A few things
I keep thinking about.

I am interested in how ideas become systems—and how those systems earn our trust.

That interest takes me from questions about compression and intelligence to the practical details of agent environments, electromagnetic simulation, and physical design. I care about both the abstraction and the artifact that survives contact with reality.

Why “Bits to Mind”?

I find the relationship between compression and intelligence deeply compelling. A useful representation seems to do more than retain observations: it captures something reusable about them. But a compact model is only part of the story. Reliable action also needs tools, feedback, state, and a way to check the result.

The name describes that direction of inquiry. It is not a claim that fewer bits necessarily produce more intelligence, or that a single principle explains everything.

What I build

My engineering focus is on environments in which AI agents can do meaningful RFIC design work: exploring authorized information, making circuit and geometry choices, running experiments, repairing failures, and producing results that remain inspectable and editable.

I prefer reusable general-purpose agent runtimes where they are sufficient. The domain-specific investment belongs in the workbench: the sandbox, the instruments, the task definitions, and the independent evidence. The research page gives a fuller picture of these directions.

How I approach the work

Freedom inside; clear authority at the boundary. I want agents to reason and experiment, without giving them the authority to rewrite what counts as success.

Artifacts over transcripts. A persuasive explanation is useful, but it is not a replacement for the design state, the evaluation conditions, and the results.

Complexity should earn its place. I prefer a small, recoverable workflow to an elaborate organization whose costs and responsibilities are difficult to explain.

Claims should carry their scope. A demonstration, a repeatable benchmark result, and a production-ready capability are different achievements. Each deserves its own evidence.

What belongs here

This is a personal notebook for essays, working frameworks, research questions, and small experiments. It is not a product catalog or a substitute for a technical paper. Ideas can be useful before they are complete, provided that the uncertain parts stay visible.

The recurring subjects are compression and intelligence, AI × RFIC, agent systems, and the engineering methods that connect them.

Find me online

You can find my public profile on GitHub and follow new essays through the RSS feed.

Email: yxiaohan@bitstomind.com.

Technical corrections, alternative explanations, and careful counterexamples are especially welcome.