Never doubt the AI exponential, because it can catch up with your entire career while you’re still explaining why your profession is special. Just listened to Cédric Villani, a Fields Medal winner, talking about how shocked he was by the progress of AI. Funny because I met him twice, including around 2018 when he was a Member of Parliament leading France’s AI mission, and we were already discussing how fast NLP was progressing with the new transformers. France has 13 Fields Medals, second only to the US, and some of the brightest people I know come from that world. A lot of them have been caught by surprise despite following AI for years and using the products themselves. I think it’s very hard to accept the exponential when it threatens something you’ve spent your entire life building, especially when your career is such a big part of your identity. You can see what’s happening to graphic designers or software engineers and still convince yourself that your own work will take another 20 years. My guess is every intellectual profession will go through this, and honestly, even believing in the exponential doesn’t prepare you for the consequences. I work on this stuff all day, I expect progress to keep compounding, and I still struggle to imagine what that means a few years from now. I don’t think nature has equipped us to really grasp exponential growth. We can understand the math and still struggle to imagine what it means for our own lives, even while watching it happen.

Today we're launching Sunra MCP, with the most accessible pricing in the industry. Every leading image, video and audio model, right inside Claude, ChatGPT and the AI assistant you already use. Each one at its maker's own API price, with no markup added. Now is the time to build. Go create something no one has seen before.

Today, we are launching OpenRouter Armor in beta. Tool calls allow agents to interact with the world, and regex-based filtering often isn't enough to detect unwanted tool calls. Armor uses @typesafeai's Jev model to guard against out of policy tool calls before your agent executes them, with better performance than Claude Code's auto classifier at a fraction of the cost.