About.
Deep-Agnt is where AI engineering is practised as it is done: agents built, broken and fixed, judged by what they do on real runs.
An agent is a loop, not a reply.
It calls a model, runs a tool, reads the result, decides, and goes round again. So a problem here is rarely a function in a box: it is a loop to close, a tool to describe, a run to read.
Every step is evidence.
When a run goes wrong you should be able to point at the turn where it diverged, see what went in and what came out, and fix that. Problems are graded on runs for the same reason.
Restraint keeps the loop legible.
The site is black and white, with colour kept for meaning: pass, fail, a tool call, a result. A stream of agent events is easier to read in a quiet frame.
Who makes it
Deep Machine Learning LLC, the makers of Deep-ML, where machine learning is practised the same way: by writing it. One account signs you in to both sites.
Questions, problems you would like to see, or something broken: info@deep-ml.com. Common questions are answered in the FAQ.