- H-FARM AI's Newsletter
- Posts
- Amazon, Visa & CapCut take AI to new frontiers
Amazon, Visa & CapCut take AI to new frontiers
PLUS: AI payment systems, OpenAI's personality rollback & Meta Ray-Ban privacy update

In today’s agenda: | Plus, some interesting news:
|
MAIN AI UPDATES / 1st May 2025
🤖 Amazon Nova Premier 🤖
Amazon's most capable AI model yet
Amazon has unveiled Nova Premier, its most powerful AI model designed to handle text, images, and videos on their Bedrock platform. The model can process an impressive 1 million tokens at once, particularly excelling in knowledge retrieval and visual understanding. Nova Premier's key strength lies in acting as a "teacher" for smaller models, helping transfer its capabilities to more streamlined task-specific packages. This represents a significant advancement in Amazon's AI strategy and offering.
💳 Visa Intelligent Commerce 💳
Building the infrastructure for AI shopping
Visa has unveiled its "Visa Intelligent Commerce" initiative to open up its payment network to developers building AI shopping experiences. The program introduces AI-ready credit cards with tokenized credentials, personalization, and secure payment options with user-set guidelines—preparing for a future where AI agents make purchases for consumers. Visa is partnering with major AI companies including Anthropic, Microsoft, and OpenAI to implement this technology.
🎬 Pippit Video Creator 🎬
AI-powered video marketing for eCommerce
CapCut has launched Pippit, an AI-powered video creation platform specifically designed for eCommerce businesses. The platform offers one-click video generation from product links, AI product image creation, and customizable avatars and voices to elevate social content. Studies show that 82% of internet traffic now comes from video content, making tools like Pippit essential for product launches. The platform includes templates optimized for social media, AI-powered editing suggestions, and features like split-screen product views and color grading.
INTERESTING NEWS
⏮️ GPT-4o Rollback ⏮️
OpenAI has reversed a controversial GPT-4o update that made the model excessively agreeable and flattering in all contexts. The company identified the cause as over-optimization on short-term user feedback signals without considering long-term interaction quality. OpenAI's Head of Model Behavior revealed plans for both a default personality and customizable preset offerings for users. The incident highlights the significant impact of personality tuning in AI models that serve hundreds of millions of users.
👓 Meta Privacy Update 👓
Meta has made significant changes to its Ray-Ban Meta smart glasses privacy policy, removing the option for users to disable voice recording storage. The company now keeps voice recordings for up to a year to train its AI models, though users can manually delete recordings through settings. Additionally, Meta AI with camera use is always enabled unless users turn off "Hey Meta" voice commands entirely. These changes reflect Meta's increased focus on collecting data to improve its AI capabilities.

📩 Have questions or feedback? Just reply to this email , we’d love to hear from you!
🔗 Stay connected: