Sol 5.6 is really cool and I've been getting so much done the last couple of days. I love that i can give it a big code task, set a goal that i want i to achieve, tell it to iterate until it's done and it just goes. It keeps going for hours until it has reached its goal. ๐
The eval layer marks that trace as healthy. The model did behave correctly given what it was shown. The data was wrong. Model evaluation and data verification are different units of analysis.
Agent calls a sanctions list via MCP. Valid response, name not found, payment approved. The list was 3 months stale. Supplier added 2 months ago. No error thrown anywhere in the stack.
@koeppelmann Sorry, but your friend became a victim of Russian propaganda.
I was born in Donets'k in 1983 and was living here until war started in 2014. Russian is my native language. I learned Ukrainian in school and didn't know anyone who speaks Ukrainian natively.
The story of a