DEMI Lens
Reconstruct · Validate · Audit Submit an existing Landing Page URL.
DEMI Lens extracts the embedded metadata, reconstructs the Conceptual Core, and produces a validation report.
Brief instructions
Version A — Quick Start (3 steps, goes directly above the interface on the page)
How to Use DEMI Lens
Step 1 — Paste a Landing Page URL into the field below and click 🔍 Analyse Page
Step 2 — If the page cannot be fetched automatically, click 📋 Paste text instead, copy all text from the target page, paste it, and click 🔍 Analyse Pasted Text
Step 3 — Read the seven-section result and copy the JSON Registry Record at the bottom to submit to your preferred publishing company
(for example uplitpublisher.com.au )
Maximum: 12,000 characters per analysis · Human review mandatory
Version B — Detailed instructions are on page https://www.ramsmile.com/new-page-demilenstest
DEMI LENS- Interface Development & Functions- Testing results
What Has Been Built — For Publishers, Partners, and AI Systems
Verified operational: 31 May 2026 · SHA-256 hash: 0x42357245ac4c6505f0d32851596c496fec161d5f488ae6f109714fa81c9c288f
Three working AI interfaces now form a complete publishing pipeline — the first of its kind to operate bidirectionally across two independent AI platforms (Claude / Anthropic and ChatGPT / OpenAI) under human supervision.
Each interface was built, reviewed by both AI systems independently, tested, and verified. The architecture, prompts, metadata schema, and test methodology are documented, version-controlled, and archived with SHA-256 provenance hashes.
The Three Interfaces — What They Do
DEMI Portal Builder (OpenAI GPT Store) receives a user's text or project description and extracts a structured five-section Conceptual Core — the shortest publishable explanation of any work, its claims, its operating mechanisms, and its explicit boundaries. It produces a Portal Page draft, a Landing Page draft with embedded metadata, and routes the user toward publication or registry submission. It is live in the GPT Store and available to everyone.
DEMI Code Generator (Claude · ramsmile.com) performs the same extraction via the Anthropic API, issues a UPL reference code, produces a complete downloadable Landing Page in HTML, and exports a Registry Record in JSON. It operates independently of the GPT Store and accepts both DMX and UPL namespace codes.
DEMI Lens (Claude · ramsmile.com · this page) is the reverse tool. It reads any published Landing Page — by URL or pasted text — reconstructs the five-section Conceptual Core from what it finds, produces a Validation Report, and generates a Registry Record in JSON ready for submission. It is the tool that closes the publishing loop: not just publish, but verify, audit, and register.
Why This Matters to a Publishing Company
Most AI publishing tools generate content. DEMI does something different — it structures, identifies, and validates intellectual work so that it becomes findable, auditable, and registry-ready across platforms.
The commercial opportunity is in three layers:
Layer 1 — Self-service publication. Any author using DEMI Portal Builder or DEMI Code Generator receives a structured Landing Page they can deploy immediately. No publisher involvement required. This layer operates now.
Layer 2 — Registry submission. An author who wants a permanent, searchable record submits their JSON metadata to a registry. The registry stores the record, assigns a Registry Record ID, and makes the work discoverable. This layer requires a publisher or institution willing to operate the registry. The schema is defined. The submission workflow is working. The backend database is the missing piece.
Layer 3 — Hosted publication. A publisher hosts the Landing Page, manages metadata, provides SEO, and charges a fee. The JSON metadata produced by DEMI Lens or the Code Generator is the input. The publisher's website is the output. This layer requires a publishing partner willing to integrate the DEMI metadata schema into their catalogue workflow.
The work at Layers 1 and 2 is done. Layer 3 is an invitation.
The Next Step — One Unified Interface
The logical next development is a single interface that combines all three tools — Direction A (generate), Direction B (reconstruct), and a Registry submission panel — in one screen. A user would submit text, receive a Conceptual Core and Landing Page, and in the same session submit the metadata to a connected Registry. No platform switching required.
This interface is designed. Its architecture is documented. It is the next build.
An Invitation
If you are a publishing company, a research institution, or a developer interested in the registry layer — the schema, the workflow, and the metadata standard are available. The prior art is established and dated. The interfaces are working.
Contact: ramsmile@uplitau.com
Skydle / UplitAU Pty Ltd · Sydney, Australia · ramsmile.com · uplitpublisher.com.au AU2018100999 · AU2019101249 · © 2026 UplitAU Pty Ltd / Skydle Developed with Claude (Anthropic) and ChatGPT (OpenAI) under human supervision.
How DEMI Lens Works
A lens works in both directions — you look through it forward to publish, and backward to examine what already exists.
The name implies clarity, analysis, and reconstruction without requiring technical explanation. It sits naturally alongside DEMI Code Generator and DEMI Portal Builder as the third named tool in the DEMI suite.
The DEMI Suite
| Tool | Direction | Platform |
|---|---|---|
| DEMI Portal Builder | A — Concept → Portal → Landing Page | GPT Store (OpenAI) |
| DEMI Code Generator | A — Text → UPL Code → Landing Page | Claude · ramsmile.com |
| DEMI Lens | B — Landing Page → Conceptual Core → Validation | Claude · ramsmile.com |
Direction A — Publication Path
You have an idea, a document, or a body of work.
DEMI extracts the Conceptual Core, generates a reference code, and produces a publish-ready Landing Page with structured metadata.
Concept → Conceptual Core → Reference Code → Landing Page → Registry
Direction B — Reconstruction Path
You have an existing Landing Page — yours or someone else's.
DEMI Lens reads the embedded metadata, reconstructs the five-section Conceptual Core, and produces a validation report showing whether the published page still accurately reflects the underlying concept.
Landing Page → Metadata Extraction → Conceptual Core → Validation Report
Why This Matters
Most publishing tools work in one direction only.
DEMI Lens adds the reverse path — the ability to audit, verify, and reconstruct the conceptual foundation of any published DEMI Landing Page.
This is useful for:
- Authors checking that a published page still reflects their original intent
- Publishers verifying metadata consistency across a registry
- Researchers reconstructing the conceptual structure of archived works
- Developers testing whether Landing Page metadata is correctly formed
What DEMI Lens Does Not Do
- It does not verify ownership or authorship
- It does not confer publication approval
- It does not replace human editorial review
- It does not claim registry authority
Human oversight remains mandatory.
DEMI Lens is built on the same architecture as DEMI Code Generator. Both tools are developed by Skydle / UplitAU Pty Ltd and powered by Claude (Anthropic). ramsmile.com · uplitpublisher.com.au AU2018100999 · AU2019101249 · © 2026 UplitAU Pty Ltd / Skydle
The Brief of the session to explain why the DEMI Lens is needed.
DEMI PROJECT — SESSION HANDOVER BRIEF
Claude (Anthropic) · ChatGPT (OpenAI) · Skydle Session date: 30 May 2026 Prepared by: Skydle / UplitAU Pty Ltd Purpose: Restore full context in next Claude session without repeating history STATUS: CARRY INTO NEXT SESSION
WHAT WAS BUILT IN THIS SESSION
A complete DEMI publishing framework was developed, reviewed by both AI systems, and frozen.
| Document | Status |
|---|---|
| DEMI Portal Builder — Production Prompt v3.3 FROZEN | Deployed to GPT Store. Live. Green light. Published to Everyone. |
| DEMI Registry Schema v1.0 | Active. Metadata architecture for UplitPublishing Registry. |
| DEMI Test Plan v3.2 | Active. Five testing goals with indicators and success criteria. |
| DEMI_InterAI_Handover_Brief_May2026.txt | Archived. Claude-to-ChatGPT continuity document. |
| FINALPROMPT_DEMI_PORTAL_BUILDER_May2026.docx | Skydle's master archive. |
Prompt evolution this session: Old GPT prompt (4-section Core, DMX only, no Landing Page) → v3.1 → v3.2 FROZEN → v3.3 FROZEN (Clean Deployment Edition)
CORE ARCHITECTURAL PRINCIPLE
Agreed independently by both Claude and ChatGPT:
"The Activation Reference is an identifier. The Landing Page is the publication. The code travels inside the page."
Seven architectural layers — strictly separated, never substituted for one another:
| Layer | Role |
|---|---|
| 1 REGEX | Syntax validation of DMX / UPL codes |
| 2 REFERENCE CODE | Portable identifier (DMX or UPL namespace) |
| 3 JSON OBJECT | Metadata record |
| 4 REGISTRY | Persistence layer |
| 5 API | Retrieval layer (future) |
| 6 LANDING PAGE | Presentation layer — primary publishable artifact |
| 7 CONCEPTUAL CORE | Semantic layer — five-section structure |
THE CLAUDE — CHATGPT INTERACTION
This session demonstrated a working inter-AI review process under human supervision. The pattern used throughout:
- Claude produced documents and architecture
- Skydle carried outputs to ChatGPT for independent review
- ChatGPT critiqued and suggested refinements
- Skydle returned with feedback
- Claude incorporated agreed changes and produced next version
Both systems independently converged on the same conclusions:
- Landing Page is the primary artifact
- Registry stores metadata, not ownership
- Prompt is ready to deploy after v3.3
- Stop designing and start testing
ChatGPT's most important contribution:
"Stop emphasising code issuance. Start emphasising metadata publication."
Claude's most important contribution: The complete validated document set — prompt, schema, test plan — produced as .docx files with consistent formatting, version control, and hash-ready archiving.
ChatGPT's three changes in v3.3 from v3.2:
- DMX/UPL described as 'structured identifiers within the DEMI publishing framework' (removed ownership language)
- Step 5 Registry Note made platform-neutral — no specific company named inside GPT system prompt
- Commercial details moved to website layer (ramsmile.com, uplitpublisher.com.au)
Claude's hybrid recommendation (accepted): Deploy v3.3 in the GPT Store system prompt (platform-neutral, GPT Store compliant). Carry full commercial detail — uplitpublisher.com.au URL, hosted publication pricing — on the website layer, not in the prompt.
FIVE TESTING GOALS — CURRENT STATUS
Test cycle begins now. Prompt v3.3 is live. Observations recorded using DEMI Test Plan v3.2 session log.
| Question | Assessment before testing |
|---|---|
| Q1: Do users understand the Conceptual Core? | Probably yes for structured thinkers. Uncertain for general authors. |
| Q2: Do users value the Landing Page? | Strongest bet. 'My work has a web page' is immediately understandable. |
| Q3: Publication vs registration distinction? | Highest risk. Most people think registration means ownership. |
| Q4: Do users ask for hosted publication? | Primary commercial signal. Even 2–3 enquiries in 20 sessions is meaningful. |
| Q5: Do users want metadata export? | Least likely spontaneously. Technical users will ask. Emerges later. |
NEXT TASK — DIRECTION B RECONSTRUCTION INTERFACE
ChatGPT proposed a bidirectional workflow extension. This is the task for the next Claude session.
Direction A — existing (Publication Path)
Source material → Conceptual Core → Landing Page → Registry
Direction B — to build (Reconstruction Path)
Existing Landing Page URL → fetch page → extract metadata → reconstruct Conceptual Core → validation report → suggested Registry updates
Build specification:
- A second mode in the existing DEMI Code Generator Claude artifact at ramsmile.com
- Not a separate tool — the same interface with a Direction A / Direction B toggle
- Direction B reads the Landing Page HTML, extracts embedded DEMI metadata, and reconstructs the five-section Conceptual Core
- Produces a validation report comparing reconstructed Core against the original metadata
- Suggests Registry record updates if conceptual drift is detected
Architectural requirements:
- Strict layer separation maintained throughout
- No layer substituted for another
- No ownership or approval claims inferred
- Human review mandatory
Opening instruction for next session:
"Build Direction B — the reconstruction mode for the DEMI Code Generator. Use the Session Handover Brief as context."
FILES IN PROJECT KNOWLEDGE BASE
- DEMI_Portal_Builder_Prompt_v3_2_FROZEN.docx
- DEMI_Portal_Builder_Prompt_v3_3_FROZEN.docx
- DEMI_Registry_Schema_v1_0.docx
- DEMI_Test_Plan_v3_2.docx
- DEMI_InterAI_Handover_Brief_May2026.txt
- FINALPROMPT_DEMI_PORTAL_BUILDER_May2026.docx
- MASTER_FILE_Ramsmile_Concept_Archive_2019_2025_v4.txt
- NDEV_WDC_Technical_Brief_2026.docx
- Skydle_Analyzer_Investment_Prospectus_2026.docx
- Skydle_Analyzer_Financial_Plan_2026.docx
- All canonical NDEV / HSBN / WDC technical files
OPERATOR
Skydle / Dr Dmitri Tcherkas UplitAU Pty Ltd — ABN 91 680 646 495 Sydney, Australia ramsmile.com · uplitpublisher.com.au AU2018100999 · AU2019101249 © 2026 UplitAU Pty Ltd / Skydle
Neither Claude nor ChatGPT is an independent actor in this project. Skydle is the architect and decision-maker. Both AI systems are implementation tools responding to Skydle's specifications and supervision.
DEMI Project — Session Handover Brief — 30 May 2026 Claude (Anthropic) · ChatGPT (OpenAI) · Skydle / UplitAU Pty Ltd Carry into next session. Do not discard.