Reads

"The cost of abstractions with AI is very high. Every layer between your code and the model is a tax on context, speed, and clarity. That's why tools like Convex, which keep everything in code, win over abstraction heavy alternatives like Supabase when agents can grep and edit your logic directly, complexity becomes the enemy."

"To solve the hardest problems, we must first solve intelligence itself. The path to AGI isn't just about winning games; it's about creating the ultimate tool to unlock the secrets of the universe."

"Don't flood the context with every tool at once. Present them as code to explore, and let agents discover what they need when they need it—efficiency follows discovery."

"In a distributed system, the source of truth is the log. If you can get all nodes to agree on the state of that log, you've achieved consensus and solved for data consistency."

"To solve concurrency conflicts, stop having transactions operate on the same data. Instead, give each transaction its own 'version' to work with."

"True product delight goes beyond superficial features. It's achieved by creating experiences that thoughtfully address both a user's functional goals and their underlying emotional needs, turning utility into a powerful connection."

"The goal isn't just to build a great product, but to solve a problem so painful that the market pulls the product out of you. That's when you have the wind at your back"

"Care more about outcomes and impact than the exact implementation, or the tools used to solve the problem"