Agreed future directions

Status: index of product ideas and plans that emerged in conversation. This is not a dated roadmap and does not turn these directions into scope for the current version.

1. Expanding the audience without splitting the product

A single downloadable SkillNet with two future modes:

• organization: company, class, academy, team, or another group with an owner and participants;
• individual: a single person administers their own space and studies their own courses, with no employees or talent features.

class is not a third mode and user is not used as a mode name. The current company/employee product is preserved. Full design: audience-modes.md.

2. Replacing and improving the mascot

The current mascot will change in the future. It must be a replaceable visual layer and must not contain the chat, voice, or node-reading logic inside it.

It also remains pending to improve how it understands node state and when it intervenes, because current behavior doesn’t work well. The concrete solution must be validated before being fixed. Brilliant’s companion can serve as a targeted reference for investigating that behavior, not as a general direction for the courses.

Agreed boundary: conversational-modalities.md.

3. Podcast Studio with NotebookLM-like quality

Start from the Audio Overviews experience: sources, focus, format, duration, language, background generation, transcription, and references. The implementation will be agnostic, modular, and configurable.

The default design may use two voices, but it will work with one or more depending on the provider’s capabilities. Grounding, editorial planning, script, review, voices, and finishing will be replaceable stages.

Full plan: podcast-studio-plan.md.

4. Audio as chat input

The person will be able to send a voice note to the chat. SkillNet transcribes it and responds in text. It doesn’t trigger TTS or open a call.

audio -> transcription -> chat -> text

Agreed boundary: conversational-modalities.md.

5. Live voice conversations

Realtime will be a separate feature for talking by voice. Even if GPT Realtime is one option, the integration must stay behind an abstraction to support other providers.

Its interface, concrete use cases, and implementation priority haven’t been fixed yet. Agreed boundary: conversational-modalities.md.

6. Separation of modalities

Web, audio, and video are cumulative modalities, not exclusive variants of a screen. Multimedia generation stays on-demand and separate from OpenUI composition.

Architecture design: delivery-modalities.md.

Status summary

This index should be updated when a direction is dropped, changed, or moves into an implementation phase. It should not be used as an automatic list of promised features.