Field notes, written from production traffic.

Posts here come out of builds that are running, not out of a content calendar. Voice agent architecture, retrieval and evaluation, vertical playbooks for specific trades, and the unit economics vendors do not publish. Every claim that can be sourced is sourced in the post.

What the archive covers

• Voice agent engineering — latency, barge-in, failure modes, eval design.
• Chat and retrieval — knowledge bases, deflection, handoff triggers.
• Vertical playbooks — HVAC, dental, legal, roofing, real estate and more.
• Agency economics — pricing, delivery, and what a build actually costs to run.

The editorial rule

If a number appears in a post it either has a citation or it is labelled as an estimate from our own traffic. Comparison posts have to concede where the competitor genuinely wins, because a comparison that never concedes anything is a sales page and readers can tell within two paragraphs.

Where to start

If you are evaluating vendors, start with the honest comparison across Bland, Synthflow, and Vapi. If you are building in-house, start with the eval harness post. If you run a service business and want the money argument, start with the real cost of missed calls.