Research notebook · started 2026-09-30

MDLC Industry Research

Research on the evolving model development lifecycle across ML, GenAI, and agentic systems.

Scope
classical ML · GenAI · LLM apps · agentic systems
Method
sources → notes → comparisons → patterns → frameworks → architecture → thesis

How this repository works

Each page sits at one layer of the research progression. Claims move rightward only with citations. Every statement is labelled as a fact, observation, interpretation, hypothesis, open question or decision.

Research system: sources flow into notes, comparisons, patterns, frameworks, reference architecture and thesis.
Fig. 0The research system.Open SVG ↗

Current thesis

As agents take over building and validation, the MDLC reorganizes around evidence and approval. Humans set direction and accept evidence at gates. Policy, knowledge and skills become versioned infrastructure. Thesis v0 →

Latest notes

    All notes →

    Latest open questions

      All open questions →

      Explore

      Search all pages →

      How to use this notebook · 每天怎么用
      When · 频率What to do · 做什么Say to Claude · 对 Claude 说
      Daily · 每天Ask one question. It is sharpened and logged as the next Q.
      问一个问题，磨尖后记为下一个 Q。      		“我想到一个新问题：…”
      When promoted · 值得深挖时Research it into a cited, bilingual note — only if it meets 2 of 3: evidence is findable · it affects a hypothesis (H) · you care about the answer.
      满足 3 条中的 2 条才写成笔记：能查到证据 · 影响某个假设 · 你关心答案。      		“这个问题值得研究吗？” / “开始研究 Q5”
      Every note · 每篇笔记Check whether it supports or contradicts an existing H; note it, don't rewrite the thesis.
      只标注它支持或反驳哪个 H，不改论点页。      		(automatic · 自动)
      Monthly · 每月Review the thesis: collect evidence per hypothesis, adjust confidence, record the revision.
      复查论点：汇总每个假设的证据，调整信心，记录修订。      		“帮我复查一下论点”