release: 0.3.0 - #19
Merged
Merged
Conversation
Minor rather than patch: the default model changes, which changes what every new install transcribes with. VERSION plus the seven fields scripts/stamp-version.sh propagates it to. Both CLIs report `dnt 0.3.0 (6938aa8, 2026-08-25)`. The near-miss gate was re-run because the model changed, which is the condition docs/RELEASING.md names. It came back 40/48 matched with 2 regressions rather than the 0 the suite asks for; the numbers and what they do and do not mean are in the pull request rather than buried here. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015XKX6zEAGiZFE5wAxJgYbV
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Minor rather than patch: the default model changes, which changes what every new install transcribes with.
What's in it
gemini-3.6-flashas the default, on a truncation result rather than a quality oneTruncationGuard— catches a transcript too short for the speech in the recordingDefaultPolicybugdocs/INCREMENTAL.md— an investigation that ended in the feature not being builtVersion stamping
VERSION→ 0.3.0, propagated byscripts/stamp-version.shto all seven fields. Both CLIs reportdnt 0.3.0 (6938aa8, 2026-08-25). Suites on the stamped tree: 548 Swift, 468 C#, Android green.The near-miss gate, recorded rather than buried
Re-run because the model changed, which is the condition
docs/RELEASING.mdnames:person-name: "send the draft" → "send a draft"real-version-number: "…including Gemini 1.5" → "…including Gemini 2.5"Eight focused passes on that second case, per model:
gemini-3.5-flashgemini-3.6-flash3.6 delivers the correct answer twice as often and regresses four times as often. Not a contradiction: "regressed" is defined against the no-context baseline, and 3.5 mishears the number regardless, so it has less correct baseline for context to break. By delivered accuracy — what reaches the user — 3.6 wins that case 50% to 25%.
Shipped on the maintainer's explicit call with these numbers in hand. The
score-review.pydictation gate remains unrun for 3.5-vs-3.6 and is the outstanding follow-up.🤖 Generated with Claude Code
https://claude.ai/code/session_015XKX6zEAGiZFE5wAxJgYbV