Stop Eating Other People’s Secondhand Leftovers

I recently understood something: much of my time in the past was spent eating other people’s leftovers.

Reading other people’s summaries, listening to their interpretations, asking everywhere “how do I do this?”, buying courses and reports, and joining communities all share one quality: someone else has digested them once and spit them out for me. No amount of leftovers can grow your own judgment.

Here are the four rules I set for myself.

1. Get as close as possible to first-hand information

You can learn about the same thing from three places:

1. What the person involved says: their posts, blog, code, financial reports, and words from their own interviews.
2. What someone who understands the field relays: their interpretation plus their biases.
3. What someone who does not understand the field repeats: clickbait, community screenshots, and “a big shot revealed”.

At each layer, information is lost once and noise is added once. Most people encounter the third layer every day, the second occasionally, and almost never the first.

The ceiling of your judgment is the quality of your sources. If your source is third-hand, even intelligence only lets you calculate precisely inside noise.

So the first rule is: for anything important, find the first-hand source first. To follow a model’s progress, read the vendor’s release page and papers, not commentary; to see how a founder works, read their own posts and product, not someone else’s “ten secrets”; to study a company, read the original filings and earnings-call records, not a research-summary excerpt.

This is boring, but it is the starting point of the entire chain of cognition. If the starting point is dirty, everything afterward is dirty.

2. Do not eat other people’s leftovers

Asking everyone is eating leftovers.

What you get from another person is their conclusion, not your understanding. Their conclusion comes from their situation, information, and interests and may not apply to you. Worse, asking others can itself degrade your learning ability: once you get used to waiting for answers, you stop getting used to finding them.

Paying for secondhand information is also eating leftovers. Courses, reports, paid communities, and “inside information” are essentially things someone else has chewed once. This business used to work because first-hand material was hard to obtain and expensive to digest; you paid to buy time. Now first-hand material is there and AI is beside you. The value of someone else “digesting it for you” has fallen to zero.

So what should you do? Improve your own ability to learn and build.

My process is:

1. Find first-hand material and read it once yourself, marking what you do not understand.
2. Take those questions to people and AI. There is only one goal: understand the principle.
3. Once the principle is clear, hand the rest to AI.

Here is the key dividing line:

In the past, principles and details were learned together because you ultimately had to do the details yourself. Now it is different. Details are what AI does best: parameter tuning, code, and formatting. If the principle is right, AI does them faster and better than you. Learning details is competing with AI for its job—and losing.

Conversely, AI cannot replace principles. If you do not understand the principle, you do not know what to ask AI to do or whether it did it correctly.

So I changed my goal when talking with people. I used to want a complete “how to” from them; now I want to understand the one critical point hidden in their long explanation. In any field, one or two things usually decide success; everything else is detail around them. For a product, it may be “will anyone pay?” For a technology, it may be “what fundamental problem does it solve, and what does it cost?” Grab that point and let AI fill in the rest.

3. Beware AI long-form content taking over your attention

AI has made the cost of producing long articles zero. The result is thousands or tens of thousands of words everywhere, with complete structures, neat headings, and arguments that look reasonable. These pieces are excellent at occupying attention: you feel you learned something while reading, then remember nothing afterward.

They share one feature: they elaborate, but do not explain. They split one core idea into ten headings, fill each with three paragraphs, add examples, and end with a summary. Word count rises while information density falls.

The true core of a thing is often one or two sentences. Everything else is packaging.

So the third rule is: before reading anything, ask what the core principle is and whether it can be said in one or two sentences. If it can, those sentences contain all the value and the rest can be skipped. If it cannot, the piece probably has no core and is simulating depth with length.

This rule applies to me too. Instead of writing ten thousand words about how to use a tool, explain why it was designed that way and what fundamental problem it solves. The former is obsolete in six months; the latter remains useful.

4. Build resistance to interference: learn while doing and correct direction with real feedback

Even if you follow the first three rules, one last thing remains: hold your ground.

There is more interference now than ever. Every day brings a new model, tool, and trend. Every group has someone saying one direction is dead and another is taking off. Recommendation algorithms constantly put “what other people are doing” in front of you. Their shared effect is to make you doubt what you are doing and keep changing direction.

When I reviewed my failure to get results, the biggest reason was being led off course. It was not a lack of judgment, but an inability to hold a judgment. I saw someone doing A and doubted my B; I heard someone say C was the future and dropped what I had to look at C. Every switch cost a lot, and by the time I switched back, the earlier accumulation had gone cold.

What can correct direction? Not other people’s voices, but real feedback.

Learning and action should spiral upward:

• Action obtains feedback. Build it, put it out, and see whether users use it, pay, and stay. This is the only signal that will not lie to you.
• Thinking corrects direction. After feedback, look back: what was right, what was wrong, and was the key judgment accurate?
• If positive feedback does not arrive, return to learning and review rather than pushing blindly. Blind effort repeats the same mistake. Returning means finding first-hand sources again, understanding principles again, and checking whether the direction or the method is wrong.

Learning without thinking is confusion; thinking without action is emptiness.

Learning without doing fills your head with other people’s frameworks, without knowing which applies to you. Thinking without doing has no feedback to verify it, however clear it seems. Learning, thinking, and acting must turn together, one layer upward at a time.

Resistance to interference is not stubbornness. Stubbornness ignores facts; resistance looks only at facts, not the wave of opinions. Direction can change, but only through first-hand information and your own real feedback—not someone else’s sentence.

Finally

The four rules are really one sentence: do not eat leftovers.

Use first-hand information, understand principles yourself, reserve attention for the core, and correct direction with your own feedback. AI can accelerate these things, but cannot replace them. Details and execution are what you hand to AI.