This Humanoid Robot Is a Terrifyingly Competent Office Intern
Swiss startup Flexion Robotics has developed a simulation-first training method that allows humanoid robots to autonomously perform complex office chores.

Google's newly deployed AI Overviews feature began failing to return standard search results for specific command-like words on Friday, instead displaying conversational chatbot acknowledgments that effectively hid relevant web links. Users searching for terms like "disregard" or "ignore" were met with blank text boxes or conversational replies rather than definitions or search results.

The glitch highlights the unpredictable behavior of integrating large language models directly into core search infrastructure. While Google pitched the AI-driven overhaul as its most significant upgrade in decades, the system's inability to distinguish between a user's search query and a system instruction has drawn widespread criticism online.
According to reports, typing the word "disregard" into the search bar prompted the AI Overview to respond with phrases like, "Got it! Let me know if you need help with anything else." Another test returned, "No problem at all! How can I help you today?"
The search engine treated the query as a conversational command to ignore previous instructions rather than a search term. The issue extended to other synonymous terms. When users searched for the word "ignore," the AI Overview generated a message stating:
"Message received! I’m here and ready to help. What would you like to focus on today? Just let me know if there’s a specific topic, task, or question you’d like to explore."
The failure mode appears to stem from prompt injection vulnerabilities inherent to large language models. Because the AI model processes the user's search query as part of its prompt context, command verbs like "disregard" or "ignore" trigger the model's instruction-following behavior. Instead of summarizing web pages about the word, the model executes the command.
While traditional search results like dictionary definitions still exist on the page, the AI Overview pushes them far down. Users must scroll past a massive block of empty space or conversational filler to find actual links. For many users, the broken AI response is the only visible content above the fold.
The bug arrived immediately after Google's annual I/O developer conference, where the company announced a massive restructuring of its search results. Google showcased an ambitious AI vision designed to change how people interact with the web. The update prioritizes AI-generated summaries over the traditional list of blue links.
Elizabeth Reid, head of Google's Search division, called the update the biggest upgrade to the search box since its debut. However, the aggressive rollout has forced conversational interfaces onto users who may only want quick, direct information. The "disregard" bug illustrates how easily these conversational elements can disrupt basic search utility.
The search engine's new behavior has sparked significant frustration among users who feel the core utility of the web is being degraded. There is a growing interest in alternative platforms as users seek to escape forced AI integrations. Many users are actively looking for ways to bypass the AI-driven interface.
Unlike Google's aggressive implementation, competitors like Microsoft's Bing have taken a more conservative approach to AI summaries. The contrast has led some long-time tech observers to note that alternative engines are temporarily delivering cleaner, more predictable results. Google has not yet detailed a permanent fix for the instruction-word override bug.
Subscribe wiring is coming soon. For now, follow the daily news feed or connect on LinkedIn for updates.