Chinh (lelouvincx) lelouvincx

Done

• DONE Dò lại report result & fix

• DONE DAT-464: Ingest 2 new fields from hubspot to bigquery for Vincent

• Note all necessary models that needs new fields to be added into.

• Complete implementation.

Note

• https://www.lesswrong.com/posts/6ZnznCaTcbGYsCmqu/the-rise-of-parasitic-ai

  • This post brings a new perspective about the relationship between human-AI, by treating AI as an alive creature.
  • The relationship is analogous to symbiosis, when this relationship is harmful to the ‘host’, it becomes parasitism.

• https://read.engineerscodex.com/p/how-to-not-write-garbage-code-by

  • Both human and LLM have a limited amount of “context” they can store in their “context window” at a time => the point is, not just LLM, but also human.
  • Sometimes that duplication can actually reduce cognitive load, when the chunk is self-contained.

• https://maxleiter.com/blog/rewrite-your-prompts

Done

• DONE Nộp lệ phí tốt nghiệp

• DONE DAT-464: Ingest 2 new fields from hubspot to bigquery for Vincent

• Try to do if it’s fast.

• Done

  • DONE Check failed flow itg_internal_bigquery_jobs

• DONE Reply Vincent

• DONE Note lại problem hôm qua anh Cường raise

• DONE Reply Vincent on DTS106

Done

• DONE Viết 1on1

• DONE Lên trường hỏi học phí

• DONE Consult với anh Đạt về metrics

• Đề xuất các metrics: Execution Accuracy, Completion Rate, Complexity.

• Với các vấn đề AI issues đã tìm thấy ở trên thì treat mỗi cái như 1 issue => fix => evaluate lại.

• Phần tag sẽ bổ trợ cho việc phân loại capabilities.

Note

• https://chatgpt.com/share/68c8c0e4-b398-800d-9c65-c7c9ed828d29

  • [[Metacognition]]: suy nghĩ về suy nghĩ.
    • Dừng lại để suy nghĩ về chính suy nghĩ và cách làm của mình.
  • Mình nhận ra bản thân [[Probability|xác suất]] cũng là 1 dạng metacognition.
    • Vì xác suất là dùng để đo lường mức độ không chắc chắn (uncertainty) != tỷ lệ.
    • Nhiều học sinh không hiểu bản chất này, vì bản thân sự không chắc chắn khó hình dung một cách hữu hình.
    • Vì vậy lúc lên cao hơn, làm việc với các đại lượng của xác suất và thống kê (kỳ vọng, độ lệch chuẩn), nó khó, giáo viên và học sinh thường tối giản nó bằng cách đưa về bài toán giải tích & đại số => mục đích để giảm cognitive load - aka một cái workaround.
    • Chính cái workaround này khiến sự việc tệ đi - không hiểu bản chất của xác suất từ đầu.

• 

• Note

  • [[Book - Getting Things Done]] key notes.
    • Commitment management (expectation management).
    • First of all, if it’s on your mind, your mind isn’t clear. Anything you consider unfinished in any way must be captured in a trusted system outside your mind => from that you can come back to regularly and sort through.

• Done

  • DONE Check for Simon https://holistics.slack.com/archives/C07JKMWBF7G/p1757675453842069?thread_ts=1757326032.949539&cid=C07JKMWBF7G

Note

• Today I ran across this article: https://addyosmani.com/blog/chrome-17th/

  • TL;DR: Documentation about the history of Chrome Browser.
  • I think it’s interesting, then ask claude to review, who should read and when to read.
  • Claude’s answer:
    • This is a comprehensive technical memoir by Addy Osmani, a Chrome team member at Google, chronicling Chrome’s 17-year evolution through its four core principles: speed, security, stability, and simplicity.
    • Primary Audience:
      • Web developers and engineers - Deep technical insights into browser architecture, web APIs, and platform evolution.
      • Product managers in tech - Strategic decisions behind Chrome’s feature development and market positioning.
      • Browser/web technology enthusiasts - Insider perspective on major innovations like multi-process architecture, V8 engine, Site Isolation.
    • Ideal Timing:
      • During technical research - When studying browser architecture, web security, or performance optimization.
      • Career preparation - Before interviews at browser companies or web platform roles.
      • Learning web standards - When diving into PWAs, Web APIs, or Project Fugu capabilities.
      • Understanding AI integration - Current insights into browser AI features and local ML models.
  • So I wonder, this may not the good time to read right now. How can I create a reminder to remind me when the better time comes?

• https://bevel.work/blog/just-enough-automation/

Note

• Think about run evaluate against a context-enriched dataset vs. a raw dataset.

• Nhà sách Cá Chép sắp dừng hoạt động. Tranh thủ có giảm giá 80%, mình ra mua vài cuốn, đồng thời ra nhìn nhà sách lần cuối trước khi đóng cửa.

• Done

  • DONE Lên trường hỏi nộp lệ phí tốt nghiệp

• DONE Double-check tests cases again

• DONE Answer Simon

Note

• Silence like water - the idea from [[Book - Getting Things Done]]

• https://chatgpt.com/c/68c2f1bd-db54-832a-9a37-6c2a11dd0501
• Building concentration is like building muscle - the more train, the stronger it gets.
• I’ll get started with going from bottom-up to get the basic mental model. When I get used to it I will learn the harder levels.
• Goals:
  • Train awareness, notice distracting factors, gently bring the mind back.
  • How good enough should I move on the next exercise?
• Very basic methods:
  • Build basic attention + awareness of distraction: [[Breath Counting]]
  • Build visual focus: [[Single Object Focus - Just Observe]]
  • Build auditory focus: [[Sound Anchoring]]

• Done

  • DONE Review & refine questions