Skip to content

Commit 96954d2

Browse files
norandomclaude
andcommitted
research(annotation-driver): Rust sizing spike with exact parity (R65)
An isolated Rust package ports the driver's model-free core on the crate's pinned parser and reproduces the Python prototype exactly on five scopes: sites, mutants by hash, 98 controls, pack admission and classification. Core 898 canonical lines, admission 413 as ported, against 753 canonical lines of removable adapters. Corrects the earlier size estimate and the unchecked statement about the apparatus cap, whose counted set contains neither the adapters nor the deconvolution command code. Co-Authored-By: Claude <noreply@anthropic.com>
1 parent 101dc63 commit 96954d2

19 files changed

Lines changed: 3107 additions & 6 deletions

File tree

‎.kiro/handover.md‎

Lines changed: 5 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -23,9 +23,11 @@ has the result table, the open decisions, the owner's rules and the commands.
2323
2. **Owner decided (2026-11-10 evening):** the apparatus is ported to Rust,
2424
one runtime; breaking codebases must be avoided. Feasibility report:
2525
[docs/reports/agnostic-driver-feasibility.md](../docs/reports/agnostic-driver-feasibility.md).
26-
Owner approved: mutation and analysis ported together; sizing spike as
27-
the first design step; cap decided after it. **Requirements are
28-
generated (nine areas) and await the owner's approval;** then design.
26+
Owner approved: mutation and analysis ported together; requirements
27+
approved. **Sizing spike done (R65): exact parity with the prototype,
28+
core 898 canonical lines against 753 removable.** Waiting for the owner:
29+
whether the driver counts toward the 7,173-line cap (its counted set
30+
does not contain the adapters). Design continues with the batch runner.
2931
3. **Second priority, not to be forgotten (owner, 2026-11-10): Slice B,**
3032
hardening against LLM inference drift, in `constraint-guided-compression`
3133
(requirements Amendment 5). Not started. It starts when the driver's

‎.kiro/specs/agnostic-annotation-driver/prototype-gate.md‎

Lines changed: 15 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -573,6 +573,21 @@ run needs a new dated owner decision.
573573
port the kind check and one operator, count the lines); the cap is decided
574574
with that measurement. Requirements are in [requirements.md](requirements.md).
575575

576+
## Sizing spike, 2026-11-10 ([R65](../../../docs/research/2026-11-10/11-annotation-driver-rust-sizing-spike.md))
577+
578+
A Rust port of the model-free core, in an isolated package on the crate's
579+
pinned parser, reproduces the prototype exactly on the five scopes (sites,
580+
mutants by hash, 98 controls, pack admission, classification). Measured:
581+
core 898 canonical lines, pack admission 413 as ported, against 753
582+
canonical lines of adapters it makes removable; two grammar crates added.
583+
584+
Correction to "Result of attempt 2" and to the owner-decision note above:
585+
the statement that the apparatus is at its cap and that the port needs the
586+
cap raised or an equal removal was not checked against the cap's counted
587+
set. That set (R41) lists fixed files and contains neither the adapters
588+
nor the deconvolution command code. Whether the driver enters the counted
589+
set is the open owner decision; the batch runner is still unsized.
590+
576591
## Result of attempt 1, 2026-11-10
577592

578593
Outcome `revise`; full tables and limits in R61. No design approval and no

‎.kiro/specs/agnostic-annotation-driver/spec.json‎

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -36,7 +36,8 @@
3636
"docs/research/2026-11-10/07-annotation-driver-gate-result.md",
3737
"docs/research/2026-11-10/08-annotation-driver-parallel-batches.md",
3838
"docs/research/2026-11-10/09-annotation-driver-attempt2-setup.md",
39-
"docs/research/2026-11-10/10-annotation-driver-gate-attempt2.md"
39+
"docs/research/2026-11-10/10-annotation-driver-gate-attempt2.md",
40+
"docs/research/2026-11-10/11-annotation-driver-rust-sizing-spike.md"
4041
],
4142
"notes": "2026-11-10: owner set priorities (agnostic driver first, inference-drift hardening second) and approved a new spec with mutation as the first annotation consumer. Brief and prototype gate written for review; gate not run. No requirements, design or tasks generated. Owner go (proceed) 2026-11-10: gate checklist step 1 scope screening done (R57, gate Amendment 1: four scopes admitted, control expectation corrected). Arm D (rules as data, model as offline rule author) approved by the owner 2026-11-10. Owner requirement 2026-11-10: hard-coded operators to be removed completely (gate Amendment 2b). Step 2 done (R58): validator controls. Gate attempt 1 completed 2026-11-10 with outcome revise (R61): Rust, Python and Java pass in at least one arm, TypeScript fails in every arm; model-written packs admitted and equivalent to hand-written ones. 21 model calls. Attempt 2 ran 2026-11-10 after the owner chose the full revision (R63, R64): arm D (rules as data, model-written packs for held-out languages) meets every mechanism row on four languages and five scopes; arm A misses yield in one TypeScript run; apparatus ceiling is the open row and an owner decision. Owner decision 2026-11-10: port to Rust, single runtime, no second runtime; breaking codebases must be avoided (copies only, verified restore). Feasibility report: docs/reports/agnostic-driver-feasibility.md. Owner approved 2026-11-10 the port scope (mutation and analysis together), a sizing spike as the first design step, and the cap decision after the spike. Requirements generated and approved by the owner 2026-11-10 (basis: \"make sure we don't forget slice B this time. other than that approved\"); criterion 9.6 added to keep Slice B scheduled. Design not started; first step is the sizing spike."
4243
}

0 commit comments

Comments
 (0)