Each week, we bring you 5 stories that resonated the most in our internal Slack channel #AI-news. We write the newsletter using various AI tools because we're an AI company and our marketing wants to move with the times too. π
Today you're reading the 61st issue in a row.
#1
OpenAI introduces new tools for agent development π€: Responses API and Agents SDK π§ π»
OpenAI has released Responses API and Agents SDK, new tools for developing AI agents π€π§βautomated systems that can browse the web, read corporate documents, and control websites using OpenAI models ππ. The new API replaces the Assistants API, which will be phased out by 2026 β³.
The tools include internet and file search capabilities, as well as the Computer-Using Agent, which simulates mouse clicks and keyboard typing π±οΈβ¨οΈ. While the technology is still in its early stages and not always reliable, OpenAI views it as a key step toward more practical and autonomous AI agents π.
The Agents SDK is open-source, making it freely available π§©, and offers easier integration, tracking, and debugging π οΈπ. OpenAI hopes that these tools will transform impressive AI demos into real products that people will use daily and seamlessly π. What do you think π§?
#2
China unveils Manus β the first fully autonomous AI for general purposes π€π
Chinese company Butterfly Effect has introduced Manus, claiming it to be the first fully autonomous AI for general purposes π€π. Unlike typical chatbots, Manus can independently tackle complex tasks such as booking travel, purchasing real estate, or developing software π«π π».
The company asserts that Manus surpasses current models from OpenAI and others π. Early users and AI experts have called it the most advanced system theyβve tested so far π. A viral demo video has generated significant interest, with invitation codes reportedly selling for thousands of dollars πΈποΈ.
However, some testers have reported factual errors and incomplete answers β οΈ. Despite this, Manus is seen as a potential step toward Artificial General Intelligence (AGI) ππ§ .
#3
Applications using existing LLM models gaining popularity among startups in Silicon Valley ππ€!
In Silicon Valley, AI startups focused on applications utilizing existing large language models (LLMs) ππ€ are gaining increasing popularity. Companies like Harvey, which tailors OpenAIβs technology for lawyers βοΈ, and Anysphere, creators of the popular coding tool Cursor π», are seeing rapid revenue growth π.
Harvey recently surpassed $50 million in annual recurring revenue, while Cursor reached $100 million π°. These AI tools are growing quickly and attracting investors who find them more practical and cost-effective than investing in developing new AI models π‘πΈ.
Legal and programming industries are already reaping the benefits of advanced AI through the right software ππ».
#4
Nvidia unveils Blackwell Ultra and Rubin: New AI chips for the next generation of AI ππ»!
At its annual GTC conference, Nvidia introduced two major innovations in AI hardware: the Blackwell Ultra and Rubin chips β‘π». Blackwell Ultra boosts AI performance, enabling cloud companies like Microsoft, Google, and Amazon to offer premium AI services ππ‘, generating up to 50 times higher revenue than its predecessor π°π.
Rubin, Nvidia's new GPU for the 2026 generation, will feature its own CPU Vera, which is twice as fast as previous versions π. With the capacity to handle up to 50 petaflops (an immense amount of computing power) π’ and the ability to combine multiple chips into one, Rubin promises a huge performance boost for AI tasks π§ β‘.
With its new annual chip release cycle, Nvidia is reaffirming its commitment to staying at the forefront of AI development π‘π.
#5
Meta AI in Europe πͺπΊ: After a yearβs delay, Meta launches its AI in Europe with restrictions π!
Meta is launching its AI chatbot in Europe πͺπΊ after a year-long delay due to privacy concerns π. Starting this week, users in 41 European countries and 21 territories will have access to Meta AI on WhatsApp, Facebook, Instagram, and Messenger π¬π±, but with text-only capabilities βοΈ.
The chatbot can assist with brainstorming, trip planning πΊοΈβοΈ, and answering questions using web data ππ. However, users wonβt be able to generate or edit images πΌοΈβ or interact with photos, and the model isnβt trained on EU user data ππ.
This move follows extensive negotiations with European regulators ποΈ. Meta plans to expand features in the future and continues collaborating with regulators to roll out more options for European users π. What do you think π§?
