Author: Wang Jianshuo, Founder of Baixing.com
Many people readily proclaim, "Large models are everything." I don't quite buy it.
Every time I hear phrases like "devour everything," I feel it's mostly because our understanding of the future hasn't reached that level, hence we casually throw out such a broad statement. Otherwise, how could one thing devour everything? Take the internet—it's been shouting about devouring everything for years, has it truly devoured everything now? So which one is it, the internet devours everything, or large models devour everything? Both are devouring, leaving nothing behind?
So I'd rather put it differently: it is a very important foundation. Without this foundation, the whole world can't develop, just like the internet without its underlying backbone network, or like using electricity without power plants. That, I acknowledge.
But after the foundation is in place, that's where the real excitement happens above.
Take electricity. As soon as electricity was generated, what was the first application people saw? The light bulb. Thomas Edison lit the first one, and it just kept shining, shining, shining. If the world stopped there, with only a light bulb, then I could completely say: the power plant is the core of the whole world, the power plant devours everything.
But that's not how things went. Later came motors, to drive machines; and then you find that once a fundamental thing like electricity exists, countless appliances grow on top of it to utilize it. Washing machines are for washing clothes, televisions are for watching TV, vacuum cleaners are for vacuuming—they are all applications of electricity. Without electricity, none of these things would exist. But if you say "electricity devours everything," I don't believe it.
It's the same with large models. They provide foundational intelligence. But this intelligence must be placed into a specific, scenario-oriented "machine" or "device" before it can play its role and truly change the world.
Claude Code is for writing code, Claude Design is for design, VoiceDrop is for writing articles. Using the same large model, placed into different devices, solves completely different problems.
With only electricity, only water, without a washing machine, clothes still can't be washed. Imagine, a power plant produces a massive amount of electricity, electricity is very powerful, and then what? Without a washing machine, can this pile of electricity wash clothes by itself?
Intelligence is great, but most things in the world require the combination of multiple elements to work, just as a washing machine needs to combine elements like electricity, water, and even a tub; large models might be able to replace a lot in the field of software, but application scenarios in the world that require only one element are not many.
Here's a current example. Now we have large models, but having just a large model isn't enough. There needs to be a layer above it called Harness—this layer has only recently emerged—it interacts with code, and finally forms something truly usable. A large model by itself can't write code. Of course, the core of Claude Code, frankly, I could write it with a little over fifty lines; a bit longer with a few more lines, and it can run and write programs. But you have to see: with only a large model and without this outer layer, it's still not easy to use—that is, the intelligence of a large model, if not combined with the code execution capabilities provided by the operating system, relying on the large model for calculations isn't economical, and sometimes even impossible.
The core value of this interface layer is to help us put that intelligence, which is like electricity or water, into a specific application scenario, turning it into a machine that can solve specific problems.
Now, of course, the logic behind "devour everything," I don't completely disbelieve it either.
What it primarily refers to is existing software. So far, we have built up a very, very large layer of software—things pieced together by many rules, forms, buttons, workflows, a lot of it. Massive numbers of filters, fixed templates, a bunch of backend operations, many SaaS detection features. And all the various things we used to know as "M" systems, whether CRM, or HIS (hospital information system), all sorts of so-called "systems," "software," and the like, a whole pile.
This layer, I believe large language models will indeed devour quite a bit of.
Why? Because these original software programs were, by nature, clear, computer-executable instructions, solidified and repeatedly executed—we call this software. And this is precisely what large language models are most adept at chewing through.
But.
Within this layer, besides software, there are many other things. Customer information. Execution capabilities—like when you book a flight ticket, the actual capability to move a plane, to transport a person from here to there. Also trust. Many things in the physical world. These, I don't think will be devoured.
After devouring that layer, it actually opens up a much larger space—the new type of software on top of it.
The interface of the new software will certainly be flow-based, it likely won't solidify as many rules as before. After handing all those rules to AI, think about it, the CRM we could achieve like Salesforce before, that was already the pinnacle of human effort, using all possible strength. But if this part becomes relatively easier to solve, then what everyone will do next is to unleash even more imagination and possibilities on top of it—and that part is precisely what we haven't seen yet.
This is where we often make mistakes. When a new technology arrives, because we can't see the bigger road beyond it, we can only stare at the part in front of us. The leaf blocks the view, obscuring the mountain.
Let alone this kind of trend judgment. I remember back in 2004, a group of friends were complaining, saying that the entire internet could no longer produce companies bigger than Sina, Sohu, and NetEase, that the internet was nearly over, and they would monopolize everything. And look how many years have passed since then? The world turned upside down. We would be crying at our own shortsightedness back then.
So my proposition is this: Are large models important? Yes, they are the foundation, the main driving force in recent times. But once they become stable, continuously provided things, then various kinds of "machines" and "devices" on top are needed to solve specific problems. That thick layer—where it's used, how it's used—is precisely the mainstream of the second wave of this trend.
The phrase "devour everything" is too imprecise. In this world, which thing, which social form, which technology has ever truly devoured everything?
Finding opportunities in the places it does devour—that's the truly important thing.
