算法艺术创作
algorithmic-art
通过代码和可调参数创作生成艺术与交互式视觉效果。
argumentation【设计与创意】Construct well-structured arguments using the hypothesis-argument-example triad. Covers formulating falsifiable hypotheses, building logical arguments (deductive, inductive, analogical, evidential), providing concrete examples, and steelmanning counterarguments. Use when writing or reviewing PR descriptions that propose technical changes, justifying …
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
由設經辯至實證建嚴辯。技稱皆循三:明設述何信、辯釋何以、例證之確。
述稱為明、可駁設。設非見或好——乃可對證測之具體斷。
可駁與不可駁:
| Unfalsifiable (opinion) | Falsifiable (hypothesis) |
|---|---|
| "This code is bad" | "This function has O(n^2) complexity where O(n) is achievable" |
| "We should use TypeScript" | "TypeScript's type system will catch the class of null-reference bugs that caused 4 of our last 6 production incidents" |
| "The API design is cleaner" | "Replacing the 5 endpoint variants with a single parameterized endpoint reduces the public API surface by 60%" |
| "This research approach is better" | "Method A achieves higher precision than Method B on dataset X at the 95% confidence level" |
得:一句設具體、定範、可駁。讀者可即想何證可確或駁。
敗:設似糊→施「吾何駁此」測。不能想反證→稱為見非設。窄範或加可量準至可測。
擇最支設之邏構。異稱呼異推法。
| Type | Structure | Best for |
|---|---|---|
| Deductive | If A then B; A is true; therefore B | Formal proofs, type safety claims |
| Inductive | Observed pattern across N cases; therefore likely in general | Performance data, test results |
| Analogical | X is similar to Y in relevant ways; Y has property P; therefore X likely has P | Design decisions, technology choices |
| Evidential | Evidence E is more likely under hypothesis H1 than H2; therefore H1 is supported | Research findings, A/B test results |
配設於最強辯類:
考合類為強辯(如類比輔以歸納證)
得:擇辯類(或合)附明由何配設。
敗:無單類合→設或須分子稱。析為部,各有自然辯構。
築接設於證之邏鏈。
例——碼察(演+納):
設:「取驗邏為共模將減三 API 處之蟲重」
前提:
- 三處(
createUser、updateUser、deleteUser)各施同入驗微異(觀於src/handlers/)- 末 6 月 5 驗蟲中 3 修一處而未傳他(見 #42、#57、#61)
- 共模行邏一源(演:一施則一處可修)
邏鏈:因三處重同驗(提一),一修而他漏(提二,自 3/5 例納)。共模意一修達諸呼者(共模語演)。故取將減蟲重。
對辯(鋼化):「共模引耦——驗變影一處可破他」
駁:諸處已共同驗意;耦隱難維。經共模附參選(如
validate(input, { requireEmail: true }))使耦明可測。今隱重更險,蓋藏依。
例——研(證據):
設:「域語料前訓較增通語料更增生醫 NER 下游性能」
前提:
- PubMed 前訓 BioBERT(4.5B 詞)勝通英前訓 BERT-Large(16B 詞)於 6/6 生醫 NER 基(Lee 等,2020)
- Semantic Scholar 前訓 SciBERT(3.1B 詞)勝 BERT-Base 於 SciERC、JNLPBA 雖前訓料小
- 通域擴(BERT-Base 至 BERT-Large、3x 參)於生醫 NER 益小於域適(BERT-Base 至 BioBERT、同參)
邏鏈:證一致示生醫 NER 域料擇勝料規(證據:此果於域特較規重時更可)。三獨比同向,強納。
對辯(鋼化):「此或不普於生醫 NER 外——生醫獨技詞膨域適優」
駁:有效限。設專為生醫 NER。然似域適益見於法 NLP(Legal-BERT)與金 NLP(FinBERT),示紋或普於他特域,雖此分稱需自證。
得:完辯鏈附前提、邏接、鋼對辯、駁。讀者可步循。
敗:辯似弱→察前提。弱辯多由不證前提非邏誤。為各前提尋證或認為設。對辯強於駁→設或須改。
以獨可驗證輔辯。例非繪——乃使辯可測之經基。
例擇準:
| Criterion | Good example | Bad example |
|---|---|---|
| Independently verifiable | "Issue #42 shows the bug was fixed in handler A but not B" | "We've seen this kind of bug before" |
| Specific | "createUser at line 47 re-implements the same regex as updateUser at line 23" | "There's duplication in the codebase" |
| Representative | "3 of 5 validation bugs in the last 6 months followed this pattern" | "I once saw a bug like this" |
| Includes edge cases | "This pattern holds for string inputs but not for file upload validation, which has handler-specific constraints" | (no limitations mentioned) |
得:讀者可獨驗之具例。至少正一邊一。各引具產(檔、行、議、論、集)。
敗:例難尋→設或太廣或非觀現實接地。窄範至可指。例缺乃信號非以糊引補之缺。
合設、辯、例為脈適式。
碼察——構註:
[S] <one-line summary of the suggestion>
**Hypothesis**: <what you believe should change and why>
**Argument**: <the logical case, with premises>
**Evidence**: <specific files, lines, issues, or metrics>
**Suggestion**: <concrete code change or approach>
PR 述——構體:
## Why
<Hypothesis: what problem this solves and the specific improvement claim>
## Approach
<Argument: why this approach was chosen over alternatives>
## Evidence
<Examples: benchmarks, bug references, before/after comparisons>
ADR——用標 ADR 式三映於脈(設)、決(辯)、果(例/預證)
研書——映標構:引述設、法/果供辯與例、論對對辯
察合辯:
得:完、格之辯適於脈。讀者可評設、循推、察證、考對辯——皆於一連構。
敗:合辯似散→設或太廣。析為焦子辯,各自有設辯例三。兩緊辯勝一蔓辯。
review-pull-request —— 施構辯於構碼察饋review-research —— 構基證辯於研脈review-software-architecture —— 證構決以設辯例三create-skill —— 技本身為構辯如何成事write-claude-md —— 文益於明證之例與決高險決,合此技與 advocatus-diaboli 為決前察環。紋:
何時合對獨用:
例——PR 應精: 構辯構應(設:合 PR 善、附證辯、合作邀)。advocatus-diaboli 後捕兩關症:代過程識稱為揣非實(安 PR 致愧)、「吾實踐已測之」不可驗。皆去。末應短 40-50%——過釋示不安。
例——系設分流: 構辯(經 Plan 代)設全 500 行分流管。advocatus-diaboli 殺之:9 項時系太早自為維負(遞陷)。末解:25 行加既腳。