Don't attribute to (#AI) #malice what can be attributed to (#human) #ignorance
A couple of days ago, I opened a #PR with a new BESSER feature touching both the backend and our front-end web editor. Built mostly with Cursor and Anthropic's Claude 4.0
The icon to represent the new feature was referenced in the code but missing in the hashtag#commits. And Armen Sulejmani assumed the LLM had #allucinated and created the icon it again.
But it wasn't that. It was just my mistake. I had forgotten to commit the new icon file. The hashtag#LLM was right and the hashtag#human (me) wrong.
I'm the first to point out all the limitations of current llms, such as their hashtag#biases. But this time it's us that were biased. #AI capabilities advance so fast that we blame hashtag#LLMs based on our past tests. (even if this "past" was just 3 months ago). We struggle to catch up with how fast they improve.
#LLMs are often wrong. But so are #humans. Be careful who you hashtag#blame next time!