Predicting Expert Now Evaluations in Software Code Reviews

“””<p>I’m at Stanford and I research software engineering productivity.<p>We have data on the performance of &gt;50k engineers from 100s of companies.<p>Inspired by @deedydas, our research shows:<p>~9.5% of software engineers do virtually nothing: Ghost Engineers (0.1x-ers)<p>“””<p><a href="https:&#x2F;&#x2F;x.com&#x2F;yegordb&#x2F;status&#x2F;1859290734257635439?s=46" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;yegordb&#x2F;status&#x2F;1859290734257635439?s=46</a><p>Uses LLMs to build a classifier to evaluate the usefulness of commits on company-private Git repositories. The classifier has r=0.85 correlation with a 10-person panel of software engineer evaluators.