Ai

My previous article covered the semiconductor cycle, and I feel like there’s a piece of background/context missing.

Your judgment/conclusion regarding this point—the overall direction is correct. Furthermore, I believe it is a prerequisite that is easiest to overlook when trying to understand this current semiconductor boom.

A more accurate way to put it is not that “all internet giants are fighting in the same field,” but rather: Large Models have, for the first time, brought together major players previously scattered across different domains—such as search, advertising, social media, e-commerce, office productivity, cloud computing, and content distribution—into direct competition within the same technical stack.

This technology stack includes models, computational power, inference, cloud, Agents, distribution gateways, and commercialization closed loops. Everyone’s original “moat” is different, but now we must all fill the same gap. Those who fail to do so will see their future search entry points, ad pricing, office suites, e-commerce conversion, and social traffic distribution rewritten by others.

The most unusual aspect of this current AI market cycle is not that Nvidia has risen sharply, but that the increase in value has been transmitted throughout the entire industrial chain: first GPUs, then servers, switches, ASICs, HBM, and finally to NAND, hard drives, power, and data centers.

If it were just a concept, the market trend shouldn’t last this long. But saying that it has already formed a complete profit cycle might be premature.

I prefer to view it as a “bull market driven by certain expenditures”: cloud vendors and model companies are genuinely spending money, and upstream companies are indeed collecting revenue, which is why stocks rose first; however, terminal applications have not yet proven that these investments can reliably generate enough profit, meaning the risk of a bubble also exists.

When I used to see phones with capacities like 512GB or 1TB, I always felt it was a bit excessive/wasteful.

The storage capacity here rivals that of a standard laptop. What exactly do phones even store that requires this much space? My previous understanding was very simple: just photos, videos, and WeChat data. You regularly transfer them to your computer, and you’re done cleaning up the phone.

I later realized that this judgment was actually heavily influenced by my own personal biases/habits.

I recently revisited this issue, and the conclusion is quite straightforward: It’s not that Huawei “cannot use TSMC”; rather, it is that U.S. regulations have severely constrained its compliant pathways. Xiaomi is able to access TSMC simply because it was never included in the same export control list as Huawei.

The most interesting thing about AI short dramas is not that they can immediately replace live-action short dramas, but that they turn genres that were previously considered “too risky to bet on” into viable subjects for experimentation.

Zombies, armies, spirit beasts, fantasy—these elements are incredibly difficult to execute in traditional live-action short dramas. It’s not a failure of screenwriting; it’s that every single step requires funding: extras, costumes and makeup, sets, special effects, staging/choreography, safety measures, and post-production. Furthermore, the commercial logic of short dramas demands speed, low cost, and high-frequency uploads. The greater the imaginative scope of a theme, the easier it is for costs to spiral out of control.

AI first revised this budget sheet. It doesn’t guarantee that every shot will be high-end, but it manages to constrain elements that previously required building sets, hiring massive crews, or doing elaborate special effects, down to a feasible/experimental scope. Zombies no longer need to organize 100 extras; mythical creatures don’t automatically drain a cinematic-level VFX budget; and army scenes don’t necessarily have to start with costumes and locations.

I saw that Fanqie Novel was embroiled in another controversy about AI-written content in the past couple of days. My initial reaction wasn’t surprise, but rather a sense that this issue will eventually come to the surface. Considering the combination of free platforms, the pressure for daily updates, and algorithmic distribution, it is almost inevitable that authors will turn to AI to supplement their content capacity.

But honestly, I have a consistent feeling about many books on Fanqie: they are readable, and even the first hundred chapters are often quite good. However, the further along you get, they tend to be nothing more than tropes and speed, lacking that inherent power that classic Xuanhuan or cultivation novels possess. This ‘power’ is hard to explain—it’s probably that feeling when you know it’s a bit over-the-top/cringey (“zhong er”), but you still want to follow the characters all the way through.

It’s something worth noting, and this isn’t meant to criticize the platform. Fanqie being completely free definitely attracts a lot of readers; there’s no arguing that point. However, for someone like me whose taste was spoiled early on by Tian Can Tu Dou, Wo Chi Xi Hong Shi, Er Gen, and Chen Dong, AI can match the output volume, but it cannot replicate the flavor/quality.

In the last few months, while writing code using tools like Claude or Codex, my most striking realization wasn’t that “programmers are obsolete,” but rather that many tasks that used to be given to newcomers for practice can now generate a basic first draft themselves. Whether it’s creating a scaffold, adding several tests, or making small modifications on the fly—after running through these operations, the speed is genuinely fast, so fast it feels almost bittersweet.

For someone like me, who graduated ten years ago, frankly, this is more about increasing efficiency. Because I generally know where it’s reliable and where it isn’t; where something looks functional but actually has pitfalls hidden further down the line. But for fresh graduates, this topic isn’t so straightforward. AI isn’t just here to take over a few hours of manual labor; it feels more like it is compressing the traditional path of how a newcomer goes from zero knowledge to proficiency. This is also why I wanted to write about it separately.

Stunned. / Dumbfounded.

The official ChatGPT side doesn’t make it easy to track tokens and costs directly, so I found a third-party platform and ran a round of similar tasks using GPT-5.4 and GPT-5.5 in Codex, setting the thinking mode to high. The result was very straightforward: simple questions were relatively mild (in terms of cost), with GPT-5.5 being about 30% more expensive than GPT-5.4; however, once complex tasks were involved, the costs shot up to 2.6 times, and both the request count and token consumption increased simultaneously.

My current assessment is very straightforward: this isn’t something that can be decided just because of the statement “5.5 has a higher unit price.” In simple tasks, the cost mainly comes from the unit price; but in complex tasks, what is actually expensive is the entire calling chain (or execution flow). However, looking at it another way, 5.5 does genuinely feel like it’s absorbing your rework costs for you. The model is more willing to think through multiple steps, perform more actions, and check things more thoroughly. Ultimately, the billing isn’t based on a single answer; it’s based on the complete set of actions, which also minimizes the number of back-and-forth cycles required from the human user.