Learn - AI Stack
Learn what people mean by an AI stack, how modern AI systems are typically structured, which patterns are showing up in the market, and what trade-offs sit behind the design choices.
AI systems are no longer just a model and a prompt. Teams now combine models with retrieval, orchestration, memory, execution environments, safety controls, and application layers to build systems that can do useful work reliably.
This guide is designed to help you understand the space in a practical way: start with the basics, move into architecture and reference patterns, then look at the major trade-offs and our view.
Recommended reading order
If you are new to the topic, read these in order. The section is arranged to move from definition, to structure, to examples, to decision-making.
What is an AI Stack?
Start here: what the term means, why it matters now, and how AI stacks relate to older ideas like RAG, LangChain, microservices, and MLOps.
Open topic ->Architecture
The major layers of a modern AI stack - plus where cross-layer ideas like harnesses fit without becoming their own layer.
Open topic ->Reference Stacks
How published stacks and real-world patterns map to the AI stack: five-layer models, agentic RAG, GStack, and Paperclip.
Open topic ->Key Trade-offs
The main design decisions teams run into: local vs cloud, single-agent vs multi-agent, structured workflows vs autonomy, model fleets, and async execution.
Open topic ->Our View
After the explainers: our current view on what is genuinely useful, where teams overcomplicate things, and how to think practically about AI stack design.
Open topic ->What you should come away with
By the end, you should have a clear mental model of what an AI stack is, which layers matter most, what common setup patterns look like, and where the main architectural trade-offs actually sit.