WEBVTT

1
00:00:00.000 --> 00:00:03.193
The final challenge is to
demonstrate and defend your result.

2
00:00:03.518 --> 00:00:07.152
You have built a documented sensor-monitor
workflow with source control,

3
00:00:07.361 --> 00:00:12.005
project context, skills, tests,
replay, and a capstone feature.

4
00:00:12.446 --> 00:00:16.533
Now you will connect those artifacts to
the evidence that supports them. The core

5
00:00:16.580 --> 00:00:20.724
outcome is a complete private candidate
with reproducible software checks,

6
00:00:20.968 --> 00:00:23.813
not a public launch or a
physical kit demonstration.

7
00:00:24.254 --> 00:00:26.831
Begin with the source
identity and the requirement.

8
00:00:27.156 --> 00:00:30.941
Show the current candidate and explain
what changed from checkpoint eight.

9
00:00:31.266 --> 00:00:34.703
The capstone adds configurable
alert behavior and bounded

10
00:00:34.749 --> 00:00:38.012
commands while preserving
missing, nonfinite, stale,

11
00:00:38.035 --> 00:00:41.460
and recovery handling. A reviewer
should be able to understand

12
00:00:41.495 --> 00:00:44.839
the result without reading the
entire conversation that produced it.

13
00:00:45.350 --> 00:00:48.844
Present the changes as a small
sequence of engineering decisions.

14
00:00:49.355 --> 00:00:53.186
Explain the policy and
configuration slice, the parser,

15
00:00:53.360 --> 00:00:57.818
the host and firmware integration, and
the verification/documentation work.

16
00:00:58.097 --> 00:01:00.825
Point to coherent commits
or a reviewed diff.

17
00:01:01.208 --> 00:01:05.400
A large amount of generated source
is not evidence of quality by itself;

18
00:01:05.725 --> 00:01:09.498
the relationship between the requirement
and the final behavior matters.

19
00:01:10.009 --> 00:01:13.132
Defend the hysteresis
rule with a concrete case.

20
00:01:13.573 --> 00:01:17.045
Twenty-seven point five degrees
can be normal or alert because

21
00:01:17.080 --> 00:01:20.690
the prior non-fault state determines
retention inside the band.

22
00:01:21.015 --> 00:01:25.416
At twenty-eight, normal enters alert.
At twenty-seven, alert clears.

23
00:01:25.799 --> 00:01:29.572
Show which tests establish those
equality and history conditions.

24
00:01:29.955 --> 00:01:33.543
A statement that alerts work is
too broad without those details.

25
00:01:33.868 --> 00:01:35.691
Explain recovery next.

26
00:01:36.016 --> 00:01:40.137
Startup and fault both need two
consecutive fresh valid samples.

27
00:01:40.416 --> 00:01:43.853
An invalid reading or
stale gap resets the count.

28
00:01:44.178 --> 00:01:46.918
A valid threshold command
cannot clear fault because

29
00:01:46.953 --> 00:01:50.169
it changes configuration
rather than supplying a sample.

30
00:01:50.552 --> 00:01:55.091
After recovery, the state is classified
from fresh input and the entry threshold,

31
00:01:55.300 --> 00:01:59.793
not an old remembered alert. Show
the command contract and its tests.

32
00:02:00.072 --> 00:02:02.859
The complete bounded line must
match the promised grammar.

33
00:02:03.242 --> 00:02:07.119
Trailing numeric text, out-of-range
values, oversized input,

34
00:02:07.143 --> 00:02:10.533
and truncation are rejected
without changing the prior setting.

35
00:02:10.916 --> 00:02:15.200
Reset restores the default threshold
because settings are RAM-only.

36
00:02:15.525 --> 00:02:17.430
These are observable requirements,

37
00:02:17.604 --> 00:02:22.120
not optional implementation details hidden
behind a successful command example.

38
00:02:22.561 --> 00:02:24.976
Present actual host and replay receipts.

39
00:02:25.301 --> 00:02:29.214
Name the source, input, command,
result, and interpretation.

40
00:02:29.597 --> 00:02:32.465
Distinguish an expected fault
state during valid replay

41
00:02:32.523 --> 00:02:35.449
from a malformed fixture
and an assertion failure.

42
00:02:35.890 --> 00:02:39.675
Keep synthetic labels visible. A
reader should not have to infer whether

43
00:02:39.744 --> 00:02:42.287
temperatures were measured
or supplied by a fixture.

44
00:02:42.670 --> 00:02:46.815
Show the meaningful defect-rejection
evidence. The bad hysteresis

45
00:02:46.873 --> 00:02:51.122
and command variants should compile and
then fail the relevant behavioral checks.

46
00:02:51.563 --> 00:02:54.454
Preserve the good solution's
passing result as well.

47
00:02:54.733 --> 00:02:58.251
This demonstrates specific defects
the verification can detect.

48
00:02:58.576 --> 00:03:01.269
It does not prove the absence
of every possible bug,

49
00:03:01.513 --> 00:03:04.752
and an assertion count should not
be presented as that guarantee.

50
00:03:05.193 --> 00:03:08.641
Add the firmware-build receipt and
the fresh-checkout reproduction.

51
00:03:09.083 --> 00:03:12.728
The build supports compatibility
with the recorded target inputs.

52
00:03:13.170 --> 00:03:15.956
The second checkout tests
whether the distributed source

53
00:03:16.002 --> 00:03:19.520
and written setup are sufficient
without untracked local files.

54
00:03:19.845 --> 00:03:22.899
A clean status alone does
not establish reproduction;

55
00:03:23.177 --> 00:03:26.440
the required commands must
actually run in the fresh copy.

56
00:03:26.881 --> 00:03:29.006
Now apply the one-hundred-point rubric.

57
00:03:29.331 --> 00:03:33.035
Requirements and sourced hardware
facts receive fifteen points.

58
00:03:33.360 --> 00:03:38.120
Git history and review receive fifteen.
Firmware behavior receives twenty-five.

59
00:03:38.398 --> 00:03:41.011
Verification quality receives twenty-five.

60
00:03:41.336 --> 00:03:43.716
Documentation and handoff receive ten,

61
00:03:43.890 --> 00:03:46.653
and project instructions
and skills receive ten.

62
00:03:47.164 --> 00:03:52.110
Score using linked artifacts rather than
how polished the presentation feels.

63
00:03:52.493 --> 00:03:54.385
The suggested target is eighty,

64
00:03:54.594 --> 00:03:57.648
with mandatory gates that a
high score cannot replace.

65
00:03:58.089 --> 00:03:59.924
The build must reproduce.

66
00:04:00.307 --> 00:04:03.221
Critical software behavior
must meet the requirements.

67
00:04:03.662 --> 00:04:08.503
Host and replay checks must have actual
results. Evidence labels must be honest.

68
00:04:08.782 --> 00:04:12.068
If one of those gates is open,
name the missing evidence and next

69
00:04:12.137 --> 00:04:16.050
action instead of averaging the problem
away with points from other categories.

70
00:04:16.491 --> 00:04:19.254
Include the actual
fresh-session and skill records.

71
00:04:19.498 --> 00:04:22.552
Project instructions should
have guided a bounded task,

72
00:04:22.795 --> 00:04:25.721
and the skills should have
been evaluated on relevant,

73
00:04:25.837 --> 00:04:30.029
clean or unrelated, and
incomplete-input cases as appropriate.

74
00:04:30.412 --> 00:04:34.731
A valid skill file or a summary of
instructions is not a behavior result.

75
00:04:35.056 --> 00:04:36.913
If selection was unobservable,

76
00:04:37.122 --> 00:04:40.803
preserve that uncertainty separately
from the quality of the output.

77
00:04:41.314 --> 00:04:43.368
State the physical boundary plainly.

78
00:04:43.809 --> 00:04:47.943
No board measurement, wiring
result, upload, sensor accuracy,

79
00:04:48.152 --> 00:04:52.772
USB behavior, or physical timing
follows from the core software checks.

80
00:04:53.051 --> 00:04:56.244
Optional physical work has
separate procedures and evidence.

81
00:04:56.569 --> 00:04:59.251
It is not required for
the agreed core score,

82
00:04:59.460 --> 00:05:02.827
and it must not be implied by
a screenshot of passing tests.

83
00:05:03.337 --> 00:05:05.892
Prepare for one challenge
to your evidence.

84
00:05:06.170 --> 00:05:10.280
A reviewer might ask whether a passing
result belongs to the current source,

85
00:05:10.524 --> 00:05:12.776
whether the second checkout actually ran,

86
00:05:12.985 --> 00:05:16.108
or whether a failed mutation
was only a compiler typo.

87
00:05:16.491 --> 00:05:19.858
Answer by opening the relevant
receipt and explaining its scope.

88
00:05:20.241 --> 00:05:23.736
If the record cannot answer, preserve
the gap and perform the missing

89
00:05:23.794 --> 00:05:27.335
check instead of treating the
question as a presentation problem.

90
00:05:27.776 --> 00:05:31.410
Finish the presentation with your
transfer task and maintenance plan.

91
00:05:31.793 --> 00:05:36.797
Explain one observation about total
effort, including review and repair.

92
00:05:37.122 --> 00:05:39.932
Name a realistic trigger
for rechecking a tool,

93
00:05:40.141 --> 00:05:42.672
source, or skill and
the check you would use.

94
00:05:43.113 --> 00:05:46.340
The workflow should remain
useful after the course because

95
00:05:46.619 --> 00:05:49.522
you know how to maintain its
assumptions and evidence,

96
00:05:49.765 --> 00:05:52.366
not because you collected
a fixed set of prompts.

97
00:05:52.970 --> 00:05:55.756
Your submission is the
complete candidate bundle,

98
00:05:55.965 --> 00:05:59.622
scored rubric with evidence
links, transfer result,

99
00:05:59.901 --> 00:06:03.999
maintenance plan, and a concise
defense of these engineering decisions.

100
00:06:04.440 --> 00:06:06.426
If a result cannot be reproduced,

101
00:06:06.704 --> 00:06:09.932
preserve the failure and return
to the relevant bounded repair.

102
00:06:10.373 --> 00:06:14.669
Do not defend a passing claim that the
available evidence no longer supports.

103
00:06:15.110 --> 00:06:18.233
Pause here to complete the
challenge and final module quiz.

104
00:06:18.674 --> 00:06:22.819
Four correct answers out of five
is the recommended quiz target,

105
00:06:22.993 --> 00:06:25.721
with unlimited retry and explanations.

106
00:06:26.104 --> 00:06:28.206
Quiz progress is a self-check;

107
00:06:28.531 --> 00:06:32.478
practical completion depends on
the artifacts and mandatory gates.

108
00:06:32.862 --> 00:06:36.821
When those are satisfied, you have
demonstrated a repeatable way to use

109
00:06:36.879 --> 00:06:40.536
AI in embedded engineering while
keeping the source, decisions,

110
00:06:40.571 --> 00:06:44.959
and evidence understandable to the
next person who must rely on the work.