BUILD YOUR JURY BY ZIP CODEYour case has a venue.
Build the jury around it.
Enter the case’s five-digit U.S. ZIP code. Case Lab uses local Census estimates to establish the demographic foundation, then generates 100 individual Synthros jurors for that case.
Each juror receives a portrait and a complete persona: work and household circumstances, education, personality, relationships, goals, pressures, beliefs, legal attitudes, media habits, cognitive biases, local context, private cognition, short-term memory, and long-term memory.
Explore every profile and choose the jurors you want. The pool stays with the case across focus groups, deliberation, interviews, and controlled presentation tests.
Meet the personas behind the experience ↗
A LIFE BEFORE THE CASEThe background matters.
So does the person.
A retired renter living alone brings a different set of circumstances into the exercise than a young professional balancing work and family. Financial pressures, community ties, trust patterns, aspirations, and fears give each Synthros persona a distinct starting point.
Case Lab carries that identity into the case. You can see who is responding, explore the context behind the response, and return to that juror with a follow-up question.
Explore the thirteen-dimensional architecture ↗
CIVIC TWINS · CONTEXT-AWARE INTELLIGENCEA community surrounds
every courtroom.
Jurors encounter your presentation through the world they already know. Synthros personas bring information habits, worldview, personal circumstances, and available local and national context into the exercise.
Local population preparation can add public news context for the selected geography. Case Lab records the context and its timestamps so you can understand the setting behind a response and return to it when comparing presentations.
That creates questions worth exploring: Where does this juror’s trust come from? What attracts their attention? Which assumptions carry into the discussion?
COGNITION · MEMORY · DISCUSSIONFollow the perspective
as it develops.
- Make your case. Add both positions, your key question, and supporting excerpts. Approve the brief your panel will receive.
- Hear the first reactions. Each juror responds through their own identity and context. A private cognitive component records a reflection summary.
- Watch the discussion develop. Jurors consider the panel’s initial responses alongside their recent case memory, then record their final reasoning.
- Understand what held—or changed. A longer-term case-memory summary preserves the perspective after discussion. The report brings together themes, questions, and individual responses.
- Keep asking. Return to a juror to explore an assumption, a source reference, or an unanswered question in their recorded response.
FROM INSIGHT TO YOUR NEXT MOVEChange the presentation.
Explore the response.
Test a different framing, an additional piece of evidence, or an alternative assumption. Reuse your panel and compare the new exercise with the baseline.
Your working record keeps the approved brief, sources, persona identities, context, and responses together. Follow the reasoning, identify the questions that remain, and bring the findings back to your team.
See a real response from a fictional case test ↗
A CLOSER LOOKDepth you can explore.
Explore the sources, construction, and evaluation behind Case Lab.
Census sources, ZIP-code coverage, and population alignment
ZIP-code preparation uses Census ZIP Code Tabulation Areas across the United States. Most residential ZIP codes have a corresponding Census area; postal-only and other codes without the required estimates cannot support local Census preparation. The workspace checks ZIP coverage before building a population.
The prepared population uses largest-remainder allocation for adult age × Census sex and, separately, education for ages 25+. The source record shows Census target percentages, allocated counts, and actual counts. Other characteristics use weighted draws or Synthros generation; their full joint distribution is not calibrated.
Occupation estimates describe employed civilians aged 16+, housing describes occupied households, and race/ethnicity estimates describe all residents. These source populations differ from adult jury eligibility. The audit states these differences; Census sex categories do not measure gender identity.
The report preserves the chosen geography, ACS source year and retrieval date, demographic allocation, and source links. Missing or suppressed required estimates stop preparation. No national fallback is labeled local.
A county or Census ZIP area may differ from a court’s jury catchment. The selected twelve-person panel can differ from the prepared population’s proportions. Census alignment does not establish human-jury representativeness or predicted verdict accuracy.
Persona construction and cognitive components
The inspected Synthros model-creation code builds a persona instruction from profile attributes and attaches it to a base model. That is a configured persona; it does not by itself establish that each juror has separately trained model weights.
Case Lab addresses each selected juror through its existing model identifier. It sends the case exercise without replacing that model’s configured system persona. Local preparation creates new Case Lab model identities and their cognitive components. Existing Synthros models are not overwritten; shared base-model weights are not separately trained for each persona.
The thirteen dimensions describe the Synthros architecture. Case Lab connects four model components: juror identity, private cognition, short-term memory, and long-term memory.
Context sources and freshness
Case Lab uses the model context available at the start of the run and shows the model’s update timestamp. New local population preparation can capture public news search results for its Census geography, with source links and publication dates. These results are not independently verified. A failed fetch supplies no invented headlines. Analysis does not automatically refresh news or run thirty-day preparation. The original Idaho models inspected during this build were last updated in June 2026.
County and Census ZIP-area preparation are available in the workspace. Extended venue histories and enterprise deployment customization still require separate work.
Source STM models can contain previous user conversations. Case Lab invokes the existing memory model identities with a new case-scoped memory role and the approved record. It does not import shared chat histories or write case material back into the original jurors.
Evaluation and interpreting results
Integration verifiedThe installation check found matching models for all 100 enabled adult profiles. The expanded workflow completed a fictional twelve-juror exercise using all 48 existing juror, PCL, STM, and LTM component identities. It produced reflection and memory summaries, a discussion, and a saved report. The consistency review flagged conflicting vote explanations.
Output quality is still under evaluationAn earlier test found liability votes that contradicted the same jurors’ written reasoning. The cognitive workflow adds a model-based consistency review and conservative checks. Flagged reports preserve raw responses and label the result “Review required.” These checks do not establish correctness; successful execution is not evidence of predictive accuracy.
We have not established accuracy against human jury studies, demographic representativeness, calibrated win probabilities, or reliable damages estimates. Shared base models can also produce correlated responses. Use Case Lab to inspect questions and presentations, and independently review every finding.
Data handling and account details
Case Lab stores accounts, briefs, excerpts, and results in its own application database. Existing Synthros jurors remain read-only. A private bridge writes each new case pool into a hidden Synthros simulation so the established identity, cognition, memory, and portrait pipeline can prepare all 100 personas without exposing internal controls to legal users.
Uploaded files are processed for text extraction; the application retains the saved excerpts, not the original upload. Analysis sends the approved material to the configured Synthros inference service. Administrators control infrastructure access, service logs, and backups.
Deleting a case removes its material and linked results from the active application database. It does not independently erase backups or inference logs. Email verification and automated password recovery are not currently enabled.
Read the workspace data notice ↗Start with a case.
Plan for your firm.
Use your account workspace to prepare briefs, select local jurors, review analyses, ask follow-ups, and export reports. Case records and memory stay organized around your work.
For larger teams, discuss an annual license, dedicated hosting, or deployment on firm-controlled infrastructure. The same local persona foundation supports the conversation; capacity, integration, and deployment are scoped with your firm.
