Chinese AI models are getting harder to ignore

Last week, during our community call, we started comparing the models we use.

One member had moved from Claude to Codex.

I had done the same.

Codex simply fit my current workflow better.

Then Kimi K3 came up.

─────────

🌏 Chinese models are entering the market

I haven’t tested Kimi K3.

When I tried to subscribe, paid access placed me on a waitlist.

Its published API pricing starts at $3 per million uncached input tokens and $15 per million output tokens.

For a model positioned around coding, reasoning and longer agentic work, that deserves attention.

AI is becoming a serious monthly expense.

ChatGPT, Claude, APIs and longer agent sessions add up quickly.

Naturally, people are starting to look for alternatives.

Most AI information reaching us comes through American platforms, developer tools and media.

Those channels naturally give American products more visibility.

Chinese models may become cheaper, faster or good enough for the same work without appearing regularly in our feeds.

To understand where the market is moving, we also need to follow what is being built in China.

For now, I’m watching.

I’ll test Kimi K3 when access becomes practical.

─────────

🧠 AI is changing what we remember

Another discussion last week started from something I noticed in my own work.

Since delegating more coding to AI, I remember less of the application architecture and fewer details about individual features.

My code-review skills are improving because reviewing is now a larger part of the job.

One member shared that he gives the requirements to an agent, lets it propose the architecture and improves the plan afterward.

That workflow is becoming normal.

I want to keep more of the technical thinking before the agent starts.

One: sketch the architecture.

Two: divide the feature into smaller steps.

Three: delegate one step.

Four: review it before continuing.

Technical ownership now includes understanding the architecture, the decisions behind it and every change the agent makes.

This was one of last week’s discussions inside The Way of the Dragon.

Rares 🐉