feat(perf): #263-A — the measurement instrument (CALIBRATION_ONLY; no admissible baseline yet) by PhysShell · Pull Request #351 · PhysShell/Own.NET

feat(perf): #263-A — the measurement instrument, calibrated and firewalled

claude

committed

f09fa00

evidence(perf): the CALIBRATION_ONLY validation report

claude

committed

3118095

fix(perf): a contended runner is a result, not a job failure

claude

committed

e9d3bce

docs(perf): the Windows finding is sharper than "the runners are noisy"

claude

committed

790c872

fix(perf): the D7 gate must verify a freeze, not a file that echoes it back

PhysShell

and

claude

committed

5d6dab4

evidence(perf): re-record the calibration against the repaired instrument

PhysShell

and

claude

committed

35befb2

fix(perf): four review findings — the instrument was measuring the wrong things

claude

committed

55aad67

evidence(perf): re-record the calibration on a toolchain that is actually there

claude

committed

0590ea7

fix(perf): the harness digest named the checkout, not the instrument

claude

committed

2feb496

evidence(perf): re-record the calibration against the content-addressed digest

claude

committed

438a567

fix(perf): a phase list must mean the interval did those things

claude

committed

889c285

evidence(perf): re-record the calibration against the corrected phase taxonomy

claude

committed

dfdce55

fix(perf): a runner that will not hold still is not a broken instrument

claude

committed

b3080e4

fix(perf): three axes for reproducibility, and three names for one constant

claude

committed

6aacbc5

evidence(perf): preserve the n=5 sizing evidence, and record that n=5 is intermittent

claude

committed

ad4c302

evidence(perf): the calibration of record, at n=15 on a clean tree

claude

committed

63e57db

fix(perf): warmup is not a sizing knob, and a pair needs both its halves

claude

committed

6c33e2a

evidence(perf): run A of both calibration pairs, measured on a clean tree

claude

committed

742dad7

evidence(perf): both sizing pairs failed; the calibration of record is withdrawn

claude

committed

29deb17

fix(perf): a pair is one experiment, or it is two

claude

committed

b64ecee

evidence(perf): re-record both pairs under the corrected identity procedure

claude

committed

388598e

fix(perf): implement D7 §6 literally — two commits, not two sections

claude

committed

637a619

evidence(perf): re-record both pairs under the D7 §6 and reference-identity repair

claude

committed

2128ec4

fix(perf): the D7 anchor is provenance, and a preregistration is not three strings

claude

committed

aaf0d42

fix(perf): stop git housekeeping racing the fixture teardown

claude

committed

f8e7112

evidence(perf): re-record both pairs under the anchor and protocol repair

claude

committed

055b8e4

fix(perf): close the boolean N/A escape hatch and check the whole cell universe

claude

committed

43b6a4a

evidence(perf): re-record at 43b6a4a; n=5 failed and ships as it landed

claude

committed

34f7ef3

fix(perf): D7's metrics are not the instrument's attribution taxonomy

claude

committed

843d6fe

evidence(perf): re-record at 843d6fe under the D7-vocabulary repair

claude

committed

0374793

docs(perf): ratify the D7 metric mapping, and preregister Round 6

claude

committed

71c5832

feat(perf): Round 6 metrology helper and driver, no measurement yet

claude

committed

8c61a7b

evidence(perf): Round 6 setup pass — the calibrated duration ladder

claude

committed

4f7171c

evidence(perf): Round 6 metrology dataset and its reading — outcome O2

claude

committed

406bc85

Navigation Menu

feat(perf): #263-A — the measurement instrument (CALIBRATION_ONLY; no admissible baseline yet) - #351

feat(perf): #263-A — the measurement instrument (CALIBRATION_ONLY; no admissible baseline yet)#351
PhysShell wants to merge 34 commits into
mainfrom
claude/p-022-analysis-wiring-mwhqlw

Commits on Sep 10, 2026

feat(perf): #263-A — the measurement instrument, calibrated and firewalled

evidence(perf): the CALIBRATION_ONLY validation report

fix(perf): a contended runner is a result, not a job failure

docs(perf): the Windows finding is sharper than "the runners are noisy"

fix(perf): the D7 gate must verify a freeze, not a file that echoes it back

evidence(perf): re-record the calibration against the repaired instrument

fix(perf): four review findings — the instrument was measuring the wrong things

evidence(perf): re-record the calibration on a toolchain that is actually there

fix(perf): the harness digest named the checkout, not the instrument

evidence(perf): re-record the calibration against the content-addressed digest

fix(perf): a phase list must mean the interval did those things

evidence(perf): re-record the calibration against the corrected phase taxonomy

fix(perf): a runner that will not hold still is not a broken instrument

fix(perf): three axes for reproducibility, and three names for one constant

evidence(perf): preserve the n=5 sizing evidence, and record that n=5 is intermittent

evidence(perf): the calibration of record, at n=15 on a clean tree

fix(perf): warmup is not a sizing knob, and a pair needs both its halves

evidence(perf): run A of both calibration pairs, measured on a clean tree

evidence(perf): both sizing pairs failed; the calibration of record is withdrawn

fix(perf): a pair is one experiment, or it is two

evidence(perf): re-record both pairs under the corrected identity procedure

fix(perf): implement D7 §6 literally — two commits, not two sections

evidence(perf): re-record both pairs under the D7 §6 and reference-identity repair

fix(perf): the D7 anchor is provenance, and a preregistration is not three strings

fix(perf): stop git housekeeping racing the fixture teardown

evidence(perf): re-record both pairs under the anchor and protocol repair

fix(perf): close the boolean N/A escape hatch and check the whole cell universe

evidence(perf): re-record at 43b6a4a; n=5 failed and ships as it landed

fix(perf): D7's metrics are not the instrument's attribution taxonomy

evidence(perf): re-record at 843d6fe under the D7-vocabulary repair

docs(perf): ratify the D7 metric mapping, and preregister Round 6

feat(perf): Round 6 metrology helper and driver, no measurement yet

evidence(perf): Round 6 setup pass — the calibrated duration ladder

evidence(perf): Round 6 metrology dataset and its reading — outcome O2