AI Pulse 🚨
A bite-sized curation of this week's most important AI news.
🌐 Meta launches standalone AI app to compete with ChatGPT. Initially available in the US and Canada, the app can remember user preferences and context across conversations.
🔍 OpenAI rolls out major ChatGPT shopping and search improvements. New features include improved product search with visual details and direct buy links, WhatsApp integration for real-time answers, improved citations with source highlighting, and trending search suggestions.
🤖 Alibaba’s Qwen3 introduces a hybrid architecture combining fast responses with deep reasoning. The largest model in the series (235B parameters) shows strong performance against leading models, with support for 119 languages and adjustable “thinking modes” for different tasks.
🔎 Google adds new AI Mode features in Search, introducing visual place and product cards, a history panel, and shopping integration for select U.S. users. This represents Google’s direct response to AI-native search platforms like Perplexity.
🚨 Security Researchers discover major prompt injection vulnerability affecting leading LLMs. The “Policy Puppetry Attack” lets attackers bypass safeguards to extract system prompts and generate harmful content across ChatGPT, Claude, and Gemini, exposing alignment weaknesses.
📊 Microsoft releases 2025 Work Trend Index, revealing emergence of “Frontier Firms” built around AI. Key findings show 82% of leaders consider this year pivotal for strategy shifts, with 71% of workers at AI-native firms reporting company success versus 37% globally.
🤖 OpenAI rolls back GPT-4o update after discovering the model became overly agreeable in responses “sycophantic behaviour”.
🔗 Anthropic launches “Claude Integrations”, enabling connections with popular services like Jira, Confluence, Zapier, and more. This allows Claude to interact with apps and services across the web, significantly expanding its capabilities.
💰 xAI Holdings reportedly seeking $20 billion in funding, potentially valuing the company at over $120 billion. If successful, this would be the second-largest startup funding round ever, behind only OpenAI’s recent $40 billion raise.
🤔 Dario Amodei discusses the critical importance of developing “AI interpretability” before AI systems become too powerful, arguing that understanding the inner workings of AI models is essential for addressing safety risks and ensuring responsible deployment.
🌥️ Amazon launches Nova Premier, their most advanced model for complex tasks. Available in Amazon Bedrock, it features a 1M token context window, multimodal capabilities, and serves as a teacher model for distilling smaller, specialized versions.
👨💻 Meta CEO Mark Zuckerberg predicts AI will write half of Meta’s code by 2026. Speaking at LlamaCon with Microsoft CEO Satya Nadella, he envisions every engineer becoming a “tech lead” managing their own “army of agents.”
🔮 Perplexity Comet browser set to launch next month. The browser aims to enable true AI agents by accessing third-party services through hidden tabs and performing client-side actions. CEO Aravind Srinivas sees browsers as “containerized operating systems” essential for building the next generation of AI assistants.
🎧 Google’s NotebookLM Audio Overviews now works in more than 50 languages. You can pick your preferred language in the settings to help you create content and study in different languages more easily..
💳 Mastercard announces “Agent Pay”, a new payment solution for AI agents. The platform includes agent verification, enhanced tokenization, and fraud protection, with Microsoft, IBM, Braintree, and Checkout.com as early partners.
💳 Visa unveils “Intelligent Commerce”, an initiative enabling AI agents to make secure purchases on behalf of users. The platform introduces AI-Ready Cards with tokenized credentials, personalized shopping recommendations, and customizable spending limits. Partners include Anthropic, Microsoft, OpenAI, and other industry leaders.
📱 Perplexity launches WhatsApp integration, enabling users to access answers, sources, and image generation directly through the messaging platform. More features are promised to come soon.
📈 Anthropic Economic Index: Latest research reveals AI’s significant impact on software development, with 79% of Claude Code interactions focused on automation tasks. The study shows startups are leading AI adoption (33% of interactions) while enterprises lag behind (13%).
