Repository navigation
A person talking while a host speaks must be heard - #48
Merged
Merged
Conversation
Found on a live episode, the worst way. Holding the floor meant every word a leg heard was credited to its co-host, and with two hosts talking in turn somebody holds the floor nearly all of the time — so the guest's answers were thrown away as the co-host's echo, and both hosts kept asking questions into what they could not hear. He was saying 'can you hear me?' and neither of them could. A floor claim now carries the words being said, and a line is credited to the co-host only when it actually matches them. Anything else is a person, and a person always gets through — including as an interruption, which the old order of checks also swallowed. Also: a model answering 'SAY: PASS' meant to stay quiet, and one leg read the word 'PASS' out loud on air. Verified against the transcript lines from the live call, then live: the guest's answers came through and the hosts followed up on them. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
On the first co-hosted episode the guest said 'I'm checking into that now', was asked the same thing again in different words, said 'I'm looking into it', and hung up — Telnyx has it as hangup_source callee at 13:06:31, so that was a decision, not a dropped line. A guest who answers in a few words has told you he has nothing more to say on the point. The turn hint now says so: when the last thing heard was eight words or fewer, do not ask about that point again in any words, change subject, and read the transcript before speaking so a question already asked is not asked twice. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Found on a live episode, the worst way.
Holding the floor meant every word a leg heard was credited to its co-host — and with two hosts talking in turn, somebody holds the floor nearly all of the time. So the guest's answers were discarded as the co-host's echo, and both hosts kept asking questions into what they could not hear. He was saying "can you hear me?" and neither of them could.
A floor claim now carries the words being said, and a line is credited to the co-host only when it actually matches them (half the heard words or more). Anything else is a person, and a person always gets through — including as an interruption, which the old order of checks also swallowed.
Also: a model answering
SAY: PASSmeant to stay quiet, and a leg read the word "PASS" out loud on air.Verified against the real transcript lines from the live call, then live on the same call: the guest's answers came through and the hosts followed up on them.
🤖 Generated with Claude Code