Agent Loop Foundry.

Field reports.

Long-form, interactive guides to how the work actually gets done — the frameworks the Foundry runs, laid out in full. Read straight through or jump to any section.

A dark operations room at night, walls of rack-mounted equipment; a worn industrial robot sits at the console, reading the glowing dashboard.
Report 01 · 12 sections · 18 min read

The AI Conversion Engine

Why most AI programs stall, and how to build one that compounds.

Every company can buy the same models. The only durable advantage left is how fast you convert new capability into a measured, adopted, cost-optimized part of how the business runs.

Read the report →
A large paper process map spread across a table lit by a single lamp in a dark room at night, surveying tools resting on it.
Report 02 · 12 sections · 14 min read

The Agent Loop Foundry

Reconnaissance and loop design for the agentic era.

A small frontier team that turns undocumented business processes into compounding agent loops. Find the work, size it honestly, match it to the right architecture, and prove the first loop before anyone scales it.

Read the report →
A worn brass orrery on a cluttered workbench at night, lit by a single lamp; ringed orbits around a central sphere, hand-drawn orbit diagrams and loose gears on the paper beneath it.
Report 03 · 8 sections · 10 min read

Digging Gravity Wells

Human intent should be the center of gravity.

Agents borrow every goal they pursue. What holds a fleet of them on course is not supervision — it is the mass of human intent at the center. A stage-by-stage field guide to digging that well.

Read the report →
Dozens of small handwritten paper slips laid out in a grid on a dark workbench under a single lamp, three of them lifted out and pinned upright on a wire stand, with calipers, a magnifier and a ruled tally sheet beside them.
Report 04 · 8 sections · 13 min read

Market Research for Machine Answers

The category is measuring it like SEO. It behaves like a customer study.

AI visibility tools are not scarce, and the best of them are growing quickly. What the category inherited is SEO's mental model: a list of queries, a position, one number that goes up. But an assistant does not rank you. It describes you — differently every time somebody asks. Describing is what research measures, and research is designed, not tracked.

Read the report →