WallMock — Catalog Director

What the company appears to be: WallMock — Catalog Director is a self-reported tool for high-volume wall-art sellers that automates visual merchandising by generating mockup packs from uploaded artwork collections using deterministic geometry and GPT-5.6 for orchestration.

What changed: The project evolved from an idea to a working prototype during the OpenAI 2026 hackathon, with a focus on reducing manual labor in creating visual listings for wall art. It uses AI not just for generation but also for reviewing and validating outputs through adversarial loops.

The single most important open question: Is there evidence of traction or early adoption by actual sellers? The description states no revenue, customers, or adoption data beyond the demo and self-reported claims.

• The description states that WallMock — Catalog Director is a tool for high-volume wall-art sellers.
• It generates mockup packs (Hero, Lifestyle, True Scale, Product Detail, Options Card) from uploaded artwork collections.
• It uses GPT-5.6 as an engineering orchestrator and adversarial reviewer.
• The system analyzes each artwork, ranks certified rooms, solves physical size and placement, applies product construction, and composites exact pixels into scenes.
• It verifies fidelity and geometry before delivering outputs.
• It supports resumable collection planning with diversity-constrained fallbacks and deterministic artifact IDs.
• It includes fail-closed commercial authority and catalog certification.

Not evidenced: No information on actual product delivery, user interface, or integration points beyond the demo.

• The description states that WallMock asks a different question than existing visualization tools: instead of manual placement for each listing, sellers upload once and software directs the whole visual merchandising job.
• It positions itself as solving repetitive work for sellers with hundreds of designs.
• The author claims that the strongest use of GPT-5.6 was not a single prompt but a disciplined loop involving repository analysis, implementation, execution, refutation, visual review, and evidence-backed acceptance.

Inference: This suggests a shift from generative AI as a tool to generative AI as part of a structured validation process — an evolution in how AI is used within the product lifecycle.

• The description states that WallMock targets "high-volume wall-art sellers".
• These are described as individuals or businesses who repeat the same production work for every listing, including choosing rooms, deciding ratios, sizing pieces, placing them relative to furniture, selecting frames, and exporting multiple channels.
• The system aims to reduce manual labor in this process.

Not evidenced: No specific customer segments, personas, or buyer intent data are provided. No evidence of whether these sellers currently exist or have been engaged.

• The description does not state any pricing model or business model.
• It mentions future features like secure managed billing and durable jobs, implying a monetization path may be planned.
• There is no mention of revenue streams, subscriptions, usage fees, or licensing models.

Not evidenced: No evidence of pricing, monetization strategy, or customer acquisition costs.

• The system is built with TypeScript on Node.js using Sharp/libvips for image processing.
• Room geometry, homographies, frame construction, shadows, product footprints, and fidelity checks are described as deterministic and test-pinned.
• Generative models (Codex + GPT-5.6) were used for orchestrating tasks, reconciling legacy behavior, drafting specs, implementing and reviewing fail-closed catalog authorities, diagnosing bugs, generating law tests, and inspecting visual evidence.
• The system intentionally uses disagreement: one pass builds a change and another tries to refute it before owner approval.
• Challenges included ensuring trustworthiness of results; the system now measures and withholds untrustworthy outputs based on spatial, rights, or release authority.

Inference: This indicates a strong engineering focus on reliability over pure automation — suggesting an approach that balances AI creativity with structured validation.

• The description states that the app generates four mockups in 19.2 seconds.
• Exact-art SSIM scores are reported as 0.9512, 0.9651, 0.9669, and 0.9716.
• It includes role geometry for crop-safe Hero, truthful True Scale, and coded Product Detail.
• It supports resumable collection planning with diversity-constrained fallbacks and deterministic artifact IDs.
• A review trail records rejected sources and preserves the live catalog until explicit owner approval.

Not evidenced: No data on actual users, adoption rates, or performance in real-world use cases. No evidence of product maturity beyond demo-level functionality.

• The description states that existing visualization tools are useful editors but still require manual decision-making for each listing.
• WallMock is positioned as a solution to the inefficiency of manual placement and repeated visual work.
• It contrasts with tools that do not automate the full workflow from upload to mockup delivery.

Not evidenced: No mention of direct competitors or competitive advantages beyond automation. No evidence of market size, competition landscape, or differentiation strategy.

• The description is entirely self-reported and unverified.
• There is no evidence of revenue, customers, or traction beyond the demo.
• The project is described as a hackathon submission with only one team member (A K).
• The system relies heavily on GPT-5.6 for orchestration, which raises questions about scalability and consistency without human oversight.
• The claim that AI makes code and source imagery cheaper but reliability comes from accumulated decisions and constraints implies potential complexity in maintaining quality at scale.

Inference: If the tool is not yet adopted or monetized, it may be premature to assess its viability as a commercial product.

1. What specific wall-art sellers have you engaged with? Have they provided feedback?
2. How do you plan to scale beyond the current demo-level functionality?
3. What is your path to monetization and customer acquisition?
4. Can you demonstrate how the system handles edge cases or incorrect inputs?
5. How does the system validate that generated mockups meet real-world seller expectations?
6. What are the technical limitations of relying on GPT-5.6 for orchestration?
7. Are there any legal or compliance concerns around using AI-generated content in visual merchandising?