A Get-Together on Where AI Is Heading
Yesterday's "meeting." If I told you we talked about AI Agents, Vibe/Agentic Coding, or Sơn's broken CI/CD (cursed by Nguyện's bad luck)... well, that would be too ordinary.
Yesterday, we actually talked about... quantum physics
The cat sits in a sealed box, both alive and dead at once — the outcome depends on you opening the box to observe. The state only "settles" when someone looks.
So why is a quantum computer so much more powerful than a regular one? Because it uses the superposition principle. Where a regular computer only has bits of 0 and 1, a quantum computer uses qubits — which can represent 0, 1, or both at the same time.
Sizing up the coding tools
Excellent UX. It indexes (RAG) the codebase, so it's sharp at finding issues and related logic. Yet somehow it just doesn't get the FOMO the others do.
The cheap-token era is ending
Coding Agents keep getting pricier. The days of cheap tokens for "coding all day long" are almost over.
Even a Claude Max 20X plan at nearly 6M VND/month isn't enough for a Solo Builder who really grinds. Are we heading toward a day of "I fired the AI and hired back a developer"?
The fix: package your own Skills
You should package your own Skills. It not only reduces context dilution, but is highly efficient because each Skill solves exactly your own task.
- 01Use one thread for a repeated task
Open a thread and do a task you frequently run, over and over.
- 02Have the AI Agent package it
Then ask the AI Agent itself to wrap everything you just did into a Skill of your own.
- 03Reuse it
Next time, just call the Skill — clean, efficient, less dilution.
It's not the strongest model that wins — it's whoever packages their own workflow into something reusable.
