THE BOUNDEV JOURNAL
Better questions.
Better evaluation.
Practical guidance for AI teams buying expert code evaluation, response comparison, and rubric design.
LATEST PERSPECTIVE / Code Evaluation
Is your coding agent's pass rate real? How to audit the passes
A pass means the tests went green, not that the fix is right. What SWE-bench audits found, and a five-step way to check your own pass rate.
Read the article ↗Explore our services through the journal
Inter-rater agreement for LLM evals: is your human review reliable?
85% reviewer agreement can hide a weak eval. How to measure inter-rater agreement on LLM output with Cohen's kappa and Krippendorff's alpha, and what to do when it is low
Switching to a new LLM? Run a blinded side-by-side eval first
New models ship faster than teams can test them. Here is how to compare a candidate model against your current one on real traffic, blinded, before you switch.
How to write a rubric for grading AI-generated code
Passing tests is not enough. Here are the five criteria, written anchors, and agreement checks that make code eval scores trustworthy.
GO FROM THEORY TO PRACTICE
See an expert review, step by step.
Explore a sample report, a review rubric, and the evidence behind an evaluation.