WEBVTT

1
00:00:00.000 --> 00:00:06.833
This overview explains how a conceptual explanation
differs from a retained experimental result.

2
00:00:06.833 --> 00:00:12.466
A correct-looking animation cannot establish
elapsed time or a performance gain.

3
00:00:12.466 --> 00:00:12.467
A correct-looking animation cannot establish
elapsed time or a performance gain.

4
00:00:12.767 --> 00:00:17.633
Suppose a door opens only for a valid pass while access is enabled.

5
00:00:17.633 --> 00:00:21.833
Let A encode the pass and B encode enabled access.

6
00:00:21.833 --> 00:00:24.367
These are two separate requirements.

7
00:00:24.367 --> 00:00:30.467
We are choosing a small teaching rule so that every possible input can be checked.

8
00:00:30.467 --> 00:00:33.700
At zero zero neither condition holds.

9
00:00:33.700 --> 00:00:36.433
At zero one the pass is missing.

10
00:00:36.433 --> 00:00:39.433
At one zero access is disabled.

11
00:00:39.433 --> 00:00:42.100
All three cases deny entry.

12
00:00:42.100 --> 00:00:46.800
At one one both conditions hold, so the answer is one.

13
00:00:46.800 --> 00:00:53.367
Combining the factors A and B by and reconstructs exactly these four answers.

14
00:00:53.367 --> 00:00:57.600
The four matches establish this two-input Boolean identity.

15
00:00:57.600 --> 00:01:02.867
The drawing itself has not measured a compiler, a machine, or a request.

16
00:01:02.867 --> 00:01:06.267
Exact agreement is an answer about behavior.

17
00:01:06.267 --> 00:01:11.354
A speed claim needs additional evidence about an execution.

18
00:01:11.354 --> 00:01:11.367
A speed claim needs additional evidence about an execution.

19
00:01:11.367 --> 00:01:15.833
A drawing can explain a proposal without recording an experiment.

20
00:01:15.833 --> 00:01:20.467
For example, split A AND B into the factors A and B.

21
00:01:20.467 --> 00:01:25.200
An exact check confirms their conjunction for all 4 inputs.

22
00:01:25.200 --> 00:01:28.700
That establishes the displayed Boolean identity.

23
00:01:28.700 --> 00:01:33.333
It does not establish how long a program takes to find or use it.

24
00:01:33.333 --> 00:01:41.300
A retained measurement needs input identities, a procedure,
a timed boundary, measured results, and uncertainty.

25
00:01:41.300 --> 00:01:45.500
A stopwatch illustration supplies none of those records.

26
00:01:45.500 --> 00:01:48.333
Exactness describes the answer.

27
00:01:48.333 --> 00:01:52.433
Promotion describes a decision under a performance gate.

28
00:01:52.433 --> 00:01:57.528
A mathematically correct result may still be slower or remain unpromoted.

29
00:01:57.528 --> 00:01:57.533
A mathematically correct result may still be slower or remain unpromoted.

30
00:01:57.833 --> 00:02:02.500
Work a second complete case with A 1, B 0.

31
00:02:02.500 --> 00:02:06.200
Keep this assignment fixed throughout the calculation.

32
00:02:06.200 --> 00:02:10.867
Each symbol retains the meaning already introduced; a new value does not

33
00:02:10.867 --> 00:02:14.300
create a new logical rule.

34
00:02:14.300 --> 00:02:19.933
The valid pass contributes one, but enabled access contributes zero.

35
00:02:19.933 --> 00:02:22.867
Combining them with and gives zero.

36
00:02:22.867 --> 00:02:27.500
Reconstructing from the two separate factors gives the same zero.

37
00:02:27.500 --> 00:02:34.233
The failed access condition has not been removed by drawing the rule as two factors.

38
00:02:34.233 --> 00:02:38.033
This case establishes a specific contrast.

39
00:02:38.033 --> 00:02:41.067
Reconstruction preserves this denial.

40
00:02:41.067 --> 00:02:44.000
Exactness is distinct from speed.

41
00:02:44.000 --> 00:02:48.567
Keep the input, intermediate values, and conclusion together when comparing

42
00:02:48.567 --> 00:02:50.633
it with the previous case.

43
00:02:50.633 --> 00:02:57.033
The calculated answer belongs to this complete assignment under the stated rule.

44
00:02:57.333 --> 00:03:00.833
Imagine a run record for the same door rule.

45
00:03:00.833 --> 00:03:06.400
Label the source version and require all four output bits in a declared order.

46
00:03:06.400 --> 00:03:12.233
Mark whether the timer includes preparation, execution, and output delivery.

47
00:03:12.233 --> 00:03:19.100
This is a fictional record that demonstrates what must be recorded, not a new result.

48
00:03:19.100 --> 00:03:22.000
Use abstract cost units.

49
00:03:22.000 --> 00:03:26.900
The candidate spends eight units preparing and one executing.

50
00:03:26.900 --> 00:03:31.067
The control spends two preparing and two executing.

51
00:03:31.067 --> 00:03:35.333
The candidate has the cheaper execution step, but that alone does not

52
00:03:35.333 --> 00:03:38.467
settle the complete request.

53
00:03:38.467 --> 00:03:40.733
Nine is greater than four.

54
00:03:40.733 --> 00:03:44.467
Both methods may produce exactly the right vector while the

55
00:03:44.467 --> 00:03:48.300
candidate loses this complete-request comparison.

56
00:03:48.300 --> 00:03:52.933
Repeated readings, declared uncertainty, and a performance gate belong

57
00:03:52.933 --> 00:03:55.367
in the real experiment.

58
00:03:55.367 --> 00:04:01.567
Correctness and promotion therefore remain separate decisions.

59
00:04:01.567 --> 00:04:01.577
Check the mathematical claim independently.

60
00:04:01.577 --> 00:04:05.200
Check the mathematical claim independently.

61
00:04:05.200 --> 00:04:09.928
Use recorded measurements to support performance claims.