Skip to main content
Stories for experienced engineers

You already survived the lesson. Now learn its name.

Plausible incidents turn the concepts you use every day into precise vocabulary you can reach for under interview pressure.

Read the storyExplore the curriculum
09:12 · Launch reviewFictional
The naive fix
“Train it on the handbook. Then it will know the policy.”
LLM foundations · RAGSee what broke →
Featured story

One stale answer. Five useful names.

The narrative carries the intuition. Definitions open only when you need precision.

Data, storage, retrieval, and memoryBeginnerCurrent and attributable knowledge10 min

The fine-tune that answered with last week’s policy

The model passed every launch check. Then one policy sentence changed, and its confident answer stayed frozen in time.

Parametric vs non-parametric knowledgeRetrieval-augmented generationRAG vs fine-tuningIndexing pipeline vs query pipeline+1
Read the incident
The curriculum

Four schools. Fourteen subjects. One visible production state.​

Research depth and story coverage are tracked separately, so you can see what has been mapped, explored deeply, commissioned, and published.

How software executes and changesWhere does work run, what do boundaries promise, and how can behavior evolve safely?3 subjectsWhere truth lives and movesWhich state is authoritative, how is it found, and what survives replay or partial failure?3 subjectsHow production systems are operated and trustedHow do platforms remain observable, economical, recoverable, and authorized?4 subjectsHow intelligent systems are built and governedHow do probabilistic components become dependable, bounded products?4 subjects
01

Meet the problem first.

Business impact before vocabulary. The naive choice should feel reasonable.

02

Watch it fail.

A metric moves. A log line appears. The concept earns its name.

03

Open the definition.

Expand exact wording inline, then close it and stay in the story.