ONE RESEARCH FRAMEWORK · THREE PLATFORMS

Use PhygitAi
Across Platforms

Choose a platform for its current availability, access instructions, and future package downloads.

Explore how PhygitAi supports accountable human–AI decision-making in industrial, product, and physical-digital design.

For designers, students, and researchers: examine AI contributions, human decision authority, available project evidence, and opportunities for accountable revision.

OUR RESEARCH APPROACH

From Design Intention to Accountable Decisions

A consistent research approach to understanding human–AI decisions, available evidence, and opportunities for revision.


01

Choose a Platform

Select ChatGPT, Claude, or Gemini.

02

Access PhygitAi

Follow the available access or installation instructions.

03

Explore Your Design

Apply six guided workflows to examine your project evidence.


Which PhygitAi edition can I use?

The downloadable screening files, the hosted ChatGPT plugin and the MCP-backed technical pilot are separate editions. A ZIP download does not grant access to a private hosted plugin.

ChatGPT · Downloadable screening ZIP

v0.1.8 · Public download

Standalone instructions and reference materials. Manual import depends on compatible host features and is not guaranteed.

Download and installation guidance →

ChatGPT · Hosted plugin

v0.1.7 · Restricted Private Preview (author-reported)

Only authorized testers can access this separate hosted edition. Downloading the ZIP does not enable it.

Read the private-preview details →

ChatGPT · MCP-backed edition candidate

Read-only technical pilot · Public approval not confirmed

A separate hosted MCP integration submitted for platform review. A working server or pending review does not establish public plugin availability.

Understand edition differences →

Claude & Gemini · Downloadable Skills

Claude v0.3.1 · Gemini v0.1.5

Research-stage files for manual installation in compatible environments. No native directory approval or automatic updates are claimed.

See releases and checksums →

Availability information reflects documented project status as of 21 October 2026. PhygitAi provides preliminary conceptual screening, not empirical validation or certification.

How PhygitAi Works

Explore a fictional adaptive reading-lamp example to understand what you provide, what PhygitAi examines, and what you receive.

Illustration of a fictional blue adjustable reading lamp beside an open book
Fictional example

Adaptive
Reading Lamp

Illustration only · Not test evidence
01

You provide

  • Project brief and design objectives.
  • Intended users and design assumptions.
  • Reported AI involvement.
  • Available documentation and observations.
02

PhygitAi examines

  • G0–G5 research-gate evidence coverage.
  • AI contributions and their documentation.
  • Human decision authority.
  • Intended versus observed behavior, where documented.
  • Evidence gaps, disagreements and accountable revisions.
03

You receive

  • A structured preliminary G0–G5 review.
  • Documented evidence gaps and limitations.
  • A provisional seven-field Intention–Behavior Record, where supported.
  • Recommended documentation and next steps.

Fictional research demonstration. Illustrative screening only; not empirical validation, certification or independent verification.

Six workflows. One research inquiry.

Audit

Review one project across the six G0–G5 research gates and seven-field Intention–Behavior Record.

Compare

Examine two projects using consistent evidence criteria, without scoring or ranking.

Inspect

Examine project claims, supporting documents, and missing evidence.

Detect

Identify documented discrepancies, evidence gaps, and unclear decision authority.

Explain

Clarify existing PhygitAi assessment findings and their methodological limitations.

IBR

Trace one design decision using the seven Intention–Behavior Record fields.

These cards describe the workflows. Platform access and installation details are provided in the guides above.

Availability & research limitations

ChatGPT is in restricted private testing. The Claude Public Screening v0.3.1 and Gemini Public Screening v0.1.5 packages are available for download. Direct native Gemini Skill sharing is not yet available in the researcher’s account.

PhygitAi provides preliminary conceptual research screening, not empirical validation or certification. Platform names and logos identify compatibility; they do not indicate endorsement or directory approval.

Fictional demonstration

Adaptive Reading Lamp

Fictional reading-lamp concept illustration

Concept-stage lamp for university students. Adaptive brightness was suggested by AI; no verified prototype observations are supplied.

Preliminary G0–G5 review

G0

Intention & Assumptions

Improve reading comfort; assumptions remain untested.

G1

AI-Assisted Co-Design

AI suggested adaptive brightness; original prompts are unavailable.

G2

Authority & Delegation

Human approval and manual override are planned; rules are undocumented.

G3

Physical Realization

Concept stage only; no verified prototype evidence.

G4

Disagreement Assessment

No observations support an intention–behavior comparison.

G5

Accountable Revision

No correction or retest is documented.

Provisional Intention–Behavior Record

Intention
Improve reading comfort.
AI Input
Suggested adaptive brightness.
Permission
Not documented.
Event
No observed event supplied.
Outcome
No verified observation.
Decision
Human approval pending.
Retest
Not performed.

Recommended next steps

Record AI contributions, define human authority and override rules, prototype, document observations, and plan retesting.

Fictional research demonstration. Illustrative screening only; not empirical validation, certification or independent verification.