Google Gets In The Mode; Branched Convos; Ca(i)ll Centers

Google AI Mode May Finally Be Killing the Blue Links

Google is fundamentally transforming its search experience, with product lead Logan Kilpatrick signaling that google.com's classic search page will be replaced by AI Mode.

The AI-first experience is live in 180+ countries with 100M monthly users and a 90% zero-click rate, representing a structural shift from directing traffic outward to answering questions in place.

Additionally, Google released EmbeddingGemma, a compact 300M parameter multilingual embedding model optimized for on-device performance, allowing users to search emails, notes, and web pages using plain English questions without data leaving their device.

Read more: AI Secret, The Rundown AI, TLDR AI, The Neuron

OpenAI Tackles Hallucinations, Inbreeding

OpenAI published groundbreaking research suggesting that AI hallucinations can be solved by teaching models that it's okay to say "I don't know" rather than making confident guesses.

The research found that standard training methods reward confident guessing over admitting uncertainty, creating a fundamental conflict where models learn to always guess even when completely uncertain.

This comes as the industry grapples with AI's "inbreeding" problem, where models trained on AI-generated content experience bias multiplication and quality degradation, potentially leading to an industry-wide decay spiral.

Read more: The Rundown AI, AI Secret

Custom Chips for OpenAI

OpenAI will begin mass production of its own custom AI chips next year through a $10B partnership with Broadcom, joining Google, Amazon, and Meta in reducing dependence on Nvidia's hardware.

The custom chips will help OpenAI double its compute capacity within five months to meet surging demand from GPT-5 and address ongoing GPU shortages. The chips will be used internally only and won't be available to external customers.

Read more: The Rundown AI, TLDR AI

🔧 UX Innovations

At Last... OpenAI Launches Chat Branching

OpenAI introduced chat branching functionality in ChatGPT, allowing users to branch off conversations to explore tangential questions without derailing the main chat thread.

This prevents context rot in long conversations and enables users to ask follow-up questions like "give me three examples of collaborative filtering" while maintaining the original conversation flow about recommendation algorithms.

Read more: The Neuron

Claude Pro Enhances Conversation Continuity

Claude Pro users can now have Claude reference past chats, making it easier to pick up where conversations left off. Users can toggle this setting on or off, improving continuity and context retention across Claude conversations.

Read more: The Neuron

🏢 Acquisitions & Funding

Atlassian Acquires AI Browser Company

Atlassian acquired The Browser Company for $610M cash, focusing on creating an AI-powered browser optimized for knowledge workers. The acquisition builds off Atlassian's existing suite of productivity SaaS applications, while The Browser Company will continue developing its unreleased Dia browser under independent operations.

Read more: TLDR AI, The Neuron

Anthropic Pays Up for Stealing Book Content

Anthropic agreed to pay at least $1.5B to settle a class-action lawsuit from authors for using copyrighted works to train Claude, marking the first major payout from an AI company for copyright violations. The settlement covers approximately 500,000 books at $3,000 per work after authors discovered Anthropic downloaded over 7M pirated books from shadow libraries. This comes alongside Anthropic's recent $13B funding round at a $183B valuation, highlighting the company's "run-rate revenue" of $5B rather than traditional metrics.

Read more: The Rundown AI, AI Secret

Sierra Raises Massive AI Agent Funding

Sierra announced raising $350M at a $10B valuation to expand its AI agent platform, backed by Greenoaks. The platform supports hundreds of major companies, reaching 90% of Americans in retail and 50% of US families in healthcare, demonstrating significant investor confidence in AI agent platforms for enterprise applications.

Read more: TLDR AI

🇨🇳 International Moves

China's DeepSeek Challenges OpenAI

Chinese startup DeepSeek is developing a multi-step AI agent software similar to OpenAI's Agent tool, designed to handle complex tasks without human supervision. The system is expected to launch by end-2025, representing direct competition to OpenAI's agent capabilities from a major Chinese AI company.

Read more: TLDR AI

ByteDance Advances Robotic AI

ByteDance introduced Robix, a unified vision-language model for robots that combines reasoning, task planning, and natural language interaction in one framework. The system enables robots to handle complex, long-horizon tasks and communicate naturally with humans through verbal responses and proactive dialogue.

Read more: TLDR AI

💼 AI's & Employment

Call Center Transformation

AI has fundamentally restructured call centers, with routine "break/fix" work shifting to bots while complex cases still require human expertise.

Klarna, for example, famously replaced 700 service reps with chatbots in 2023 but had to rehire humans when customer satisfaction tanked for identity theft cases.

The trend shows AI eliminating $12/hour script reader jobs while making $40/hour problem-solvers indispensable.

Read more: AI Secret

🛠️ Notable AI Tools & Launches

Recent AI tool launches include...

• Spinach AI for meeting transcription and CRM updates
• Moonshot AI's Kimi K2-0905 with enhanced coding capabilities and 256k token context
• Google's NotebookLM for transforming documents into customizable podcasts with four formats - each adjustable - helping you digest complex information through audio
• Midjourney's new style explorer for visual prompt testing, and various specialized tools for outfit recommendations (Ella), interface prototyping (Prototyper), and AI agent simulations (AIvilization).

Read more: The Neuron, TLDR AI

Today’s Sources:

The Inernet
The Neuron
The Rundown AI
TLDR AI
AI Secret