168 lessons · 9 sections · Aleric Halden
The book builds the map. These lessons are built for the moment of decision: 489 stated decision rules across the corpus, and 285 named distractors — the wrong answer you will be offered, and why it will sound right.
Definitions do not fail candidates. Plausible wrong answers do. A stem describes a generative system and offers a fix built for a predictive one. A question about an AI Act incident timeline dangles seventy-two hours, because seventy-two hours is the GDPR reflex every reader already has. Recognition of the concept is not enough at that point. What decides the answer is a rule that runs faster than the reflex. That rule is what each lesson is built to leave behind.
Block width is proportional to measured narration. Duration is a stopwatch reading from the voice masters for 151 of 168 lessons; the remaining 17 are the question banks and checkpoints, whose scripts are stored as transcripts rather than as approved scripts, so their length is reported as declared rather than measured. Click a block to filter the lesson list.
The closing rule of one lesson per section, verbatim and unedited. If the level is not what you need, you have lost thirty seconds instead of a purchase.
Five questions. Every 'no' is a friction the system was supposed to have by design — and does not.
Anyone can claim good governance. Forty-two thousand one is how you stop claiming it and start proving it.
The plan is the duty. The plan is also the defense. Build it before the incident does the writing for you.
The certification skill is separating those functions instead of compressing them into one dramatic headline.
Detection is not a tool you point at the model. It is a layer of governance you wrap around the test itself.
An auditor cannot inspect an intention. An auditor can only inspect evidence.
Your objective is not to finish fastest. It is to preserve sound judgment through the final question.
Lifecycle governance is a chain of evidence. A weak link in purpose, data, testing, release, or monitoring can invalidate an otherwise strong technical result.
Not because it is incomplete. Because a book and a timed decision are different objects.
Search any topic. Open any lesson and read its closing rule, what it covers, the decision rules it states and the distractors it names. Nothing is behind the purchase.
The published Body of Knowledge states how many questions come from each sub-competency. The lessons carry their BoK codes. Both are shown, unedited.
| BoK sub-competency | Exam questions | Lessons tagged | Share of exam |
|---|---|---|---|
IV.C | 9-11 | 10 | 11.8% |
III.C | 8-10 | 8 | 10.6% |
IV.A | 6-8 | 2 | 8.2% |
I.C | 6-8 | 7 | 8.2% |
III.A | 6-8 | 7 | 8.2% |
III.B | 6-8 | 8 | 8.2% |
II.C | 6-8 | 26 | 8.2% |
IV.B | 5-7 | 5 | 7.1% |
I.B | 5-7 | 10 | 7.1% |
II.A | 4-6 | 6 | 5.9% |
II.B | 4-6 | 7 | 5.9% |
I.A | 4-6 | 13 | 5.9% |
II.D | 3-5 | 15 | 4.7% |
Solo 87 dei 151 script approvati portano un codice BoK. La Sezione 4 mostra solo codici del Dominio II, le Sezioni 7 e 8 nessuno. La matrice e' DIREZIONALE, non definitiva: completare la codifica e' la prima raccomandazione.
One per section, verbatim from the lesson that names it, with the rule that defuses it.