Hook

Their other posts in the index, biggest breakout first.
The old AI skill was prompt engineering. The new AI skill is loop engineering. The shape of stuff today is roughly "Humans have ideas, and the models are able to em-placement, test and evaluate them an [order of magnitude] faster than before." A Claude Code lead at Anthropic basically said he doesn't spend his time writing perfect prompts anymore. He writes loops. It means instead of asking Claude one thing, waiting, correcting it, asking again, and babysitting the entire process, you build a system that keeps Claude working. You give it one goal, and the loop tells Claude what to do next, checks the output, remembers what's already finished, and runs again until the task is done. Inside Claude Code, there's five things. No. 1 automation, so it runs without you. No. 2 worktrees, so multiple AI agents can work at once. No. 3 skills, so Claude remembers your project context. No. 4 connectors, so it can use your actual tools. And No. 5 subagents, so one agent builds while the other reviews. That's the shift. You're not just chatting with AI anymore. You're building AI employees that work in loops.