A few days ago I watched a video of an interview with Wang Xingxing, the founder of Unitree Robotics. At the opening roundtable of the 26th China International Industry Fair, he said something I had to rewind twice:
“The biggest mistake I’ve made over the past dozen years is that I didn’t believe in AI enough.”
That carries different weight coming from someone who builds robots and stands at the very front of the technology. He was not being modest. He said AI’s progress delivers “very big updates every month or two”; after GPT-6 shipped, Unitree put it on various robots to run task tests and “the results were actually very good.” He now expects GPT-7 next year may “punch straight through the problem” of embodied intelligence.
What says more is the timeline he has given: in August 2025, one to three years if things go smoothly; in August 2026, “two to three years at best, five to ten at worst”; and this time, “one to two years, if it goes fast.”
(The quotes come from the opening roundtable of the 26th China International Industry Fair, Shanghai, October 12, 2026, reported by The Paper, STAR Market Daily and Financial Times Chinese, among others.)
Someone who lives inside this industry every day still has his estimate swinging back and forth — and it has ended up at the most aggressive setting yet.
I kept thinking about that video. Not about robots. About companies like ours.
1. Even the people at the front admit they underrated it
Wang Xingxing is not an outsider. He was building robots as a student, built the quadruped XDog during his master’s, and later turned direct-drive motors and motion control into products. He understands this industry better than almost anyone.
Precisely because it is that person saying “I didn’t believe enough,” it is worth listening to. What he is really saying is this: I judged the future with the method I am best at, and I still judged it low.
The handiest tool for judging the future is past experience — take the slope of the last ten years and extend the line. In a stable industry that method almost never fails. It asks for exactly one precondition: that tomorrow looks like yesterday.
AI is the one thing that refuses that precondition.
2. What tightens the chest is not the speed — it is the gap
The speed itself is not news; everyone has been talking about it for two years. What tightens the chest is the distance between what we assume and what actually happens.
That distance is invisible most of the time. It does not hurt. Nothing shows up in the monthly report, orders keep coming, the spindles keep turning. It becomes visible all at once, at a single moment.
The moment looks roughly like this: you wake up, open the same door you always open, and for a second you do not recognize what is outside. Customers stop asking for quotes the way you are used to. The operation you were proud of gets done in one pass by somebody else’s method. The people you cannot hire are replaced next door by a few software tools and one operator.
The shock in that moment is not because the world moved too fast. It is because we kept measuring a world that had changed its scale with the old ruler in our hand.
Once I got that clear in my head, I actually felt calmer. Because once the gap is visible, it stops being fate and becomes a problem.
3. First thing: pick the tools up
My first conclusion was plain: adapt to the environment, and don’t fight the tool.
A shop like ours is naturally slow with new things. The machines turn, the orders come, the customers are still there — why change? That is honest, and it is the inertia of nearly every traditional shop. It is not laziness, it is steadiness. Steadiness is why we survived. It can also become the burden of the next decade.
But what I have learned these two years is this: your attitude toward a tool should not be decided by “will it replace me,” but by “does using it let me do the work in my hands better.” Those two sentences look close. They are not the same thing at all.
What we have been trying, piece by piece, are very concrete jobs: reading material off a drawing at the quoting stage, pulling dimensions out of a drawing, the explanations we have to write over and over. Let the tool take the repetitive, time-consuming part first, so the person can spend the freed-up hours on the judgment a tool cannot do.
It does not need to be grand. Even handing one job over changes how it feels. Standing there with a tool in your hand and standing there with nothing are two completely different postures.
4. A tool has no morals — the heart sets the direction
The same blade cuts vegetables and it hurts people. The blade does not know. The hand holding it knows.
More and more, I think a tool is an amplifier: it amplifies whatever the person using it already is. If the person wants the job done right, the tool amplifies the right. If the person is looking for a shortcut, for a way to fudge it, the tool amplifies the sloppiness just as fast.
So the stronger the tool, the heavier the decision about where to point it.
And this is where AI has an edge it cannot cross: it can give you an answer, but not a should. It can compute which route is cheapest, but it will not decide for you whether to skip a stress-relief step to hit a ship date. It can make a report look excellent, but it will not decide for you whether that report should tell the truth.
Once I worked that through, most of the unease I had been carrying dissolved. Because the thing that sets the direction cannot be replaced by it, and cannot be taken from me.
5. “Extending liangzhi” — a phrase I have understood slowly
This next part is personal.
Wang Yangming taught “extending liangzhi” — his term for the innate moral knowing every person already has. Honestly, when I was younger I filed that kind of talk under slogans: it all sounds right, and none of it lands.
Only in the last two years have I read my way a little further in. Liangzhi is not a moral slogan. Wang Yangming is explicit: “Liangzhi is nothing but the sense of right and wrong; right and wrong are nothing but liking and disliking.” He took that from Mencius, from the line about what a person can do without learning and knows without thinking — the capacity to tell right from wrong is already there. It needs no study and no deliberation.
Late in life he compressed his whole teaching into four lines: “The substance of the mind is beyond good and evil; good and evil arise when intention stirs. Knowing good from evil is liangzhi; doing good and removing evil is the investigation of things.”
What follows is my own reading, not Wang Yangming’s — there was no AI in his century. I am borrowing his frame to look at today.
A tool is like the first line, the substance of the mind: in itself it is neither good nor evil, it is just a thing. Good and evil are born where intention stirs — in what we decide to do with it. And “knowing good from evil is liangzhi” says this: the ability to judge whether that intention is right is something a person already has. The only question is whether you are willing to listen to it.
So “there is a weapon in our hands” needs its second half: a weapon has no morals. The morals are entirely in the heart that holds it — and in whether that heart still keeps to the thing it first set out to do.
Wang Yangming also said something harsher: “It is easier to defeat the bandits in the mountains than the bandit in your own heart.” We spend our time guarding against the chaos outside, and forget that the wish for a shortcut inside is the one to deal with first.
6. Conscience is not in grand theory; it is in today’s batch
Back to the shop. The word “conscience” sounds large. In a shop like ours it comes down to a few very concrete actions.
There is a line of Wang Yangming’s that has felt more solid to me every year: “A person must be tempered in the midst of affairs before he can stand firm — then he is settled in stillness and settled in motion.” Sitting quietly and feeling that your mind is settled does not count. The moment something happens and you fall apart, that settled mind was never real. Real steadiness is ground out in the work.
For us, what are “affairs”? Today’s batch.
When a customer places an order, we run every inspection step the quality plan calls for, and we do not drop one because we are in a hurry. When we say it ships in five days, it ships in five days — and if something genuinely goes wrong in the middle, we tell the customer early instead of hiding it and explaining on the last day. When a problem shows up, we do not point first at the material, the tooling or the outside vendor. We own it, then we fix it.
None of this is grand. But whether the customer ends up with “no worries” or “one more thing to chase” is decided exactly by these small things.
Reputation in this trade is never built by marketing. It is built, piece by piece, by customers not having to worry on your behalf. When the customer is at ease, the money follows on its own — they give you the next batch, and they introduce you to somebody else. And the reverse: if a customer has to chase us for progress every day and ride us about defects, whatever cost we saved has already been paid back double.
I have also always believed that the best state a company can be in is one where nobody on the team is shortchanged — in spirit or in money. That is not a slogan. When a person is not shortchanged inside, his hands are steady; steady hands make steady parts; steady parts keep the customer; and keeping the customer is what turns the whole thing back around.
7. Eat, drink, and keep living
Which brings us back to the person.
Everything above sounds tense, as if the sky were coming down. But the world is broadly the same world, and the days still get lived.
I have watched too many people lie awake all night over a trend they cannot influence. That anxiety is not worth it — it will not make you faster or safer. It just spends the days you already have.
Life is for experiencing. Eat well, drink well, enjoy the meal in front of you, and do the work in your hands solidly today.
Things that move fast can be left to time. What you can hold is whether this one step is planted firmly.
The real foundation of a factory was never an accurate forecast of the future. It is the kind of steadiness where whatever the outside turns into, today’s batch still comes out something you would put your name on.
So my answer to myself is short. Pick the tools up, and don’t reject them. Then let the ruler inside you decide where to point them, and how far to go.
I cannot control the speed out there. But the tolerance on the batch in front of me is mine to decide.
One line for you
A tool has no morals. It simply magnifies the heart that picks it up.
About Eternal CNC
We have made precision structural parts for twenty-three years. Among our thirty machines we run a five-axis, and we often take parts like robot joint-module housings, reducer components and load-bearing structures — the kind where “a little off and it fails.”
We would not claim to have “extended liangzhi” any great distance. But one thing we do hold to: use the tool when it is time to use the tool, and do not cut the corner where the corner must hold.
If a failure on your part is not an option, send us the drawing and we will get it right together.