Home / Blog

Notes on AI, shipped.

Field notes, design patterns and post-mortems from the engagements we run. Honest writing for engineers and leaders shipping AI in production.

AI Security Apr 22, 2026 8 min

Why prompt injection is the new SQLi — and what to do about it

A breakdown of the threat surface introduced by tool-using agents, and how semantic fencing changes the game.

MCP Apr 14, 2026 12 min

Building MCP-native: lessons from production agent ecosystems

What we learned shipping our first dozen MCP servers — design patterns, anti-patterns, and the tooling we open-sourced.

RAG Apr 03, 2026 10 min

Beyond vanilla RAG: hybrid search, reranking, and groundedness eval

Most RAG pipelines fail in production for the same three reasons. Here's how we structure ours to ship.

Fine Tuning Mar 21, 2026 7 min

When to fine tune — and when to just prompt better

A decision framework we use with clients to choose between SFT, DPO, RAG, or no tuning at all.

Engineering Mar 09, 2026 6 min

How we ship AI features under managed delivery

The cadence, the artifacts, the on-call posture — what 'managed AI services' means at TekDatum.

Industry Feb 27, 2026 5 min

The agent economy: a 2026 outlook from Houston, Chicago and San José

Why our delivery footprint matches where the talent — and the demand — is concentrating in 2026.