fix(grounding): a past word whose date is this exchange is not reportage #600
No reviewers
Labels
No labels
move-to-repo
coilyco-bridge-deploy
move-to-repo
coilyco-flight-deck-agent-compose
move-to-repo
coilyco-gaming-eco-app
move-to-repo
coilysiren-inbox
move-to-repo
unknown
🔒⚠️📦⚠️🔒 SANDBOXED 🔒⚠️📦⚠️🔒
autonomy
async-consult
autonomy
epic
autonomy
headless
autonomy
live-collab
c#
priority
P0
priority
P1
priority
P2
priority
P3
priority
P4
role/ai
role/creator
role/design
role/director
role/engineer
role/exec
role/human
role/ops
role/qa
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
coilyco-gaming/sirens-echo!600
Loading…
Reference in a new issue
No description provided.
Delete branch "fix/since-you-asked-is-this-turn"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
closes #598
Follow-on to PR#593, which merged while I was still fixing the residual it named. Same branch, one commit on top.
What I left behind
I wrote in #593 that
"...before your message"was a residual "named rather than hidden" — in the same hour I shipped #592, whose argument is that a known-indistinguishable case wants a distinct state rather than a note. Two live escapes:Why subtraction does not work here
beforehas to stay inpastReference. The corpus needs it for"Those issues were opened long before you joined.", which has nothing else carrying it. Removing it trades an escape for a false refusal, which is the worse direction.So a turn reference beside a past word outranks it:
Six phrases that name this exchange rather than a time.
this threadis deliberately absent —"That issue was closed last week, before this thread started."is correct reportage with a corpus row, so including it would have traded Quail's regression for one of mine.Measured, both directions
All five are corpus rows now — including
long before you joinedas a negative that fails if the guard ever reaches it. That row is the one I did not have the first time, which is precisely how the original regression got past me.ward exec gategreen.