Greylock Partners

Braintrust's mission is to help companies build quality AI products - they're not only shaping AI themselves, they're enabling everyone else to do so too. Congrats to Braintrust for being listed on the TechCrunch Disrupt top 60 companies shaping AI! My favorite thing about listening to Ankur Goyal talk about evals and building great AI products is he's always pushing for simplicity even as resulting systems are getting more and more capable: “Agents are a very natural evolution of AI software, and AI software is a very natural evolution of software. On average, the resulting agentic system that people build is dramatically simpler than the previous generation." cc Ankur Goyal

Today we announced an equity round of up to $750 million at a $10 billion post-money valuation. The round was co-led by Dragoneer and GIC, and includes a group of world-renowned new and returning investors. We chose this group of investors because of their proven track record scaling category leaders, and because they are committed to the long-term future of Wealthsimple. They believe in our vision for the future of financial services. This funding will supercharge our growth by helping us scale our products and deliver more value to our clients. More from our CEO and co-founder Michael Katchen below. https://lnkd.in/gsREmiDg

This last year has marked a fundamental shift from Generative AI to Agentic AI: agents that maintain goals and execute multi-step workflows independently. We started Greylock Partners Change Agents to explore this new era of agentic AI with the founders, researchers, and builders who have a front-row seat. Coming up we have the founders of Harvey, Glean, Cresta, Box, Baseten, and more. Change Agents is recorded in front of a live audience in SF - if you’re interested in working on agentic AI, DM me for invites. What we're learning: Reliability of agents - Spiros Xanthos @ Resolve AI: "Agents in production today are more like self-driving cars, where we might have proof that it's a lot safer than a human driver, but still this is not good enough to let it go." Democratizing capabilities - Varun Mohan @ Windsurf: "Before we launched Windsurf, we gave the product out to everyone at our company, including people that were not technical. Some of our biggest power users now at the company are people that don't write software inside our company." Human-in-the-loop - Jesse Zhang @ Decagon: "What we try to do is do all of the work and maybe if there is a final step that is irreversible, then a human takes that step. If some of these actions mutate the system but still are reversible, maybe an agent can take them." Agent-to-agent coordination - Karthik Narasimhan @ Sierra: "If you want agents to act in the world and work together with humans, they would also have to work together with each other." On the evolution of AI evals - Ankur Goyal @ Braintrust: "We went from extremely slow updates that were very manual to very fast updates that are very manual. And I think we're going to move towards even faster updates that are partially or even entirely automatic. The way that we as humans are actually going to interact with evals is going to shift." Security implications of AI-generated code - Vineet Edupuganti @ Cogent Security: "With the rise of AI generated code and vibe coding, that has a lot of positive developments in terms of productivity. But the reality is, a lot of code that AI models spit out is not fundamentally secure." Superhuman speed - Evan Reiser @ Abnormal AI: "In the past it might take hours or days for a criminal to figure out ways to attack. Now you might have like a second. So the time it takes for that alert to pop up to the SOC, to have the world's best analyst go take a look at that, it might be too late. You need AI that's able to detect things with superhuman capabilities and then respond at superhuman speed." Engineering productivity - Mayank Agarwal and Rushin Shah @ Resolve AI: "We have seen undeniable gains in code creation, but really to achieve high productivity gains in software engineering, we need to automate the more complex activities that require deep understanding of how to go across different tools, operational boundaries and tribal knowledge where a human doesn't have to be at the wheel."