Blog

What we learned building it

Working notes from production AI work — how we measure an agent before it ships, what retrieval actually costs, and the failures that were worth writing down.

How to measure an AI agent before it ships
A demo proves an AI agent can work once. An eval harness proves how often it works and what it does when it fails. Here is the smallest one worth building.
SEP 15 · 8 MIN READ
RAG vs fine-tuning: Why knowledge and behaviour differ
The choice between RAG vs fine-tuning depends entirely on whether your model lacks facts or formatting. Here is how to measure what you actually need.
SEP 15 · 13 MIN READ
How much does it cost to build an AI agent? The drivers
When evaluating AI agent development cost, the final number is dictated entirely by the structural engineering choices made in the first week of work.
SEP 15 · 10 MIN READ
Why does my RAG chatbot hallucinate? The exact fixes
A retrieval-grounded chatbot making things up is almost never a model problem. It is a retrieval, prompt, or data issue. Here is how to fix RAG hallucination.
SEP 15 · 12 MIN READ

Showing 4 of 4

What we learned building it

How to measure an AI agent before it ships

RAG vs fine-tuning: Why knowledge and behaviour differ

How much does it cost to build an AI agent? The drivers

Why does my RAG chatbot hallucinate? The exact fixes