Meta's Muse can help you browse. What does it ask in return?
WIRED's Reece Rogers found Muse useful for browsing and shopping. He also saw repeated prompts to connect personal data. Meta disputes his take.
AI AgencyRead story
Lead story
An AI lie detector is a budget request, not proof
The Pentagon has asked for $30.3 million over five years to build Polygraph+. Experts warn that body signals do not reliably prove a person is lying.
Source: MIT Technology Review · Sep 25, 2026Read original source
AI Agency
Read story
More top stories
HelpCoach asks students to name the help they need
A study with 40 college programming students found more focused questions and better retention when HelpCoach guided their requests. It did not test K-12 classrooms.
Covered by both lensesRead story
Featured
Fewer false alarms can help teachers see the real ones
MagicSchool's team says its AI review tools cut confirmed false alarms by 99%. The test did not prove that every serious failure in real classrooms would be caught.
Covered by both lensesRead story
Research & Science
Teachers set the chatbot's goal. Does it follow it?
In workshops with 27 middle-school teachers, chatbot behavior matched intended learning goals less often than its tone or response style. This was not a student-outcome study.
Beyond PromptingRead story
Culture & Human Agency
A polished AI answer can hide the step that made the claim
In one study of a personal AI tool, a reviewer found that some claims came through a middle summary that the final citations did not show. The case is limited.
AI AgencyRead story
Beyond Prompting
More from this edition
-
Research & ScienceRead story
Scientists checked the AI's work in tools they knew
Ten biologists in an observational study checked AI-assisted analysis against familiar views and their own data. It does not prove the method works for everyone.