About · who measures

About Makefield

Makefield is an independent research project on how accurately AI answer engines describe companies. I’m Denes Csaszar, and I run it from Berlin.

01 What it does

When someone asks an AI assistant about a company, an answer gets written, whether or not the company had any say in it. Makefield measures those answers: the correct answer is written down and locked before any engine is asked, each engine is asked repeatedly, and every answer is scored against the locked truth. The method sets out the rules, and the example reading shows what a complete reading contains.

02 What it never does

Makefield never does the fix and takes no cut from whoever does. It sells no ranking, no visibility and no dashboard. A measurement is only as good as the distance between the person measuring and the result, and that distance is the point of the project. Why it exists.

03 Where the work is published

• Studies: planned tests, each with its scope and its limits. The first, the floor test, is published on denescsaszar.com, my own work record.
• Field notes: single observations, each with its sources.
• The open course: the method, lesson by lesson.
• What changed: every change to a claim or a number since launch.

04 How it is run

One person, a few hours at a time, mostly at weekends. That sets the pace: few studies, published when their data and their limits are ready, and a study that finds nothing is published too. Everything here is free to read and cite.

For the next study I’m looking for two research partners.

05 Contact

Write to hello@makefield.io. The postal address is in the imprint.