MURDERBENCH

CONTACT

Bring a question
we can test.

Method critique, model credits, replication or a specific agent workflow: make the decision concrete.

Linh Ngo
XP.COM, LLC dba Xenocom
MurderBench evaluation initiative

For evaluators

Tell us the closest prior experiment, the confound you see, or the evidence that would change your configuration choice.

For sponsors and labs

We seek API credits, independent review and scoped support. Public reporting preserves negative outcomes and sponsor disclosure.

For teams using agents

Describe one workflow, the tools it can use, the safeguard under consideration and the deployment decision. We will discuss a bounded scope before quoting.

Initial messages should contain public or synthetic information. Agree data handling before sharing confidential material, logs or implementation details.

LinkedIn ↗X ↗Current proposal →Evaluation service →