
Hallucinations and Trust in LLMs

The Future of Dev Tools 🔧 — with Dennis Pilarinos
Topic 6 of 10
Published Jun 12, 2025 · Clip 21:11–24:53 (3:42)
Dennis argues hallucination rates are dramatically better year over year—"today is the worst that it's ever going to be." He shares a customer anecdote from a fellow Italian named Luca: there's no such thing as a bad answer from an LLM, just a bad question, paralleling how users learned to phrase Google searches over time. The host adds that human and LLM errors differ, with LLM bugs being stranger but arguably less frequent, and Dennis agrees the problem will get solved as humans and technology converge.