Shayne WinterAI Systems Lab

HomeAI Systems LabPersonality Synthesis Engine

Case A6 / privacy and explainability

Private alpha & tested

Private self-reflection with stable scoring and understandable explanations.

Personality Synthesis Engine is a private self-reflection application that works without an account, keeps the profile with the user, gives consistent results, and explains those results in approachable language.

  • Designed
  • Implemented
  • Tested
  • Private preview
  • Readiness gates active
Platform: TypeScript · Next.jsCore boundary: AI excluded from scoring and routingCurrent purpose: Private self-reflection and explainable synthesis

The problem

A self-reflection tool should not hide how it reached its result.

These tools can overstate certainty, expose sensitive answers, change results unpredictably, or let generated wording quietly affect the score. This system keeps scoring consistent and explainable while the user controls and can restore the profile.

User needs

Privacy, recovery, and plain-language explanations were required.

  • No account is required and profiles remain client-held.
  • Scenarios and answer choices are governed content rather than generated at runtime.
  • Routing is bounded and deterministic.
  • Assessment state can recover and restore with version-aware behavior.
  • Relatable and Precise wording serve different explanation needs without changing the underlying score.
  • AI does not decide scoring, confidence, routing, or compatibility.

What it enables

Repeatable self-reflection without giving AI control over the score.

Stable assessmentGoverned scenarios and deterministic rules keep results repeatable instead of allowing generated prose to change the score.
User-held privacyProfiles remain with the user and can be restored without requiring a central account.
Useful explanationRelatable and precise language can explain the same underlying evidence with appropriate confidence.
Evidence-led releaseRelease evidence grows through deterministic checks, recovery testing, compatibility review, and participant study.

The public portfolio uses product architecture and synthetic examples; profiles, answers, messages, and personal inference data remain private.

My role

Turn the original idea into a private, testable, understandable product.

Product framingConverted a broad personality-test idea into a bounded private self-reflection tool.
Privacy and authorityRequired accountless local use, client-held profiles, deterministic decisions, and no AI authority over scoring.
Usability directionShaped paired Relatable and Precise wording, recovery behavior, and evidence-aware explanations.
AcceptanceRequired governed content, regression tests, version-aware restoration, explicit readiness gates, and fail-closed release decisions.

How AI assisted the build

AI helped with development, but fixed rules determine the score.

AI agents contributed implementation, review, testing, and explanatory content work. Shayne retained the product and privacy model, deterministic scoring and routing boundary, content governance, interpretation of user feedback, and acceptance. The implementation uses AI as an engineering collaborator while keeping scoring and interpretation rules deterministic and governed.

Verification

Strong automated testing with clear requirements for a wider release.

Governed modelVersioned constructs and interpretation rules.
Adaptive scenariosFixed and routed scenarios support bounded adaptation.
Paired wordingRelatable and Precise answer language is governed together.
Deterministic verificationBroad automated coverage supports the private alpha.

Release gates map every broader-release requirement to named evidence, keeping automated verification, participant research, compatibility evaluation, and release readiness distinct and reviewable.

Result & next stage

A well-tested private alpha built around user privacy.

The result is designed, implemented, and strongly tested with a trusted private HTTPS preview. Its current purpose is private self-reflection and explainable synthesis; participant research, compatibility evaluation, and broader release readiness are the next validation stages.

Release gates keep required evidence visible and define the acceptance path from private alpha to broader evaluation.
Next AI Systems Lab case studyCodeConstellation