# 01 — Research synthesis

Researched at execution time (2026-08). Sources linked below.

## Direct / adjacent competitors

| Product | Expectation it sets | Strength | Weakness / gap |
|---|---|---|---|
| [Lottie](https://airbnb.io/lottie/) | Designer-authored loops | Craft, toolchains | No AI mind model; you invent state mapping |
| [Rive](https://rive.app/) / [AI SDK Persona](https://elements.ai-sdk.dev/components/persona) | High-polish stateful persona | WebGL2, variants | Heavier; context limits; not drop-in MIT menagerie |
| [OrbUX](https://github.com/vikast908/ORBUX) | Many loaders × agent states | Broad state set (streaming, tool-calling…) | Loader catalogue vs character companions |
| [Orbsona](https://www.npmjs.com/package/@accidental-revenue/orbsona) | Identity + semantic states | Voice levels, circular mask | Different API philosophy |
| [beUI agent loading](https://beui.dev/components/agents/loading-states) | Honest text/progress primitives | Truthful progress language | Not a creature companion |
| Nothing / CSS spinner | Default | Zero cost | Dead UI; trust gap |

## User language (from repo guides + launch kit)

- “Why does it feel dead while it thinks?”
- “Is the AI demo real?” → local vs live
- “Can I use it commercially for free?” → yes for 8 MIT
- Avoid: revolutionary, fake proof, buy theatre

## Authoritative standards

- WCAG 2.2 AA contrast, focus visible, target size guidance
- `prefers-reduced-motion`
- Decorative animation `aria-hidden` + live regions for captions (already FAQ)
- Performance: lazy GPU, DPR cap, offscreen pause (partially shipped)

## Adopt / adapt / avoid

| Decision | What |
|---|---|
| **Adopt** | Four-state mind as public API; honest local/live; offline activation; lazy gallery contexts |
| **Adapt** | Clearer free-vs-try labeling (OrbUX-style state honesty without becoming a loader zoo); progressive disclosure of stack tabs |
| **Avoid** | Cloning Persona/Rive chrome; fake dashboards; purple SaaS gradients; DRM theatre that contradicts npm file; present-tense wrapper claims |

## Opportunities for this iteration

1. Fix AA-failing faint text.
2. Make activation path louder and stepwise after first demo delight.
3. Surface brand + design docs for agents/humans.
4. Tighten capability language on pricing and export tabs.
5. Improve empty/error copy on demo and downloads.