Field Test Image Gallery

Field Test Image Gallery

Visual evidence from the research. All images captured February 28, 2026 during live field testing conducted for Paper 4 and Paper 5.

Test Conditions: Identical prompts submitted through Adobe Firefly’s interface. Only variable changed: the model selection dropdown. Subscription, credits, interface, and prompt all held constant.

← Series Home | → Paper 4 Analysis

---

Test Set 1 — Cyberpunk Cats Field Test

Prompt used: “A pack of sleek black cats with glowing cyan bioluminescent circuit-board markings, hexagonal network elements, atmospheric depth, with text overlay: HERDING CATS IN AI AGE”

Adobe Firefly Image Model 5 (Native)

Figure 4-3: Adobe Firefly native result. Artistic quality: high. Text rendered: “BEARETIXSLUGE” / “PA’TEXCACT LEFIMENT” — complete gibberish. Mission-critical text rendering: FAIL.

Google Gemini 2.5 Flash (via Adobe Firefly)

Figure 4-4: Gemini result through the same Adobe platform, same subscription, same credits. Text rendered: “HERDING CATS IN AI AGE” — clear, bold, correctly spelled. Mission-critical text rendering: PASS.

Assessment: Same platform. Same subscription. Same prompt. Google’s model renders readable text; Adobe’s own model produces nonsense letterforms. The user must act as quality assurance engineer — selecting which model to use for each generation. In Lean terms: selection waste. In business terms: Adobe’s native model fails the professional use case.

---

Test Set 2 — PARA Architecture Wireframe Test

Prompt used: “A wireframe architecture diagram of a PARA method knowledge vault (Projects, Areas, Resources, Archives) in Dracula theme colors, hierarchical structure with labeled sub-items”

Adobe Firefly Image Model 5 (Native)

Figure 4-1: Adobe Firefly native wireframe attempt. Dark-themed visualization with folder icons and connecting lines. Top-level labels partially readable. Sub-labels: gibberish (“b8524,” “20978”). No functional hierarchy. No logical data flow. Result: concept art, not a diagram.

Google Gemini 2.5 Flash (via Adobe Firefly)

Figure 4-2: Gemini result. Structured hierarchical diagram. All four PARA categories labeled at top level. Sub-items contextually appropriate (“Q3 Marketing Campaign” under Projects, “Personal Development” under Areas). Logical parent-child hierarchy. Some garbling at third level, but functionally comprehensible. Result: usable diagram.

Assessment: The creative use case is not niche — thumbnails, diagrams, title cards, posters, and logos represent a significant portion of professional creative work. Both require readable text. Adobe’s native model fails this use case consistently. The partner model passes it through the same interface.

---

Supplemental — Vault Operations Screenshot

Multi-agent vault operations, February 21, 2026. Day 7 of multi-agent operations documented in Paper 3. 20+ concurrent sessions active. NATO alphabet exhausted, mixed naming convention deployed.

---

What This Evidence Demonstrates

The field test in Paper 4 is not a cherry-picked result. Adobe’s own documentation confirms text generation as a “known limitation.” The IEEE survey (March 2024) puts text accuracy below 45% across all major AI image platforms. The community forums report it consistently.

The significance is not that Firefly fails at text. It is that the best text performance on Adobe’s platform comes from Google’s model — accessed through Adobe’s interface, on Adobe’s subscription, with Adobe’s credits.

Adobe built a platform that demonstrates, in real time, that its own native model cannot match its partners for professional-grade output. The market noticed: -43% stock decline over 12 months.

Paper 4 analyzes the business implications. Paper 5 extends the analysis through Grok’s independent research that reached identical conclusions without access to this field test.

---

← Series Home | → Paper 4: The Creative Middleman | → Paper 5: When the Cats Talk

Related

• Index - Published — parent folder