🤖 Google’s Search Bar Changes for the First Time in 25 Years

🔍 Google’s Search Bar Changes for the First Time in 25 Years

🍪 The Wafer-Scale AI Chip 20x Faster Than Anything

📍 6 AI Stock Sectors

🛠️ 5 Trending Tools

🥪 4 Brief Bites

💰 3 Funding Frontlines

💼 4 Job Opportunities

Google Search has been the dominant search engine for nearly two decades, processing 16.4 billion search queries per day. But answer engines, like Perplexity AI, disrupted this dominance by serving up AI-assisted answers to your search queries, eliminating the hassle of sifting through endless blue links.

To compete with the rising popularity of answer engines, Google Search introduced “AI Overviews,” which generate clear, concise, and complete summaries at the top of Google Search results. Then, they introduced “AI Mode,” which allows you to ask complex, multi-part questions by leveraging a “Query Fan-Out” technique, which simultaneously runs multiple related searches across several subtopics, pulling from hundreds of data sources.

Now, Google’s iconic search bar is an intelligent search box that dynamically expands so you can describe exactly what you’re looking for. It’s designed to anticipate your intent, formulating AI-powered search suggestions that enable “Search agents” to find exactly what you need, and it’s even capable of performing agentic transactions to browse, book, and buy on your behalf.

The Pew Research Center examined 900 U.S. adults who performed 68,879 search queries, finding that “AI Overviews” caused a nearly 50% reduction in website clicks. In other words, it’s eating away at the very foundation of Google Search’s massive blue link empire, which generated $234.2 billion in 2024. To combat this, Google’s testing conversational ad formats.

A person’s average attention span has shrunk to 8.25 seconds, a decline from about 12 seconds in 2000. Gen Z strongly favors instant gratification through quick access to immediate answers. To reflect this reality, Google Search now does the synthesizing for you.

Cerebras famously pioneered the “Wafer-Scale Engine (WSE).” Instead of slicing a large silicon wafer into smaller AI chips, they keep the entire silicon wafer intact to form a dinner-plate-sized AI superchip. The “WSE-3” is packed with 4 trillion transistors, delivering 125 petaFLOPS of AI compute through 900,000 AI cores. That means it performs 125 quadrillion calculations every second!

A token is a unit of text that bridges the gap between human input and machine output. It can be a word, part of a word, or even a punctuation mark. In English, 1,000 tokens is roughly equivalent to 750 words. Today’s frontier AI models process tokens by performing AI inference: when thousands of GPUs work together to understand the input and generate the output.

This setup requires rapid GPU-to-GPU communication, so GPUs don’t stop number crunching and sit idle. NVIDIA achieves this through NVLink, a high-speed proprietary interconnect that reduces bandwidth bottlenecks, preventing the dreaded “sampling slowdown”: that annoying pause you sometimes see before AI generates something. “WSE-3” avoids all of this by keeping everything on a large silicon wafer.

But there’s a catch: it only has 44 GB of SRAM, which isn’t enough short-term memory to store the model weights of today’s frontier AI models. To solve this, Cerebras links the large silicon wafer with external MemoryX, which sends the model weights when a prompt is submitted.

Because the dinner-plate-sized AI superchip can handle 2,100 to 2,500 tokens per second, it processes the submitted prompt within milliseconds. That means in less than a tenth of a second, faster than the blink of a human eye, the submitted prompt is processed. The moment it generates the output, the large silicon wafer resets, ready to tackle the next submitted prompt.

Blank Bio raised a $7.2M Seed Round to train RNA foundation models.

Ocean closed a $28M Series A to eliminate AI-powered email attacks.

Nourish landed a $100M Series C for AI-native metabolic health clinics.

Parspec (San Mateo, CA): Applied AI Engineer Intern, Summer 2026

NVIDIA (Santa Clara, CA): AI Software Engineer, Libkernel, Entry-Level

Anthropic (New York City, NYC): Design Engineer, Web, Mid-Level

Meta (Bellevue, WA): Data Engineering Director, Analytics, Senior-Level