Trade-off first
Every unit starts from the product decision, then opens the mechanism only as far as the decision requires.
Phase 3 in progress · 9 of 20 units published
AI fluency for product decisions.
Libella is a mobile-native Android learning path for product professionals who need to understand LLM systems well enough to make build, buy, skip, scope, and risk decisions.
Decision-grade AI literacy, not AI hype.
Libella teaches LLM systems through product trade-offs: what changes, what breaks, what it costs, what should be measured, and what decision a product leader should make next.
The v1 Android app teaches LLM systems for product managers and product-side professionals: tokenization, context windows, latency, evals, model selection, hallucination, RAG, streaming UX, rubrics, and calibrated AI grading.
Calibrated reliability
Claims are sourced and confidence-tagged. Uncertainty is treated as a valid answer.
Path, not catalog
The core surface is a sequenced learning path. Glossary material supports the path rather than replacing it.
v1 path: LLM Systems for PMs
The canonical v1 path is a 20-unit sequence for product professionals learning how LLM-backed products behave in production.
Tokenization Published
Context Window Published
Latency Published
Evals Published
Model selection Published
Prompt design basics Published
Hallucination + reliability Published
Cost dynamics at scale Published
Fine-tuning vs. prompting vs. RAG Published
Vector search / RAG fundamentals Locked
Streaming UX Locked
Tool use, multimodal, agents, safety, operations Planned