Back
RCreddit.com
20
·17 hr ago·RSS
Not on the current live radar

Most LLM features ship without the engineering discipline we'd never skip for regular software

View original

Heat trend

New
Latest 24h versus previous 24h · 7-day curve

The percentage is based on available heat signal, not comment count or independent people.

AI summary

Many LLM features are deployed without the rigorous engineering discipline typically applied to regular software. This often involves minimal testing, a lack of versioning, and evaluations based on intuition rather than concrete metrics. Consequently, issues can arise weeks later, with no clear understanding of what changed or when, due to the absence of initial measurements. This concern is highlighted by Bruno Gonçalves, PhD, founder of Data For Science, who trains engineers at Fortune 500 companies in this crucial discipline.