main abstraction

Giving a language model your documents to answer from. The standard way to make AI useful on your company’s knowledge.

Retrieval breaks before the model does. Naive chunking cuts text mid-thought, answers sit spread across two documents, and your users ask questions in words the documents never use. Production RAG means chunking that follows document structure, embeddings in a vector database fused with keyword search, a reranker over the final candidates, and a groundedness test that checks every answer against the source documents.

Software that runs a language model in a loop: reason about the goal, call a tool, observe the result, repeat until done. The model decides at runtime which tool to reach for.

Demos skip the two hard parts: stopping and state. A production agent carries a hard iteration cap and a written definition of done, or it loops until the invoice arrives. Tools connect through open standards like MCP: one integration serves every agent. Every step lands in a trace, so when the agent does something strange at 2 a.m., you can read what it saw and what it chose.

A scored test for a non-deterministic system: real cases with known good answers, run against every change.

Teams without one review quality from memory: the demo looked fine, someone rewrote a prompt on Thursday, nobody can say whether Friday’s answers got better or worse. We ship a change when it beats the previous score and revert it when it does not. Monitoring picks up the cases the test set never met.

A check around a probabilistic system: on the way in (injected instructions, personal data) and on the way out (schema, policy, facts).

For agents, put the check on the action: inspect the tool call before it runs. Anything irreversible (send, delete, pay) passes an allowlist or waits for a person. Guardrails stay simpler than the system they guard: rules and small models. A system reviewing its own actions approves them.

Data moving from where it is created to where it is needed, on schedule, without a person in the middle.

A model answers from whatever the pipeline fed it that morning. Stale loads, silent failures, and duplicated rows all surface the same way: users tell you the AI is wrong. Production pipelines run idempotent, monitored, and re-runnable from any point in history, so the day the source API dies twice costs you one re-run.