Explore the curriculum Find the failure you want to understand. Choose a curriculum subject, then move from Beginner foundations to Advanced interactions and Professional production judgment. Use content area when you want a narrower path.
Search stories and concepts ⌕ 1 · Subject All subjects AI evaluation, safety, and accountable operation Agentic and multimodal product systems Data, storage, retrieval, and memory Distributed state, coordination, and transactions Generative model behavior, adaptation, and inference Identity, security, privacy, and supply-chain trust ML systems and decision quality Performance, capacity, and technology economics Reliability, observability, and incident engineering Runtime, concurrency, and resource behavior 2 · Level All levels Beginner Advanced Professional 3 · Content area All content areas Current and attributable knowledge Delegated identity, approval binding, and containment Durable execution and authoritative workflow state Effect contracts, uncertain commits, and reconciliation Hierarchical budgets, deadlines, cancellation, and termination Operational thresholds, reviewer queues, and feedback bias Retrieval-augmented generation Retry amplification and recovery control Semantic observability and trajectory evaluation
5 stories
Data, storage, retrieval, and memory Beginner Current and attributable knowledge 10 min
The model passed every launch check. Then one policy sentence changed, and its confident answer stayed frozen in time.
Parametric vs non-parametric knowledge Retrieval-augmented generation RAG vs fine-tuning Indexing pipeline vs query pipeline +1
Read the incident → Reliability, observability, and incident engineering Advanced Retry amplification and recovery control 10 min
A fraud-review slowdown became a fleet-wide retry amplifier—and the AI orchestration layer multiplied the load before anyone saw the feedback loop.
Timeout Retry Exponential backoff Jitter +4
Read the incident → ML systems and decision quality Advanced Operational thresholds, reviewer queues, and feedback bias 16 min
The candidate ran safely in shadow mode. The incumbent still decided which field evidence could exist—and therefore which model looked better.
Decision policy Shadow evaluation Selective labels Sampling frame +1
Read the incident → Agentic and multimodal product systems Professional Durable execution and authoritative workflow state 12 min
The model proposed the right recovery. The human approved it. Then one lost cloud response split what the workflow knew from what production had already done.
State persistence and checkpointing Idempotency key Exactly-once semantics Fencing token +1
Read the incident → Agentic and multimodal product systems Professional Delegated identity, approval binding, and containment 25 min
Every service enforced its own narrow permission. Together, those permissions released one employee’s exact records to another person.
Privilege composition Source-to-sink authorization Record-subject binding Effect-bound approval +1
Read the incident →