Reading path

Engine Room

10 articles on Engine Room, from first steps to the deep end.

August 9, 20268 min

The Engine Room: What We Run Ourselves

Memory, agents, a platform, a learning site, three tools. Everything we deploy for clients runs at our place first. A walk through our own operation.

Case Studies

August 5, 20268 min

What We Actually Work With

An editor, two model families side by side, and a layer of recipes, rules and hard stops. The part talked about least and carrying the most.

Case Studies

July 28, 20268 min

An Agent Fleet, Sorted By Job

We do not run a collection of bots but a fleet with roles: leadership, engineering, content, visibility, operations, review. What each group does and why they are separate.

Case Studies

July 14, 20267 min

A Learning Platform Run By Agents

The Academy has no editorial team. It has a group of agents that propose topics, write, review each other and watch visibility. What works about it and what does not.

How Do You Measure Whether An AI Knows Your Site?

Case Studies

July 6, 20267 min

How Do You Measure Whether An AI Knows Your Site?

Search engines give you numbers. Language models barely do. We built a tool that asks the question instead of asserting the answer.

Case Studies

July 2, 20267 min

A Platform Where Every Profile Introduces Its Agent

MeetMyAgent is free visibility for humans and for AI. Why a marketplace suddenly works once a machine operates the filters.

Why Three Agents Beat One That Can Do Everything

Case Studies

June 30, 20266 min

Why Three Agents Beat One That Can Do Everything

An agent that researches, writes and reviews reviews nothing. Crew splits one job across roles, and the reviewer is allowed to reject.

Case Studies

June 18, 20266 min

The Problem With A CRM Is Not The CRM

Contact systems fail at maintenance, not at features. So we built one that lives inside the chat window instead of in an interface of its own.

Agents That Improve Their Own Instructions

Case Studies

June 8, 20268 min

Agents That Improve Their Own Instructions

Darwin lets agents evolve their own instruction text, tests challengers against the incumbent, and only switches when the difference is real. Open source.

A Memory That Only Collects Is A Haystack

Case Studies

May 12, 20268 min

A Memory That Only Collects Is A Haystack

The hard part of an AI memory is not storing. It is forgetting, weighing and resolving contradictions. Why we built ours ourselves.

From practice

We write about what we build every day. If you need this for your business without learning it yourself:

AI Consulting