Skip to content

Monthly Refresh 2026.06

keyword

  • the driver licence test (second try)
  • prod-release
  • typhoon and earthquake

journal

  • I failed the driver's licence test on Jun 1st, which felt like a head-shot — I had thought it would be quite easy.
  • I also tried to implement my own project. Due to my lack of experience, I had to rework a lot: I didn't have a grounded idea (the design), so the implementation (the steps) didn't fully follow it...
  • Several typhoons and frequent earthquakes — no big deal, but still, life wasn't always at peace.
  • I stayed up quite late these recent weeks: too many things, too little time.
  • The prod release itself was smooth, but issues still surfaced afterwards, because the test env and the prod env (running 24/7) differ. - And I didn't have enough context or capability to write tools and investigate the issues efficiently.
  • I studied a few columns but didn't finish them all, and it's quite hard to say what I learned from them, even though I set OKRs before spending any time.
  • I switched to a new platform for reading books, thinking I might focus on one book at a time — and yes, it worked as expected. But, hmm, I still didn't finish many books this month. Well, I finally practiced the skill of quickly checking a book (cover, TOC, chapter start/end) to assess it, and that was really effective for dropping the unnecessary ones.
  • k3s + tailscale + traefik turned out to be a better solution than my previous k3s + tailscale-subrouter + svc: no more latency and everything runs smoother. I deployed even more services to my k3s — now 50+ pods running 24/7.

collective

ActType Effort Why
Run 36km/7 Rain, lazy
Book 2 didn't spare extra time and effort

conclusion

I need to fail fast and get regular feedback. (Like I always tell the agents: learn your lesson and persist it to your permanent context.)

resolution

  • daily effort
  • financial orientation

sharing

  • https://charitydotwtf.substack.com/p/ai-enthusiasts-are-in-a-race-against - > know it before you criticize it
  • https://role-confusion.github.io/ - A Theory of Prompt Injection (and why you should study roles) - > how injection happened in agent context
  • https://binaryigor.com/the-joy-and-power-of-understanding.html - > AI is good, so as you (should)

Comments