Read the original at Lenny’s Newsletter ↗lennysnewsletter.com · subscriber post
By Lenny Rachitsky · lennysnewsletter.com · @lennysan on X · YouTube · LinkedIn
A glossary explaining 20+ common AI terms in simple language, from models and LLMs to transformers, training, RLHF, RAG, evals, agents, and hallucinations. It aims to help readers follow AI conversations in meetings and product work, framing jargon through plain analogies and linking to external primers.
Subscriber post — summary only01Key takeaways
- Most AI jargon maps onto a few core ideas: pre-training, fine-tuning, RLHF, prompt engineering, and retrieval at run-time.
- RAG supplies current or task-specific context at inference time, which is a common way to reduce hallucinations.
- Evals act like unit tests for AI products, defining what good output looks like and catching regressions.
- Agents sit on a spectrum, becoming more agentic as they act proactively, plan, use real tools, and loop on their own output.
- MCP is an emerging open standard that lets models connect to external tools without custom integration code for each one.
02Frameworks mentioned
Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.