Understanding LLMs · Ages 14+ & jobseekers
Open the black box of generative AI. Eight weeks from tokens and embeddings, through attention, the context window, and how a model samples each next token, to training, alignment, why it hallucinates, grounding with retrieved documents, prompting, agents, and judging output against a rubric. Lex the Lyrebird guides the way, with 15 deterministic, offline simulators (a tokenizer, an attention heatmap, a next-token probability lab, a training-loss curve, a grounding pipeline, an agent loop) so every idea is hands-on. Future-proofed on durable mechanics: no model names, versions, or prices. The mental model behind using AI tools well at work.
Standards alignment
- NEP 2020 §4.24: Artificial Intelligence as a contemporary subject (advanced)
- CBSE Artificial Intelligence skill subject (843) and beyond: how generative models actually work
- CSTA K-12: Impacts of Computing (IC) + Data & Analysis (DA)
- Industry literacy: transformer mechanics, prompting, retrieval-augmented generation, agents, and output assessment
Explain how a large language model works end to end, and reason about its prompts, limits, and costs.
Request access- 1Tokens and meaning
- 2Attention and the context window
- 3Predicting and sampling the next token
- 4The generation loop and what it costs
- 5How a model learns
- 6Truth and grounding
- 7Using LLMs well: prompts and agents
- 8Judging output and the big finish
The full week-by-week plan and every lesson in this course unlock with enrollment.