Engineers who know how to use AI effectively are already 10 times more productive than those who don’t. This is something I’ve personally experienced. CoinKarma’s quantitative trading strategy, Ultron, from data integration, cleaning, backtesting, to live trading operations—almost the entire workflow was completed through AI collaboration and vibe coding. I don’t have a computer science background. Five years ago, this would have been impossible for me. Now, Ultron’s AUM exceeds $17 million, with a 600-day performance outpacing BTC by more than 5x. Revenue from strategy profit-sharing has already surpassed subscription income, and the growth rate is accelerating. I’ve noticed that many engineers still treat AI as just a chatbot. They paste code into the chat window, wait for AI to respond, and then manually paste it back into the editor. This workflow is inherently flawed. Every time you copy and paste, context is lost. Every time you switch windows, focus is interrupted. A truly AI-native development process looks like this: AI operates directly within your codebase, while you define problems, break down tasks, and review outputs. “Humans write, AI edits” and “AI writes, humans review” are two completely different production models. The throughput of the latter is at least ten times higher than the former. This shift will force engineers to move up the value chain. When implementation details are taken over by AI, engineers will need to spend more time developing other skills: system design intuition, problem decomposition abilities, and the speed and accuracy of reviewing AI outputs. These skills can’t be accumulated through years of experience—they require deliberate practice. And these skills compound. People who collaborate with AI daily will get better at crafting prompts that make AI get it right the first time, breaking down tasks to maximize parallel processing, and reviewing outputs to quickly catch AI’s blind spots. This experience is transferable across projects, languages, and even roles. I believe good tools change the way people work, but change needs a push. So I’ve made a decision: effective immediately, all CoinKarma developers are required to use Claude for collaborative coding. The company will fully cover the cost of Claude Max subscriptions. Change starts with us. Let’s all embrace the era of big AI together.