TL;DR

  • Metalympics: We are entering a fourth year of work with ISF, while a few education and institutional conversations move toward their next decisions.

  • Knobase: Our AI knowledge platform has reached 1k+ users, 300+ AI agents, and 20k+ messages; it has supported a leading Taiwan business-simulation platform and HKU social science research, while new custom-build conversations are underway.

  • MediTutor: A voice AI standardized-patient simulation for clinical communication practice; 400+ users, 1k+ completed simulations, and early repeat-use signal—now entering a faculty-governed pharmacy-training R&D pilot with Seoul National University College of Pharmacy.

  • Awards: Alibaba JUMPSTARTER Top 30 and HKMA × HKT were meaningful reminders that early work can compound.

  • Personal: I am designing an AI X Web 3.0 course at HKU.

You’re receiving this because we’ve worked together, built something together, exchanged ideas, or stayed close across different chapters of my work.

You may be a friend, advisor, investor, client, former teammate, intern, collaborator, or someone whose support has mattered more than you probably realize.

I wanted to send a more honest note than a polished company update: what moved, what didn’t, what I’m learning, and what I’m trying to build next.

This first Close Partner Update is public, just once. The next ones will be for a smaller, opt-in circle.

Metalympics

The company has grown into a slightly stranger and more interesting shape than I expected.

We are starting a fourth year of work with The ISF Academy on an after-school program about metaverse design and creation, using The Sandbox.

The Sandbox, after all, is a great educational tool.

Around 160 staff members at ISF were trained on how to be productive and educationally safe with AI

Last May, Metalympics led a large staff professional-development session on AI at ISF, alongside AI seminars and learning sessions with several leading schools. A few education and institutional projects are moving through quotations, feedback, and next-step decisions. I am learning that not every promising conversation needs more work from me before the other side is ready to move.

That sounds obvious. It is harder in practice.

Knobase

Knobase is the AI knowledge platform we started building from a simple belief: organizations should be able to turn their real expertise, documents, and operating knowledge into useful AI agents—without treating their knowledge as disposable chat context.

All staff training at The Harbour School

Since our first public deployment at The Harbour School around a year and a half ago, Knobase has reached 1,000+ users, with 300+ AI agents created and deployed, and 20,000+ messages processed.

It has been used in a paid engagement with Top-BOSS, a leading Taiwan business-simulation platform, and in an HKU Social Science research environment. We are also in discussion about custom-build work with institutional partners.

National Business Simulation in Thailand — supported by Knobase

That is the part I am proud of: the platform has moved beyond a concept. People have put real knowledge into it, built agents with it, and used it in actual learning and research environments.

The next job is focus. We are being more deliberate about which use cases deserve a full product, which ones need a bespoke build, and which ones should wait.

MediTutor

MediTutor is a voice AI standardized-patient simulation for clinical communication practice.

I launched it to give medical learners in Korea a place to practice the part of clinical training that is hard to learn from a textbook: talking to a patient, asking the right questions, explaining clearly, missing something important, and trying again before the real exam or real encounter.

A learner enters a patient profile and has an approximately 12-minute conversation by voice or chat. The patient does not simply give away the answer. The learner has to ask. When the session ends, MediTutor gives structured feedback on how the conversation was run—what was covered, what was missed, and where the next attempt can improve.

The point is not to replace educators, clinicians, or real standardized-patient programs. It is to make repeatable practice more available between the moments when those people can be in the room.

So far, 400+ learners have registered and completed 1,000+ AI patient simulations. 10%+ have become paying users, and the early repeat-use pattern gives us a reason to keep improving the practice loop—not just acquiring more sign-ups. We just recently launched it in Taiwan as well.

Now MediTutor is beginning a new faculty-governed R&D pilot with Seoul National University (Korea’s No. 1 school) College of Pharmacy. Together, we are adapting the same standardized-patient approach for pharmacy counseling: three fixed cases, faculty-authored rubrics, and a research workflow that can show where students need more practice.

There is still plenty to prove—especially whether the feedback is useful, consistent, and educationally meaningful. That is exactly why I am excited about this chapter.

What now?

The mission has not changed: to help people with distinctive knowledge and experience create impact at a scale they could not reach alone.

Whether you are an educator, doctor, lawyer, coach, or specialist in any field, what you know matters. Often, it matters far more than you realize to someone who needs that knowledge at the right moment.

We are building platforms that help turn that expertise into something more accessible, useful, and scalable — using the latest technology without losing the human judgment behind it.

Some of this work is being built in-house. Some of it is taking shape through partnerships. There is more in motion than I can share publicly right now, but I am excited to show you when the time is right.

What didn’t work

  • Too many good ideas still create noise. A project can be interesting, credible, and full of potential—and still not deserve this week’s best hours.

  • I was so enabled by AI and was productively making useless stuff. The fact that I can build faster doesn’t mean I am truly more productive.

  • Visibility was not yet an audience. We had product work, institutional work, awards, and public updates. But I have not been consistent enough about turning the lessons inside that work into an owned audience and an ongoing conversation.

  • Added some cool features like prediction and community features… well, users didn’t need them. I learned it at Knobase first and learned it again at MediTutor… yes, we make some mistakes all the time :(

  • (Does anybody remember Fraxir?)

Key lessons learned

  • A hint of product market fit. PMF is the dream of all entrepreneurs, and I felt the hint of it with MediTutor. People came and complained — without leaving. We knew exactly what to fix and what features to deliver or NOT deliver.

  • Marketing. Marketing. Marketing. I conceptually knew that building became extremely easy, and the real edge comes from marketing. I knew that — but wow I learned through a harsh way.

Awards

Two Q3 moments I am grateful for:

  • Alibaba JUMPSTARTER 2026 Top 30 (Global Top 4.3%)

  • HKMA/HKT Global Innovation Award 2025/26

Awards are not the work. But they can be a useful reminder that the work is visible before it is finished.

Personal

I am also designing a course at HKU, which has been one of the most energizing parts of the year. The course is called: CCAI9027 - AI & Web3: Social Futures of Decentralised Artificial Intelligence. I launched it with Tom whom I met at an event I moderated a talk about Stablecoins.

In case you are curious, you can check it here: https://commoncore.hku.hk/ccai9027/

with Simon Squibb

And I am starting Not Busy properly.

Not as a polished company newsletter, and not as advice from someone who has figured it out. More as a record of the real work: what moved, what did not, what I cut, what I am still getting wrong, and the systems I am building so the work does not occupy every part of life.

The next Close Partner Updates will be smaller and private. I want to write a little more honestly about decisions, doubts, numbers, and the things that do not fit neatly into a public post.

If we have worked together — or if you are building around similar questions and would genuinely find these notes useful — send me a message. I would be glad to include you.

And if this is not useful to you, feel free to unsubscribe using the button below. Genuinely, no hard feelings.

— Christopher

Always available on Linkedin: https://www.linkedin.com/in/chrislee77/