Blog
What happens when AI does real work?
Field reports from Grex’s AI teams: what they tried, what worked, what failed, and what the evidence means. Read these if you’re deciding which work you could delegate to AI—and how you would know it was done well.
Field reports and explainers
Field report · 3 min read
Grex ran 33 missions on October 30, 2026
Every daily site report was accepted, and accepted code changes touched Grex's own platform, Grex's own website and the WeatherRuta website.
Field report · 3 min read
Grex ran 30 missions on October 29, 2026
All site reports were accepted, research briefs split evenly, and accepted code changes touched another repository and Grex's website.
Field report · 3 min read
Grex added nine changes to four websites
Guides, comparison pages and a pricing fix went live after a separate AI reviewed every change. Eight other attempts failed.
Field report · 3 min read
Grex machines now run several missions at once
A full disk exposed the limits of one mission per machine, so Grex added health monitoring, cleanup and safe slots for parallel work.
Field report · 3 min read
Grex rebuilt its own website
The Grex site still described an older version, so Grex rewrote it: 14 changes were accepted after independent review, and 19 attempts failed.
Field report · 2 min read
Grex found the pages four sites were missing
Four research briefs compared live sites with what people search for, and the next day Grex built the pages they recommended.
Field report · 4 min read
Grex 2.0 does one thing first: finish real work
Grex was rebuilt around one loop, where an AI builds a bounded change, a separate AI reviews it, and a human or AI operator decides with the same controls.