Insights

How we design and ship AI products clients can trust after launch — copilots, retrieval, and agent workflows.

RAG
LLM
Eval

Jun 12, 2026·8 min read

Shipping RAG that survives production

Retrieval quality, evaluations, and failure modes that show up the week after launch — not in the demo.

Copilot
UX
Trust

Apr 28, 2026·7 min read

Designing copilots people actually trust

The interface is the product. Confidence, citations, undo, and when to keep the human in control.

Agents
Architecture
Ops

Mar 9, 2026·9 min read

Agent workflows that fail safely

Tools, timeouts, and human gates — how to let an agent act without letting it wander.