PsychoHAistoryResearch Initiative
Enter the Laboratory

Technology · Data and architecture

A proposed modular architecture, described before it is built.

What is proposed is labelled proposed. What runs today is only this website and its demonstration model.

  • Implemented Running today (on this website).
  • Prototype A demonstration exists; not a research instrument.
  • Proposed Designed on paper; not built.
  1. 01

    Data Connectors

    Proposed

    Read-only adapters that turn work systems into time-stamped events, after a separate security and privacy review.

  2. 02

    Historical Event Store

    Proposed

    An append-only record of tasks, handoffs, reviews, approvals, failures and completions.

  3. 03

    Organizational Graph

    Proposed

    Who and what collaborates with whom: teams, agents, dependencies and approval routes, versioned over time.

  4. 04

    Feature Engineering

    Proposed

    Reproducible features — cycle times, queue lengths, coordination density — computed from declared data windows.

  5. 05

    Predictive Models

    Proposed

    Baseline statistical models first; learned models only where they beat baselines out of sample.

  6. 06

    Scenario Simulation

    Prototype

    The Interactive Lab on this site is a heuristic demonstration of the scenario layer, with documented assumptions.

  7. 07

    Evaluation & Feedback

    Proposed

    Every prediction is scored against the observed outcome; calibration and error are published per model version.

Proposed data sources

All connectors are proposed. None is implemented, and none collects data. Each integration requires its own architectural and security review, and nothing will be deployed into other Tracston products as part of this website.

First-party Tracston integrations

  • Otopia — Agent workplaces, delegation and operational coordination. Proposed
  • Observatory — Execution telemetry, AI flow and performance monitoring. Proposed
  • Excapia — Mission execution, QA results, scenario testing and outcomes. Proposed
  • Imperium — Policy, approvals and AI governance events. Proposed

Potential external connectors

  • Jira
  • GitHub
  • GitLab
  • CI/CD platforms
  • Project management systems
  • Organizational resource planning systems

Internal data domains

  • Event History

    Immutable work events with source and time.

    Proposed

  • Organizational Graph

    Versioned team and dependency structure.

    Proposed

  • Feature Store

    Named, reproducible features and their windows.

    Proposed

  • Model Registry

    Every model version, its training window and evaluation.

    Proposed

  • Prediction Ledger

    Every prediction ever issued, and how it turned out.

    Proposed

  • Scenario Store

    Configurations explored and the assumptions used.

    Proposed

  • Experiment Registry

    Pre-registered research questions and their status.

    Proposed

The Prediction Ledger

The most important long-term component. A prediction that is never checked is an opinion. For every prediction the ledger records what was predicted, from which data and model, and what actually happened — so prediction quality can be measured and, we hope, shown to improve.

Prediction Ledger record schema (proposed)
TimestampSource data windowModel versionPredictionConfidence estimateAssumptionsActual observed outcomePrediction error
No predictions recorded. The ledger is proposed, not built.

This website

  • Static site

    React + TypeScript, pre-rendered HTML; no server-side code, no database.

    Implemented

  • Rendering

    The observatory is WebGL (Three.js) with a lighter mode for small or slow devices and a still image for reduced motion.

    Implemented

  • Data

    Versioned local demo datasets only. No production data, no external APIs, no AI spend.

    Implemented

  • Privacy

    No cookies, no analytics, no forms. Nothing you do on the site leaves your browser.

    Implemented