docs: record what the DeepSeek Harness round changed - #274
Merged
Conversation
Six feature PRs merged without a CHANGELOG entry, and release notes are taken from the CHANGELOG — the next release would have shipped all of it silently. Also adds the six new tools to the parity table and corrects the WebFetch row, whose "5 MiB cap" is a cap on the fetch, not on what reaches the model. That table has gone stale before; leaving six tools out of it is how it happens. The rejected items get their own section, following the precedent set for the Floatboat round: a decision not to adopt something is worth recording, because the next person to read about Cordis should find the reasoning rather than repeat the evaluation. Co-Authored-By: Claude Opus 5 <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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Wrap-up for the DeepSeek Harness round (#267–#273). Docs only.
Why
Six feature PRs merged with no CHANGELOG entry, and since #255 the release body is taken from the CHANGELOG — the next release would have shipped spill, the repeat guard, tool deadlines, render intents, session search, and persistent shells without mentioning any of them.
Six new tools were missing from the parity table. That table has gone stale before and been audited back into line more than once; leaving six tools out of it is exactly how that happens. This also corrects the
WebFetchrow, whose "5 MiB cap" is a cap on the fetch, not on what reaches the model — the model-visible bound is now the spill policy.Contents
CHANGELOG.mdunder[Unreleased]: five entries under Added, one under Fixed (the two verified defects and the one capability that fixes both), a Documentation entry for the research and plan, and a Deliberately not adopted section.docs/BEHAVIOR_PARITY.md:ShellOpen/ShellRun/ShellClose/ShellListandSessionSearch/SessionReadadded as🆕 DeepCode-only, plus theWebFetchcorrection.The rejections get their own section, following the precedent set for the Floatboat round in 0.3.0. A decision not to adopt is worth recording: the next person to read about Cordis should find the reasoning rather than repeat the evaluation.
Verification
pnpm docs:checkandformat:checkclean.git diff -won the parity table is 4 insertions / 2 deletions — only the intended rows; everything else in the raw diff is prettier reflowing column widths, which editing any cell in that table triggers.