Roadmap

Bodhi Map is meant to be kept current — released in small, regular increments, each highlighting what’s new (see Release Process). This page sketches the direction; nothing here is a promise, and contributions can reshuffle priorities.

Guiding intent

• Stay accurate and current — track genuinely important new techniques as the field moves, and mark them with the release highlight so returning users see them.
• Depth over sprawl — keep the graph clean, well-connected, and well-sourced rather than maximizing node count.
• Stay offline and dependency-free — every addition must keep the app self-contained.

Candidate directions

• Multimodal cluster — vision/language understanding (ViT, CLIP, VLMs, multimodal fusion) and possibly image generation/diffusion. Deferred from v0.4.0; the biggest scope decision, since it broadens “LLMs” toward “AI” as advertised.
• More frontier coverage — continue filling current-era gaps across architecture, efficiency, reasoning, and agents as techniques mature.
• Filtering & views — quality-of-life on top of the existing level/cluster/New filters (e.g. saved views, “core vs extended” presets).
• Authoring ergonomics — make contributing concepts even lower-friction.

How to influence it

Open an issue or PR on GitHub:

• Propose a concept (with a source) that’s missing.
• Flag something out of date or incorrectly explained.
• Suggest a feature.

Accurate, well-sourced additions that wire cleanly into the existing graph are the most valuable contributions. See Concept Authoring and Contributing.