It's Saturday and I still haven't gotten around to writing a news drop, someone needs to remind me about these things...

DeepSeek:

This week, DeepSeek, an AI startup out of China, has released R1. R1 is an open source chain-of-thought (thinking) model that rivals even OpenAI's cutting edge o1 model. And the best part is that it's open source, and therefore free for anyone to use!

This truly is an incredible piece of technology in a space we are just beginning to understand, and shows China's commitment and ability to beat western companies in this race. I think the amount of startups coming out and beating the established players in the past few months shows that the rest of the year will likely be very exciting, and we've likely not yet met all the players.

OpenAI:

OpenAI has released Operator! Operator is a new tool that allows you to give it instructions, and it will then interact with your browser to complete those actions to the best of its ability. With neat integrations to several platforms, this seems to be OpenAI's first attempt at "agentic models."

For example, Anthropic has Claude Computer Use, which allows their Claude model to interact with a computer via keyboard and mouse. Google is also a leader in this space, with Gemini Deep Research able to browse thousands of documents to help with your research and Jules being able to review your codebase and autonomously spot, test, and fix bugs.

The idea here is to replace as many human jobs as possible with autonomous AI agents, which is where the "agentic" name comes from. As these companies seek to move beyond just chatbots and towards actual products, more and more of them are focusing on replacing humans in interacting with our world and internet.