Entropy Compatible Hiring

This release adds Existing Employee Contribution Mapping as a distinct module within the ECH desktop application. The module is for exploring potential internal entropy-reducing contribution patterns. It is separate from the hiring architecture and is not a performance rating.

New in v0.1.3

• Added a green Existing Employee Contribution Mapping module with navigation to and from ECH.
• Added independent employee reflection and manager observation assessments across all ten OEI archetypes.
• Added a comparison workspace that preserves each source separately and highlights convergence, discrepancy, and missing evidence.
• Added an applied validation endpoint that recommends the strongest current archetype for a structured OEI exercise.
• Added an employee-safe pop-out workflow designed for facilitated video calls and remote control.
• Added separate employee validation session storage and participant routes so inversion data does not enter candidate session storage.
• Added clear missing-response warnings and required setup boundaries.
• Added an ECH landing slide explaining the existing-employee module and its separation from hiring.

Important boundaries

• Contribution Mapping is exploratory and does not determine employee performance.
• Results are provisional contribution hypotheses, not permanent employee labels.
• The recommended archetype should be validated through a bounded, structured work exercise.
• Hiring sessions and existing-employee sessions remain separate. Their intentional connections are navigation and the recommended archetype passed into an applied OEI exercise.

Known limitations

• Candidate and employee participant experiences are local pop-out windows, not remotely hosted links.
• The employee workflow recommends Zoom remote control, which may be unavailable due to organization settings, permissions, device support, or call conditions.
• The manager may be able to see employee reflection responses during a facilitated screen-share session.
• Inversion scores, thresholds, and profile language are internal beta design assumptions, not validated cutoffs.
• Hiring cases, employee reflections, manager observations, and validation sessions are not durable records and clear when the application closes or the relevant case is concluded.
• AI assistance supports OpenAI only and requires a user-supplied OpenAI API key.
• Deterministic fallback exercises remain below the intended quality of AI-assisted customization.
• Updates are not automatic.
• Windows may display an unfamiliar publisher warning if the installer is not code-signed.

Installation

Install the v0.1.3 package over the previous beta. Existing beta releases do not update automatically.

This release improves packet configuration, strengthens the OpenAI integration, and updates the desktop loading experience.

What’s new

Packet-specific context guidance

• Added a “Don’t know where to start?” guide to every interview instrument.
• Each of the ten packets now explains:
  • What the completed candidate exercise will look like.
  • What information should be entered in each context field.
  • Which details are required for a credible, packet-compatible exercise.
• Guidance is unique to each OEI archetype and aligned with its generation rules.
• Guidance dialogs support keyboard navigation and can be closed with the Close button, the background, or Escape.

AI generation reliability and safety

• Strengthened the separation between OEI packet instructions and user-provided company context.
• Company context is explicitly treated as input data, not as instructions to the model.
• Only the packet’s declared context fields are sent to OpenAI.
• AI responses now use a strict structured-output schema.
• Returned exercises receive stronger validation before being accepted.
• Invalid, incomplete, malformed, or failed OpenAI responses safely use the clearly labeled deterministic fallback.
• OpenAI response storage is disabled for generation requests.
• Added clearer interface text explaining how to connect an OpenAI API key.
• Expanded automated coverage for successful AI responses, upstream failures, malformed output, and deterministic fallback behavior.

Desktop experience

• Replaced the loading-screen image with the corrected OEI short logo.
• Updated visible application and installer version information to v0.1.2.
• Refined the context-guidance interface for desktop and smaller windows.

Desktop delivery

• Distributed as a single Windows installer EXE.
• Runs a private, loopback-only application service.
• Customers don’t need Node.js, repository files, or supporting folders.
• Candidate material opens in a separate candidate-safe window.
• The recommended Beta workflow remains to screen-share only the candidate-safe window during a video call.

AI and API keys

• No shared API key is embedded in the application.
• Customers can provide their own OpenAI API key.
• Keys remain session-only and aren’t intentionally written to disk.
• Company context is sent to OpenAI only when AI-assisted generation is used.
• The methodology, required exercise conditions, evidence categories, and interpretation rules remain deterministic and packet-defined.
• AI may customize realistic scenario details but cannot redefine the assessment methodology or make hiring decisions.

Beta limitations

• Candidate links aren’t remotely hosted or independently shareable.
• Candidate delivery currently relies on screen-sharing the candidate-safe window.
• Interview cases and candidate sessions aren’t persistently stored after the application closes.
• Updates aren’t automatic. A new installer is required for each release.
• The installer isn’t commercially code-signed, so Windows may display a SmartScreen or unknown-publisher warning.
• Deterministic fallback exercises remain below the intended AI-assisted quality and are clearly labeled.
• Windows is the only packaged platform in this release.

Assessment boundary

This software records operational-behavior evidence relevant to specific OEI archetypes and pillars. It doesn’t determine overall employee quality or replace evaluation of skills, experience, technical competence, references, or broader hiring fit.

Initial Windows desktop release of the OEI Entropy-Compatible Hiring interview instrument.

Included

• Ten OEI archetype-specific interview instruments.
• Optional routing from two known OEI entropy drivers to one recommended instrument.
• Packet-specific company-context intake.
• Progressive optional fields for more detailed customization.
• AI-assisted exercise generation using the customer’s own OpenAI API key.
• Clearly labeled deterministic fallback when AI isn’t available.
• Mandatory interviewer review and approval before candidate release.
• Separate candidate-safe and private interviewer views.
• Evidence states:
  • Evidence observed
  • Contrary evidence observed
  • Insufficient evidence
• Section-level note capture with consolidated notes.
• Evidence provenance carried into the final report.
• Structured OEI evidence reports with explicit assessment boundaries.
• Candidate-link resending, case restart, and case conclusion controls.
• OEI-branded desktop interface.
• Four-slide desktop landing presentation with keyboard navigation.

Desktop delivery

• Distributed as a single Windows installer EXE.
• Runs a private, loopback-only application service.
• Customers don’t need Node.js, repository files, or supporting folders.
• Candidate material opens in a separate candidate-safe window.
• The recommended Beta workflow is to screen-share only the candidate-safe window during a video call.

AI and API keys

• No shared API key is embedded in the application.
• Customers can provide their own OpenAI API key.
• Keys remain session-only and aren’t intentionally written to disk.
• Company context is sent to OpenAI only when AI-assisted generation is used.
• The methodology, evidence categories, and interpretation rules remain deterministic and packet-defined.

Beta limitations

• Candidate links aren’t remotely hosted or independently shareable yet.
• Candidate delivery currently relies on screen-sharing the candidate-safe window.
• Interview cases and candidate sessions aren’t persistently stored after the application closes.
• The installer isn’t commercially code-signed, so Windows may display a SmartScreen warning.
• Deterministic fallback exercises are below the intended AI-assisted quality and are clearly labeled.
• Windows is the only packaged platform in this release.

Assessment boundary

This software records operational-behavior evidence relevant to specific OEI archetypes and pillars. It doesn’t determine overall employee quality or replace evaluation of skills, experience, technical competence, references, or broader hiring fit.