Shaw: DeepSeek has become the leading AI model; the team is building the next version of Eliza
Odaily News ai16z co-founder Shaw published an article on X saying that more powerful models are always a good thing for AI agents.
For years, AI labs have been outdoing each other in benchmarks and capabilities. Sometimes Google was ahead, sometimes OpenAI was ahead, sometimes Claude was ahead. Today it’s DeepSeek. The trend is that the world’s largest and most well-capitalized companies are competing for a technology that is ultimately free, open source, and costs nothing to run on your computer at home.
The winners of this race have always been hardware and consumer products. Nvidia always wins. Every model is optimized for their hardware. Apple also always wins: they invest in a unified memory architecture that enables high-VRAM machines to run the latest models (albeit slowly).
Products continue to benefit from the latest models. Cursor and Perplexity are examples of products that magically get better every few months, but as AI becomes integrated into almost all products, all of them benefit from cheaper, faster AI models.
AI agents are a new paradigm for applications - the core argument is that applications need to migrate to social media where users are, and agents are a form of application that can exist entirely on social media without users having to leave. They are self-advertising and benefit from the network effects of every user interaction.
When a new model comes out, integration into the agent framework usually only takes a few lines of code. Most model providers follow the same API conventions, following OpenAI, so this can usually be done in minutes. This gives any agent application immediate access to the latest models. Every time a state-of-the-art model comes out, the agent gets smarter.
This week has been a huge win for all of us. For the agents, for the humans, and for the AI model teams who are now motivated to work harder and do better. I’m not worried at all about our place in all of this. We’re building the next version of Eliza, and it’s only going to get better. Thousands of teams are building on our technology, and over 500 people have contributed to the core repository, a number that will continue to grow as we continue to grow. We’re creating a template for how ambitious founders can crowdfund their public product projects, and the team will be rolling out more content to solidify this strategy in the coming weeks and months.
