Recommended article this week: All about AI｜Why GPT-5.6 is changing the way we collaborate with AI, and what we can do with it. **《How GPT-5.6 Changes Knowledge Work: Don’t do your work. Tend your loop.》** Author: Dan Shipper｜Every Link to the article is at the end! This article clearly explains how the upgrade of GPT points to a new direction for the future of AI: the basic unit of using AI is shifting from completing single tasks to running a continuous feedback loop. Previously, when we used tools like GPT to assist with work, the process was something like: gather information, make decisions, execute, then review. AI would only handle one small step in this process. For example, when I needed to research a project, AI might help me gather information or assist with decision-making. But starting with GPT-5.6, the new way of collaborating with AI looks more like this: Gather and understand information → Propose decisions and actions → Human approval → Execute → Learn from feedback. AI can continuously scan information sources, summarize key points, suggest next steps, and even execute approved actions. Meanwhile, humans shift from “doing every step themselves” to setting goals, defining boundaries, correcting the system, and making critical judgments. The author calls this way of working **tend your loop**: don’t just complete tasks—nurture the system that continuously generates work. Of course, this model isn’t suitable for every task and is still in its early stages. It might take some time for everyone to explore and find what works best for them. But once you do, you might discover a lot of fun and exciting possibilities. Here are three scenarios I can think of right now that I’d immediately try: 1️⃣ **Investment loop**: Market info → Key changes → Scenario analysis → Portfolio actions → Results review. 2️⃣ **Content loop**: Ideas and fragments → Topic selection → Drafting → Publishing and data analysis → Summarizing what fits your style. 3️⃣ **Community loop**: Group chat insights → High-quality signals → Weekly draft → Member feedback → Offline events → Deeper connections. So, what is AI compounding? I think it’s about teaching AI to gradually understand your decision-making criteria, so the same process gets better and better each time. In the future, the gap between people may not be about who completes more tasks, but who builds more systems that can continuously learn and improve. Article link: https://every.to/chain-of-thought/how-gpt-5-6-changes-knowledge-work