Study reports AI advice lowered accuracy while raising confidence

Sources: The Next Web, HN discussion
Summary: A study covered on 2026-07-19 reports that participants who received AI advice on a task were less accurate yet more confident in their answers. The finding drew a large Hacker News thread on how AI assistance affects critical thinking.
Comments: HN commenters debate the study's task design and generalizability, and several relate it to their own experience of over-trusting confident but wrong assistant output.
Why it matters: Developers increasingly route decisions through AI assistants, and evidence that assistance can raise confidence without raising accuracy bears on how coding agents are reviewed and trusted.