Post by Mayfield

Last week, we hosted our Summer AI Pathfinders gathering, bringing together some of the most thoughtful builders, operators, and founders shaping the future of AI. A huge thank you to our speakers, Wen S. (Genspark), Philip Rathle (Neo4j), and Muddu Sudhakar (Salesforce), for sharing candid insights on technology shifts, revenue milestones, customer-centric product development, and what it takes to build enduring AI companies. A few themes stood out: Vibe coding is real, and we're all really starting to trust it. Unlike a year ago when teams were wondering whether AI can generate useful code, most of the AI Pathfinders are believers! Evals are becoming part of the software delivery lifecycle. Just as CI/CD made software changes testable and repeatable, AI evals will make AI-generated work measurable, comparable, and governable. Vibe coding 3.0 means reviewing outcomes, not code. The next stage of AI-assisted engineering will involve engineers reviewing intent, behavior, tests, evals, risk, and production impact. This only works if the surrounding trust infrastructure is strong. The context layer is becoming strategic infrastructure. As models become more capable and more commoditized (and model routers become more effective), enterprise advantage will come from proprietary context, workflow integration, permissions, memory, and trust. More context is not the same as better context. The most useful systems will curate, structure, govern, and refresh context rather than simply stuffing more information into the prompt. Thanks to everyone who joined us. We look forward to continuing the conversation at our next gathering! 📸 Branden Keller