ull-stack WordPress developer from Belgrade, Serbia, with 20+ years as a software engineer specializing in performance optimization, custom development, and integrating WordPress with complex systems. A long-time WordPress community contributor, he volunteers in translations and organizes and speaks at local meetups and regional WordCamps. Currently at GoDaddy, he's focused on connecting conversational AI with WordPress to shape more intuitive, human-centered site-building experiences.
[ENG] How to Stay Sane When Your AI Is an Improv Performer with a REST API
LLMs are the most confident unreliable narrators we've ever put into production. They'll invent court cases in beautiful prose, cite papers that don't exist, and tell you the capital of Australia is Sydney with the same tone they use for things they actually know. The scary part isn't that they're wrong — it's that they're wrong in a voice we've spent a hundred thousand years learning to trust. We will explore bad sides of LLMs — hallucination, bias, silent failure, and the confidence trap — and what critical thinking actually looks like when your main source of information has started freestyling in complete sentences. Equal parts for the engineers wiring these things into products and the PMs writing the specs, with jokes at the model's expense and none at yours.