crenshawdev/burnrate
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix(report): say what the composition bands hold, and pin their weights #7

Merged
john merged 3 commits from fix/composition-label-wording into main 2026-08-01 13:05:19 +00:00
Conversation 0 Commits 3 Files changed 5 +36 -6
john commented 2026-08-01 13:05:15 +00:00
Owner
Copy link

The ×1.25 composition band is computed as cc - cc1h, which is every write NOT reported as 1h rather than every write known to be 5m. Transcripts predating the cache_creation breakdown report cc1h = 0, so their 1h writes sat inside a band labeled 5m. The card's note also never named its unit while the daily chart beside it did.

The card recomputes billed-equiv in JavaScript, so the weights are a two-copy fact: be_of() and the viewer. Nothing in the suite read the viewer's copy, which is the seam the 1h weight already broke on, so it is now pinned to be_of by a test that fails when the two disagree.

Also: the ask output nests cache_write_1h inside cache_write, and now says so where a reader looks.

Ships as 1.0.2.

The ×1.25 composition band is computed as `cc - cc1h`, which is every write NOT reported as 1h rather than every write known to be 5m. Transcripts predating the `cache_creation` breakdown report `cc1h = 0`, so their 1h writes sat inside a band labeled 5m. The card's note also never named its unit while the daily chart beside it did. The card recomputes billed-equiv in JavaScript, so the weights are a two-copy fact: `be_of()` and the viewer. Nothing in the suite read the viewer's copy, which is the seam the 1h weight already broke on, so it is now pinned to `be_of` by a test that fails when the two disagree. Also: the `ask` output nests `cache_write_1h` inside `cache_write`, and now says so where a reader looks. Ships as 1.0.2.
The composition band is computed as cc - cc1h, which is every write NOT
reported as 1h, not every write known to be 5m. Transcripts predating the
cache_creation breakdown report cc1h = 0, so their 1h writes sit inside a
band labeled "5m" -- the label claims a TTL the data never carried.

The card's note also never named its unit, while the daily chart beside it
does. Both charts plot billed-equivalent tokens; only one said so.
The card recomputes billed-equiv in JavaScript, which makes the weights a
two-copy fact: be_of() and the viewer. Every existing assertion reads the
Python copy, so the viewer's could go back to pricing a 1h write at 1.25x
and the suite would stay green -- which is how that seam broke before.

The ask output has the same shape of trap: cache_write is the whole
cache_creation total and cache_write_1h is the half already inside it, so
a reader who adds them double-counts. Said so where the reader looks.
The plugin cache is version-keyed, so a release that leaves this string
alone reaches nobody. Nothing here changes a number: the composition
card's two labels now say what the bands hold, and the viewer's copy of
the billed-equiv weights is pinned to be_of by a test.
john merged commit 045420fcd5 into main 2026-08-01 13:05:19 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: crenshawdev/burnrate#7
No description provided.
Delete branch "fix/composition-label-wording"

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?