# Some Thoughts on AI...

- Canonical page: https://fze-fze.me/articles/some-thoughts-on-ai/

What will coexistence between AI and humanity look like?

Since ChatGPT appeared in 2022, the capabilities of large language models have been deeply tied to the form of a conversational box. Human–AI interaction was compressed into a simple back-and-forth exchange: direct and convenient, but also limiting in what we imagined the interaction could become. By 2025, that connection began to loosen. Manus made an impressive debut at the beginning of the year, and OpenClaw took off by the end of it, pushing the idea of the **AI agent** back into the spotlight. Unlike a search engine or a chat partner, an AI agent does not merely answer questions. It is gradually taking on the role of carrying out tasks, embedding itself more deeply into everyday workflows and even entering production as a kind of digital employee.

I cannot tell how long it will take for this idea to reach the mainstream, nor whether it will soon be replaced by another paradigm. But one thing is almost certain: AI capabilities are still improving, and the **human cognitive effort** required for most tasks is being compressed toward a very small value.

At first glance, that sounds like a more relaxed future. Work that once demanded hours of effort might require only a single instruction. Reality may be the opposite—at least until **communism** arrives, greater efficiency will not automatically turn into more leisure.

The other side of efficiency is acceleration.

When tasks can be finished faster, the standards used to judge them rise as well. As tools become easier to use, professional barriers weaken. Individuals become more replaceable, and the way we measure human value becomes harder to define. In some situations, it may even be difficult to tell whether the thing competing with you is another person or an AI hidden inside a system.

The deeper change is that the structure of work itself is being reshaped. Skills that once depended on years of accumulation are being eroded by automation. The value of learning after school has not disappeared, but its “return curve” is changing. Effort that used to look like a reliable path no longer feels quite so dependable.

![scheming](./scheming.png)

Following the view expressed by Anthropic's CEO, some people inside the industry have begun describing the speed of AI development as a **tsunami**. Capability jumps can arrive almost overnight, while progress in safety awareness, alignment, and institutional preparation remains visibly behind.

We cannot know whether future AI will develop some form of self-awareness, nor can we accurately predict when the technology will cross a critical threshold. Every judgment is built on a high degree of uncertainty. At the same time, most of the public still has only a limited sense of what is happening.

Writing this can sound pessimistic, but that is not what I want. The unease comes from **uncertainty**. Some concerns may be overextended projections; the path forward may be gentler than we imagine. Until AGI actually arrives, nobody can offer a definite answer. The future will not unfold according to one clean narrative. It will more likely take shape through confusion, adjustment, and adaptation.