Brian Armstrong just laid off 14% of Coinbase, restructured the org to become more flat, and is optimizing around 1-man teams with non-technical people shipping production code. The timeline is full of jokes about this but I think people are missing what's actually happening. Flat hierarchy + high agency trust is becoming the meta. AI is collapsing the time between "I have an idea" and "it's live in production." The orgs that figure this out first win. The ones clinging to 14-person approval chains for a button color change are going to get lapped. With every conversation, you unknowingly perform a lossy compression algorithm. Your idea exists in your head as pure intent. You translate it to English. The other person translates your English into their mental model. By the time it reaches implementation, you're three translations deep from the original intent. That friction is real and it compounds. More people, more translation layers, more drift from the original idea. One person with good judgment and the right tools skips all of that. Non-technical people shipping code sounds scary until you remember what every initial technology wave looks like. Early transistors were unreliable. First cars broke down constantly. Electric lighting was expensive and inconsistent. Nobody looked at those and said "well this is as good as it gets." They were the worst versions of themselves. That's where AI-assisted code is right now. The worst it will ever be. People aren't used to code having a probabilistic element to it, a yolo factor. But reliability curves only go in one direction. People drive cars but aren't mechanics. People use lights but aren't electricians. Things always trend towards abstraction long term. The real question isn't whether non-technical people should be shipping code. It's whether your org is structured to take advantage of the fact that they can.

One of the biggest feature requests we've gotten is the ability for strategies to restart. Today the engineering team is shipping Cycles. Cycles let your one-off strategies become perpetual. Define your logic once and it loops back to another defined point. That's it. This opens up a lot. The video shows a simple perpetual EMA crossover strat, but you can go much deeper. multiple cycles in one strategy, branching, converging, looping back at different points. It's all composable. In the future we'd like to add things like conditional logic and external data triggers to make cycles even more expressive. excited to see what people come up with!