follow-up research now maintains better context from your original query, ensuring more relevant and accurate results in multi-turn research workflows.
- Follow-up responses retain full research context from the initial query
- Improved handling of conversation history to prevent context loss
- More consistent and coherent multi-step research sessions
Fix
Document Export Formatting Improvements
Fixed numerous formatting issues in exported documents to ensure professional appearance when sharing research results.
- Tables now render correctly with proper column widths and aligned cells
- Bullet points and lists display consistently across PDF and Word formats
- Markdown formatting properly converted to native document elements
- Fixed footer placement and heading styles in generated documents