I first learned the phrase “begin with the end in mind” from The 7 Habits of Highly Effective People, and it left the deepest impression on me. It says that when we start doing something, we should already be thinking about its purpose — its endpoint. This goal-oriented, targeted way of working and living helps us reach our goals and keeps us from drifting off course or getting lost along the way.

Since learning “begin with the end in mind,” my way of working has changed a great deal: I frequently check with collaborators that we share a consensus on the desired end state, estimate whether our roadmap can get us there, and how far we still are from the final goal. A recent post on X, How to fix your entire life in 1 day, discusses a similar problem. Most goals fail to be sustained because the person executing them never truly thinks through what the end state looks like. Only when we deeply understand the goal we pursue can we execute with direction.

I used to believe “begin with the end in mind” was the ultimate method for work and life — until I read the Diamond Sutra and discovered there is another mode of thinking. In that mode, the final state (liberation, Buddhahood) and the process (saving all sentient beings) are separate, even contradictory. If you treat becoming a Buddha as the final goal, you will never become one. Conversely, if you take saving all sentient beings as your mission, thinking not at all of yourself (no self, no person, no sentient being, no life span), then you may become one after all.

A few excerpts from the Diamond Sutra to convey the “contradiction” between process and endpoint:

“However many sentient beings I thus liberate, no sentient being is liberated. And why? Subhuti, if a bodhisattva has the notion of a self, a person, a sentient being, or a life span, that one is not a bodhisattva.”

“If a bodhisattva were to say, ‘I shall adorn the Buddha-land,’ that one would not be called a bodhisattva. And why? The Tathagata teaches that adorning the Buddha-land is no adornment, and that is called adornment. Subhuti, if a bodhisattva penetrates the dharma of no-self, the Tathagata calls that one a true bodhisattva.”

“Do not say the Tathagata thinks, ‘I shall liberate sentient beings.’ Subhuti, do not think that. And why? In truth there are no sentient beings for the Tathagata to liberate. If there were sentient beings liberated by the Tathagata, the Tathagata would hold the notions of self, person, sentient being, and life span. Subhuti, the Tathagata says, ‘Those who hold a notion of self in fact have no self, yet ordinary people believe there is a self.’ Subhuti, the ordinary person, the Tathagata says, is no ordinary person — that is merely called an ordinary person.”

The Diamond Sutra is full of passages like these. Plainly put: we should practice, but not cling to the goal — everything should arise from the heart, free of utilitarian intent. Only by not chasing Buddhahood can one attain it. This is perhaps the most demanding body of knowledge to grasp.

Though it seems paradoxical, many things in life actually fit the logic the Diamond Sutra emphasizes far better than the “begin with the end in mind” logic of aiming wherever you point.

For instance, we often pursue a happy life. One approach is to imitate the people around us who look happy — perhaps they are wealthier, have more free time, or have other such traits. But if we blindly imitate and chase those traits, we may struggle to become truly happy; worse, the gap with reality may make us more anxious.

For instance, we pursue material abundance. One approach is to maximize salary as the goal of work — choosing whatever major is hot, job-hopping to wherever pay rises. Though this is extremely “begin with the end in mind,” few people actually amass great wealth this way. As one line from the Nike 10 principles goes: If we do the right things we’ll make money damn near automatic.

These things that resist the “begin with the end in mind” approach lean more toward personal psychology and career development. You could say that much of what we see on the surface is a byproduct of underlying behavior. If we chase the surface, the harder we try the further we get. As Su Shi reportedly said on his deathbed, “着力即差” — striving is already a miss. Many things simply cannot be attained through obsessive pursuit. What we can do is cultivate our inner selves, learn to live in the present, and experience inner richness. As for those surface-level outcomes, they are entirely byproducts.

After writing this piece, I found that someone had articulated a similar view. Viktor E. Frankl wrote in Man’s Search for Meaning:

Don’t aim at success. The more you aim at it and make it a target, the more you are going to miss it. For success, like happiness, cannot be pursued; it must ensue, and it only does so as the unintended side effect of one’s personal dedication to a cause greater than oneself or as the by-product of one’s surrender to a person other than oneself. Happiness must happen, and the same holds for success: you have to let it happen by not caring about it. I want you to listen to what your conscience commands you to do and go on to carry it out to the best of your knowledge. Then you will live to see that in the long-run—in the long-run, I say!—success will follow you precisely because you had forgotten to think about it.

If I were to name this mode of thinking, I would call it “indifference to the endpoint.” What we do in this moment springs entirely from present passion or values, with no expectation of results and no concern for others’ opinions.

Other

• Waymo World Model — an example of generative closed-loop simulation and data production. Well done, worth studying. Compared with other work, its highlights are multimodal vision-plus-lidar output and lots of open-set scene generation, such as inserting wild animals. On close inspection, the generated lidar point clouds are still not perfect — some structures show bending deformation. Whether this technology can be applied to real model iteration so Waymo can scale globally fast is probably what everyone cares about.
• Generative Modeling via Drifting — another tour de force from Kaiming in generative modeling. If the approach generates well, its single-step inference advantage will see it widely adopted in academia and industry. Worth following long-term.
• The Second Pre-training Paradigm. Jim Fan is still a master of riding trends — he can distill an academic trend into a movement and lead people forward under a vivid banner. From the perspective of a “traditional deep learning researcher,” providing more, cleaner, and more comprehensive supervision is always good, and the “second paradigm” here mostly refers to prediction tasks beyond text. As for whether this deserves the label world modeling, I’m skeptical. I recall it was Jim Fan who first called Sora a World Model.
• In January 2026 SpaceX filed with the US FCC to deploy up to 1 million satellites forming an “orbital data center system,” mainly for large-scale AI compute. Building such a data center involves power supply, cooling, communications, system stability, and so on. From the available information, space seems to have advantages over the ground in both power and cooling. I’m curious whether the communication latency among a million satellites can support large-scale training. Musk has said, though, that these clusters would mainly serve inference — and in RL training most compute also goes to inference. From that angle, even running each satellite independently seems workable. Let’s see how this plan unfolds.