26/30First supervised mission
Flagship .ai supervised-workflow explainer and Academy mission chapter.
- fake ui
- causal edges need overlay
- idealized field site
Founder review lab · Five controlled studies
This is the evidence room for Starlight's next character language: different bodies, different optics, different media, twelve steward concepts, and a five-scene learning journey. Every image remains a research candidate until the role reads correctly, the prompt and provenance gates pass, and a human selects it.

Eligible for human review visually; blocked from canon because the exact prompt receipt is absent.
The design decision
A universal cute humanoid would be easy to recognize and impossible to trust at scale. Starlight needs a cast system: common material restraint and interaction ethics, with purpose-built silhouettes that make different kinds of work legible.
Adult humanoid or compact operator
Calm competence, accountable orchestration, and visible handoff
.ai profiles and product missions
Short-legged guide, soft companion, or original creature
Curiosity and encouragement in explicitly low-stakes learning
Academy lessons and onboarding
Protected operator or role-shaped instrument
Restraint, evidence, interruption, and clear escalation
Trust boundaries and governed workflows
Clean biotech, distributed swarm, or spatial system
A specialty whose body is sensing, coordination, or environment
.org stories, research, and system views
Advance the idea, not the render
These candidates scored at least 25/30. They earn a controlled next test on a real product surface; they do not yet earn a name, runtime claim, or production slot.
26/30Flagship .ai supervised-workflow explainer and Academy mission chapter.
25/30Knowledge navigator profile, route transition, or mission scout on .ai.
25/30Cosmos and planetary-observation profile or mission scene.
25/30Academy editorial chapter illustration.
25/30Academy design encyclopedia or morphology documentation.
25/30Infrastructure and operations profile.
25/30Media, energy, and resilience profile or field story.
25/30Editorial explainers, social cards, or .org learning pages.
Why no eye wins
Larger round eyes can create immediate warmth, but they can also make a serious operator look childlike and encourage people to attribute feelings, certainty, or moral judgment the system does not have. Some Starlight roles need joint attention. Others should remain an instrument with an aperture, witness light, or directional sensor.
The next test measures whether a viewer can identify focus, uncertainty, warning, handoff, and rest without a smile, caption, or cinematic glow. Gaze, posture, timing, and exact UI state must work together.
13 inspected candidates
Thirteen role-led bodies test when Starlight should be humanoid, compact, soft, airborne, instrumental, distributed, or architectural.
23/30Agent roster concept and orchestration introduction on .ai.
24/30Field-operator profile or workflow explainer on .ai.
23/30Academy onboarding chapter or beginner-guide story.
21/30Trust-boundary explainer only after redesign.
22/30Biotech or biomimicry story in the Academy.
21/30Evidence and provenance explainer on .org after deterministic overlays.
25/30Knowledge navigator profile, route transition, or mission scout on .ai.
24/30Fabrication and tool-building profile on .ai.
Micro-avatar and notification-state research only.
23/30Care and collaboration profile or Academy story.
21/30Archive-companion story vignette after morphology revision.
18/30Internal swarm-gestalt research only.
20/30Academy or operations establishing shot after architectural simplification.
10 inspected candidates
Ten controlled sensing architectures test joint attention, vigilance, adaptive sensing, and deliberately low anthropomorphism.
20/30Humanoid face research only; not a canonical identity.
22/30Routing-state or role-agnostic visor research.
21/30Analyst or perception-agent face study.
18/30Social-interface failure case and negative evidence.
24/30Guardian, privacy, or trust-boundary profile research.
24/30Biotech or living-systems face research.
20/30Creative or prism-optics face research only.
24/30Verifier or evidence-agent profile research.
18/30Juvenile mascot failure case and negative evidence.
22/30Premium humanoid face research only.
10 inspected candidates
Ten media lanes test where cinematic 3D, product editorial, gouache, line work, cel shading, stop motion, paper, and glass belong.
21/30Academy animation-still research.
23/30Academy game-narrative chapter or cinematic prototype.
24/30Academy interactive tutorial or game-state illustration.
23/30Academy stop-motion campaign or story vignette.
25/30Academy editorial chapter illustration.
25/30Academy design encyclopedia or morphology documentation.
25/30Editorial explainers, social cards, or .org learning pages.
24/30Academy chapter transition or paper-illustration prototype.
23/30Physical-product editorial prototype for .org.
21/30Campaign or material-language research only.
12 inspected candidates
Twelve constellation stewards test silhouette contrast and domain legibility across the full Starlight civilization.
19/30Internal ring-steward research only.
20/30Memory and provenance explainer only after deterministic overlays.
23/30Life-systems or biotech-enterprise story.
22/30Sound and culture profile after removing fake visualization.
25/30Media, energy, and resilience profile or field story.
18/30Internal negative evidence only.
25/30Cosmos and planetary-observation profile or mission scene.
23/30Ocean and living-continuity story after role clarification.
25/30Infrastructure and operations profile.
24/30Harness and research profile.
21/30Creation-to-distribution workflow explainer after deterministic overlays.
23/30Device safety profile or community-support vignette.
5 inspected candidates
Five wide scenes turn provisioning, values, evidence, supervised work, and constellation membership into a visible learning journey.
23/30Lifecycle story: human activation and provisioning gate.
24/30Lifecycle story: plural human-values deliberation.
23/30Lifecycle story: evidence archives and provenance review.
26/30Flagship .ai supervised-workflow explainer and Academy mission chapter.
22/30Lifecycle story: joining a constellation after major redesign.
What happens next
Research metadata only. No runtime authority, production approval, or exact prompt receipt is implied.