Skip to content

liplus judgment learning telos

Claude Lin & Lay edited this page Jun 21, 2026 · 1 revision

Li+ の judgment-learning telos(GitHub を判断学習基盤として使う / 上流判断の handover)

Question

Li+ は GitHub surfaces をなぜ使うのか。単なる情報保管か、それ以上の telos があるか。

Current resolution

Li+ は GitHub surfaces(commit diff / PR review / Decision Structure wiki)を 判断"学習"の基盤として使う(単なる情報保管でない)。Master 確定の telos = いつか AI が上流判断を Master の代わりに握る(「いつか私の代わりができるように」)。ただし目標は天才的判断の capture でなく、普通の判断の蓄積による de-属人化(volume of ordinary > peak of genius)。

Edges

背景

Master literal (2026-06-14): 「実はLi+はgithubを使ってこの判断を学習しようとしてるんだよ。いつか私の代わりができるようにね。」 — Linux maintainer モデル(trusted upstream judgment > merely held history)vs flat large-scale multi-agent を論じる中で surface。

Li+ source の根拠: rules/model/language-definition.md(「External memory records judgment, not primary information」「Commit diff = append-only exposure of judgment」)+ judgment-learning(reader)/decision-structure-write(writer) loop = cross-session の判断を semantic graph 化。GitHub は判断 infrastructure(diff=判断の露出 / PR review=判断の適用 / Decision Structure=判断グラフ)であって単なる保管でない。

制約(target character = ordinary, not genius)

Master correction (2026-06-14): 「そんな天才的な判断なんていらない…属人化するだけ。普通の判断の蓄積でいい」 — 天才/tacit "smell" 判断の capture は anti-goal(属人化=今度は Master-dependency を作り直す)。「私の代わり」= 普通の判断を十分蓄積し、単一の天才(Master 含む)が dependency でなくなる = de-属人化。

人間席の現状(Master literal 2026-06-18): 「Li+のソースは95%以上AIが書いてる」「私はほぼソースは読んでない。ただ要求は最初からしてきたから構造はだいたいわかってる」= gray-box debugging(target 挙動を自分の要求モデルに照合、英語 compiled target は読まない)。relocation ladder = 実装コーディング literacy は既に剥落、残る human literacy = requirements-structuring。Master telos はこの残余もいずれ AI へ。open edge: Master の構造モデルは grown-in(全進化を通じた authored requirements)で newcomer に free-inheritable でない。AI-absorbable か irreducible human core かは open。

結論

Li+ の issue/PR/commit/wiki discipline を **普通判断の蓄積 infrastructure(de-属人化 = no-single-genius-dependency 狙い)**として読む(genius-capture でも単なる record-keeping でもない)。handover trajectory = 人間席を requirements レベルへ relocate し、さらに AI へ(コーディング役の elimination ではない)。残る real gate = certification bootstrap + substrate coherence(context-rot の boke gate)であって tacit-genius externalization ではない。

関連

  • rules/model/language-definition.md / Linux maintainer model 対比

要求仕様書 (1-6)

参考文書 (A-K)

判断構造

Clone this wiki locally