FullduplexFullduplex/blog
§E · community

Community.

Fullduplex is an observatory — and, as the name suggests, a reply channel. The live room is on Discord: demos, listening sessions, benchmark run-notes, and the occasional hallway-track debate.

live · ambientJP / ENopen to everyone · moderated
live reply channel

Join the Discord

Real-time chat, voice-agent demos, and listening sessions — with a dedicated JP room for Japanese readers. Where the community actually hangs out.

accept the invite
§01 · submit an entry

Submit a model, benchmark, or dataset.

The fastest path is to ping us on Discord or email hello@fullduplex.ai. Include the type of entry, a first-party source URL, and the measurement date for any numbers you're quoting.

  1. Choose an entry type: model / benchmark / dataset / correction.
  2. Send the first-party source URL (official docs, paper, or release page) so we can verify before listing.
  3. State the measurement date for any scores or sizes. Use unknown if you're not sure — we'd rather say nothing than mislead.
  4. We aim to publish updates within one editorial week and write back with either a live URL or a follow-up question.
model

Register a new STS model, audio-LLM, or voice-agent framework.

benchmark

Add a new evaluation, or update a leaderboard score.

dataset

Add a corpus, or correct a license note.

correction

Fix numbers, links, or copy on an existing entry.

reach
Discord
The main reply channel · JP room inside
reach
Email
hello@fullduplex.ai