When answers become worthless, judgment is your only compound interest asset

*This article originally appeared in Vista Newsletter No.133 (opens in a new tab). *

Last week a reader wrote to me asking a question, and it took me two days to figure out how to answer it.

This friend, Helen, told me that she had been really serious about it this past year. Not only have I subscribed to three AI tools and completed two online courses, I also spend an hour or two every week following the release of new AI tools. I have accumulated more than 400 reminder notes stored in Notion. However, her promotion speed was not faster, and her quotation for accepting cases could not be increased. Even the book she wanted to write was still in the outline development stage.

She finally asked me a little frustrated: “Did I not learn enough?”

I want her not to think too much, because it’s not that she hasn’t learned enough, but that she’s putting her energy into the half that will lose value!

Therefore, in this issue of the e-newsletter, I would like to integrate several articles I have written in the past six months to talk about something that I am increasingly certain of: when answers become cheap, judgment is the only asset that will compound interest. Judgment is not a talent. It is supported by three pillars, and all three can be practiced.

Learning quickly is no longer an advantage.

Let me tell you a cruel fact first.

The World Economic Forum’s Future of Jobs 2025 Report (opens in a new tab) estimates that an average of 39% of workers’ existing skills will change or become obsolete between 2025 and 2030. A study by the International Labor Organization in the same year pointed out that about a quarter of the world’s workers have some degree of exposure to generative AI in their occupations, but only 3.3% fell into the highest exposure category.

Reading these two numbers together, the conclusion is not that AI will replace you (opens in a new tab), but something more troublesome: your work content will be constantly reshuffled.

The real consequence of this is that the return on investment for learning quickly is declining.

This is not to say that learning is not important, but to say that when the half-life of a skill is only two years, if you invest your time in the skill itself, you are buying an asset that continues to depreciate. Learn prompt words today, study agents tomorrow, worry about the newly built process being eliminated by the new version the day after tomorrow. This kind of anxiety will make people keep collecting information, but they will be unable to submit works.

I wrote in “Stop Chasing Tools!” I wrote a sentence in “Build Your Undefeated System in the AI ​​Era”, and now I am more certain that it is right: tools accelerate mediocrity, and systems create value. Tools will change, and the skeleton of ability is what you can take with you.

So, what doesn’t lose value?

The answer is: the ability to decide “what to learn, what to do, and what to believe.” That is, judgment.

The problem is that judgment sounds like a metaphysics. Even if you spend money to take a “judgment course”, you can’t accumulate it using the method taught by the teacher. So most people know it’s important, but don’t know how to practice it, and end up going back to learning the next tool.

I wrote several seemingly unrelated articles in the past six months, but when I looked back, I found that they pointed to the same thing. Judgment is actually a synthesis of three specific abilities, each of which can be disassembled, practiced, and accepted.

The first pillar: attention is the raw material of judgment

Judgment cannot grow in noise.

Work data released by Microsoft in the United States in 2025 shows that high-message users will be interrupted by a meeting, email, or chat message every two minutes on average during core working hours, up to 275 times a day.

When I first saw this number, my first reaction was not “Modern people are so busy”, but “No wonder everyone is making shallow decisions.”

A brain cut into two-minute chunks can handle only reflexive responses: replying to messages, acknowledging receipt, dropping things off. Things that really require judgment, such as whether to accept this case, whether to turn in this direction, and what this customer’s silence means, all require a period of continuous thinking without interruption. And during that time, it was eaten.

I talked about this in “Concentration is your most precious asset”. At that time, my focus was on “Concentration is a scarce resource.” Now I want to go one step further: focus is not just scarce, it is the raw material of judgment. You can’t make any trustworthy judgments in a head that’s interrupted every two minutes.

Therefore, protecting attention is not for the productivity-sounding reason of “efficiency.” It’s because that’s the only environment in which you can produce high-quality decisions.

How to do it specifically? I only try to stick to the following three principles:

• Set aside a period of time every day that is not cut off by communication software, and turn off all notifications, not mute them but turn them off. During this period, you only do things that require judgment and don’t do things that can be explained to you.
• Separate responding from thinking. Respond to messages and handle chores within a fixed period of time, and do not enter the inbox during the rest of the time.
• Set a fixed review frequency for changes in the market and workplace, such as once a week. Don’t let every piece of news, every model release take over your attention on the spot.

Article 3 is particularly important. The density of AI news this year is unreasonably high. If you follow every news, you will have the illusion that you are very busy and can keep up, but that illusion comes in exchange for judgment.

The second pillar: micro-decision-making is where judgment really happens.

The second article is “Humans are worse than AI (opens in a new tab)” by Wanweigang.

I wrote about this concept in “Your micro-decisions are the key that AI cannot replace”. He said that the irreplaceability of human beings does not lie in being smarter than AI, but in those small seemingly unimportant decisions scattered in daily life.

This sentence seemed abstract to me at first, until I started to seriously observe my work.

Well, let me take writing an article as an example. AI can generate structures, supplement information, and reorder sentences. But throughout the entire process, what really determines the quality of the finished product is a series of small choices with no standard answers: Should we use stories or data for this opening? Is this argument too fast? Does the reader understand this word? Will saying this sentence hurt someone I didn’t expect?

These are not big decisions, and none of them warrant a meeting. But they add up to the entire gap between professionals and amateurs.

AI cannot provide these, not because it is stupid, but because it does not know who your readers are, what pitfalls you stepped into last time, and what kind of person you want to be. The quality of these micro-decisions comes from your experience and values, and that is what you have accumulated over ten or twenty years.

Therefore, the way I view the division of labor between humans and machines is completely different from a year ago. Instead of asking in general terms “Will AI replace me?”, it is better to break down the daily work and divide it into the following three categories:

• Which tasks are suitable to be handed over to AI as a whole? For example, check information, organize verbatim drafts, and produce the skeleton of the first draft.
• Which results require human review? Such as citations, fact checks, anything that will have your name on it.
• Which judgments involve trust, communication and responsibility and must be borne by oneself? Things like pricing, saying no, apologizing, and deciding what not to do.

The 5C framework proposed by LinkedIn CEO is actually saying the same thing. I dismantled the five Cs in “Five Things AI Can’t Do”: curiosity, courage, creativity, empathy, and communication. You will find that none of them are skills, they are all judgments.

The most counter-intuitive thing about this is:

As the AI gets stronger, micro-decisions become more important.

Because when execution becomes almost free, deciding “what to execute” becomes the only variable.

I myself had a very painful validation. At the beginning of this year, AI made development so easy that I could make a new tool every day, and I was excited for a while. It wasn’t until I stopped to take stock that I realized that very few of them were actually used and paid for. The lesson I learned from that inventory is more valuable than any other lesson: I don’t lack the ability to execute, but what I lack is the ability to judge whether it’s worth doing.

The third pillar: Rehearsal is a training ground for judgment.

The first two pillars talk about “protecting raw materials” and “recognizing the battlefield”. But the real difficulty with judgment is that it is most likely to be used when the pressure is greatest and time is least.

This is what the third pillar addresses.

This week I read Maya Rachel’s “Do the right brain exercises and your performance will improve (opens in a new tab)” and wrote “The more uncertain the world is, the more you need to run it through your mind first”. The mental rehearsal discussed in the book is not about closing your eyes and imagining success, but simulating the goals, steps, emotions, and risks in your mind to reduce confusion when you actually take action.

The book proposes five visualization techniques. I think these three are the most practical for workers.

The first is process visualization. Instead of repeatedly imagining the applause after the briefing, it is better to go through the opening, transition, being questioned, and how to pick up after forgetting the words. The former brings temporary excitement, while the latter reduces on-the-spot wear and tear. The value of mental rehearsal is not to make things happen according to the script, but to save the brain from having to process the scene for the first time when the pressure is highest.

The second is result visualization, but it is only useful if it is paired with process visualization. First see the abilities you want to have, such as completing a research report independently; then work backwards through the process, rehearsing how to define the problem, check sources, use AI to organize materials, and make the final judgment yourself. In this way, learning no longer follows a tool list, but serves a clear competency goal.

The third type is negative visualization, which is what I value most. The World Economic Forum’s Global Risks Report 2026 (opens in a new tab) compiled the views of more than 1,300 experts and ranked geoeconomic confrontation as the top risk that is most likely to trigger a global crisis in 2026. Another 50% of respondents expect that the global situation in the next two years will continue to be turbulent or even highly unstable.

These numbers are not predictions but the collective judgment of experts on risk. They remind us that policies, supply chains, markets and technological rules can all suddenly shift. For individuals and businesses, the sense of security of staying on the same path is diminishing.

Negative visualization is not about replaying the disaster in your mind, but about rehearsing a reasonable worst-case scenario: What if an important customer suddenly leaves? What should I do if the platform changes the rules? Half of the originally familiar tasks have been taken over by AI. Where should we move our capabilities?

And after the rehearsal, be sure to arrange the first action for each risk. Build your own roster, maintain a cash buffer, diversify your revenue streams, or organize your experience into transferable knowledge assets. When a record is in place, your brain doesn’t have to remind you every day with vague fears that something might go wrong.

When things are unstable, resilience is not about believing that everything will be fine, but about your ability to update your judgment and take the next step when things go awry.

How to use the three pillars together?

At this point you may feel that these three things make sense individually, but together they seem a bit forced. Therefore, I want to give you a specific way to use it together.

Think of AI as a rehearsed opponent, not a source of answers.

I talked about the structure of in-depth dialogue in “Let AI become a thinking partner, not just a machine that produces answers”. The core is the density of questions: the amount of information you give determines how deeply it can accompany you in thinking. At that time, my focus was on how to ask for insights. Now I want to add a more practical usage.

Rehearse the new workflow in your head, then throw this process to the AI and ask it to play three roles:

• Play the role of the most critical reviewer or client, write a rejection letter for your proposal, and list the most likely points first.
• Play yourself six months from now and look back and point out what you missed in this decision.
• Play the role of a person who knows nothing about your business and ask you three of the most basic and difficult questions to answer.

These three roles do not require AI to give answers, but rather require it to help you run a more complete rehearsal. Your judgment is still your judgment, but it’s stress-tested.

I did a very interesting experiment, which I wrote in “I can’t compose music, but I made a piano song that I can play”. That experience made one thing clear to me: my role changed from one of executor to one of judgement. I don’t know how to compose music, but I know which version sounds right, which version sounds a little off, and where. I don’t have the technology, but I do have the judgment, and the quality of the final product depends on the latter.

What is truly scarce is never skills, but judgment.

A ninety-minute weekly exercise

At this point, I want to give you something you can actually implement, rather than just another set of concepts.

It only takes five minutes a day:

First use thirty seconds to see the results you want to accomplish today, and then spend two minutes rehearsing a few key steps. Then take a minute to imagine the most likely disruption and plan a response to it. Finally ask yourself: What is the first action you can complete right now? Then start immediately.

Sixty minutes every week:

• In the first twenty minutes, pick a task that you have actually done this week and break it down into three categories: “AI can do it, people need to review, and people need to judge.” Write it down, don’t just think it in your head.
• In the middle twenty minutes, do a complete preview of the most important thing next week: walk through the process, think about the worst-case scenario, and write down the first action for the worst-case scenario.
• For the last twenty minutes, throw this preview to the AI, let it play the three roles above, and see what you missed.

Do this once a quarter:

Take stock of what you have accomplished this season, not what you learned, but what you handed over. Does anyone use it? Did anyone pay? If the answer is no, the problem is probably not in execution, but in the initial judgment.

This exercise only takes up ninety minutes per week, but what it teaches cannot be eliminated by the next model.

##Finally

Back to the question from reader Helen at the beginning.

I later replied to her this way: Your four hundred reminder notes will not be in vain, but they are raw materials, not assets. The real asset is your ability to decide what’s worth doing, and you don’t think of it as something that needs practice yet.

When everyone can get answers quickly, whether they can manage attention, recognize key micro-decisions, and rehearse things before they happen can actually widen the gap.

In addition to continuously upgrading tools, we must also upgrade ourselves in using tools, making choices, and continuing to move forward.

If you are a researcher, graduate student or knowledge worker and want to truly practice this set of judgment skills in your daily work process, welcome to my workshop. I will help you hand over the tedious work of document crawling and data compilation to AI, while maintaining the boundaries of academic ethics and keeping the most critical questions, arguments, and judgments firmly in your own hands.

🚀 Course information is here: AI Empowering Academic Research and Writing Practical Workshop (opens in a new tab)

Well, this e-newsletter comes to an end here. If the content of this issue is helpful to you, please feel free to forward it to a friend who is also working hard to learn tools but feels that he has not made progress.

See you next time.

---

If you like this article, please subscribe to the Vista newsletter (opens in a new tab) and have the latest issue sent directly to your inbox every week.