NOW · 2026.09

What I'm doing right now

This isn't a list of things I've completed — just a slice of the present: what I've been reading, how I'm doing, and whatever's still on my mind.

Last updated: 2026-09-02

AR-glasses-wearing figure in a white coat sitting cross-legged reading, one hand holding a clay-colored mug, a glowing brain orb floating by the shoulder

LEARNING

Learning

  • Working through the undergraduate stage of clinical medicine, keeping medicine as the foundation for understanding real practice settings and risk boundaries.
  • Reading Du Fu: A Biography and Rich Dad Poor Dad lately — completely different books, so I switch between them.
  • AI-related things still come up naturally, but I no longer assign myself a dense learning checklist.

CURRENTLY

Current state

  • I've left the company where I'd been doing AI implementation, and I'm taking a break now.
  • I'm not packing the next phase full for now — first get back to a good state, then slowly decide what comes next.
  • The personal site and existing projects are still maintained, but I'm not chasing high-intensity output these days.

QUESTIONS

Mulling over

  • When has an Agent actually completed a business task, instead of just generating an answer that looks plausible?
  • How do failure alerts, result read-back, and human takeover become default product features?
  • In high-risk settings like healthcare, how do we distinguish knowledge assistance, product recommendations, and professional judgment?

INPUTS

How I process inputs

  • The main inputs right now are Du Fu: A Biography and Rich Dad Poor Dad. I'm just reading them through normally, not rushing to turn every book into a methodology.
  • I still watch long-form AI interviews, but I no longer feel I have to follow every model release and industry story.
  • Ideas that genuinely stick get written into the personal site; the ones I haven't thought through just wait.

GAPS

Filling gaps

  • I don't have a CS background; complex architecture, security, and high concurrency still need systematic study and working with people who truly specialize in them.
  • After running medical studies and AI work in parallel, I need a more sustainable rhythm instead of relying on short sprints.
  • Some product judgment comes from personal use and real-project observation; I still need more rigorous sample and evaluation design.

PRIVATE · ENCRYPTED

My goals

This section is more personal and always changing. It's encrypted — you need a password to view it.

Enter password to view →

DIRECTION · 2026—2028

Next two years: hold the main line

  • Main line: finish the clinical-medicine undergraduate stage and prepare for the postgraduate entrance exam (kaoyan). Medicine is the long-term path; AI is a tool that amplifies understanding and execution, not a lane switch.
  • Side line: AI projects keep going, but I'd rather accumulate verifiable work, public records, and judgment — not open many directions at once.
  • Medical content: let generation obey verification. Every conclusion has to trace back to a textbook, guideline, or explicit source. AI helps break things down; the human makes the judgment.
  • Pace: finish first, then expand. When something is interrupted, pick it back up without compensatory sprints. If execution keeps stalling, change the environment and the process first.

NEXT

The next small step

Rest properly first and finish the books I've got open. The main line for the next two years is set: complete the undergraduate stage and prepare for the postgraduate entrance exam; AI projects keep moving in small iterations, without competing with the main line for time.

Read the blog See the projects