Skip to content

Commit 7adb1e6

Browse files
author
jun
committed
docs(cursor): park wp2/wp4 residuals in a numbered doc
The roadmap named residuals in prose with no home (audit B14): effort ladders on a listed fast id, claude-4-sonnet-1m staying a real row, fastMode carrying two meanings, and the five pre-existing test failures that reproduce on a clean stash of this branch. Each records what would change the decision, so a later cycle does not rediscover them as new findings. Refs devlog/_plan/260902_cursor_unified_identity/040_residuals.md
1 parent bf5ac85 commit 7adb1e6

1 file changed

Lines changed: 47 additions & 0 deletions

File tree

Lines changed: 47 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,47 @@
1+
# Residuals
2+
3+
Known-and-accepted gaps, parked here rather than left in prose (audit B14). Each says what
4+
is wrong, why it was not fixed in its cycle, and what evidence would change the decision.
5+
6+
## R1 — effort ladders on a listed fast id (wp4)
7+
8+
`/v1/models` stamps `grokEffortFields(m.reasoningEfforts, …)` from the BASE row, but a fast
9+
variant's ladder can be shorter: `claude-opus-5` runs to `max` while its `fast` spec stops
10+
at `high` (`catalog.ts` CURSOR_CAPABILITIES). With `fastMode: true` a client could therefore
11+
request `max` against a listed `-fast` id.
12+
13+
Not fixed in wp4 because the resolver clamps: `cursorVariantEffort` picks the top rung the
14+
variant actually declares, so an over-request degrades to `high` rather than failing. The
15+
cost is an advertised rung that silently clamps, not a broken request.
16+
17+
Fix when: a user reports an effort selection that appears to do nothing on a fast id. The
18+
change is to thread the resolved variant spec into the listing branch instead of reading the
19+
base row's ladder.
20+
21+
## R2 — `claude-4-sonnet-1m` stays a separate row (wp2)
22+
23+
It is a real upstream wire id, not `claude-4-sonnet` + ultra, and `claude-4-sonnet` carries
24+
no `maxModeVerified` evidence — folding it would invent a capability. So "1M" still means two
25+
things in the picker: a synthetic ultra marker for `kimi-k3`, and this genuine second row.
26+
27+
Fix when: live `GetUsableModels` proves `claude-4-sonnet` supports Max Mode, at which point
28+
the row folds into the base the same way `kimi-k3-1m` did.
29+
30+
## R3 — `fastMode` carries two meanings (wp4)
31+
32+
One flag drives OpenAI's `service_tier: "priority"` and Cursor's fast VARIANT. These are
33+
different products with different ladders. The overload is deliberate — both express "go
34+
faster" — and is recorded so a later reader does not read it as an accident.
35+
36+
Fix when: a user needs one on without the other. That is a second flag, not a re-interpretation of this one.
37+
38+
## R4 — pre-existing red outside this unit
39+
40+
`bun run test:changed` at `42731a4be` reports 14461 pass / 5 fail. All five reproduce on a
41+
clean stash of this branch, so none is caused by this unit:
42+
43+
- `tests/cli-capabilities.test.ts` — "every management route is capability-covered"
44+
- `tests/…` CL-07 task effectiveness producer (4 tests)
45+
46+
Not this unit's to fix. Recorded so a later cycle does not mistake them for a regression it
47+
introduced.

0 commit comments

Comments
 (0)