2026作品集 / 100 件Collection / 100 pieces
GPT-6.2 Sol:100 件 HTML 视觉研究GPT-6.2 Sol: 100 HTML visual studies
同一模型、100 条题目产出的网页作品:互动 23、数据 27、内容 26、工具 14、模型演示 10。全部链回原发布页,未声明许可的完整文件不复制。One hundred web pieces from one model: interaction 23, data 27, content 26, tools 14, model demos 10. All link back to the publisher; unlicensed full files are not copied.
- 作品Pieces
- 100
- 用途Uses
- 互动 23 · 数据 27 · 内容 26 · 工具 14 · 模型 10
- 运行Run
- 外链 100
- 许可Licence
- 未声明 100
为什么值得看Why it matters这是一组批量样本,适合看同一模型在五种用途上的风格分布;不是模型能力排行榜,也不把单件作品当结论。A batch sample for seeing one model’s style distribution across five uses; not a capability leaderboard, and no single piece is a verdict.
边界 / 它没主张什么Boundary / what it does not claim同题不等于同 prompt;100 件也不等于 100 次独立评测。原仓库未声明 LICENSE,本站只外链。Same brief is not the same prompt, and 100 pieces are not 100 independent evaluations. The repository declares no licence; this site only links out.
来源Sources:原仓库Source repository
2026作品集 / 86 件Collection / 86 pieces
Claude Fable 5.2:86 件互动与视觉实验Claude Fable 5.2: 86 interaction and visual experiments
86 件作品里互动体验 57、数据可视化 24、内容与视觉 2、工具与界面 3;这一组偏交互,适合看同一模型如何处理状态、动画与输入。Of 86 pieces, 57 are interaction, 24 data visualisation, two content/visual and three tools/interfaces; the set leans interactive and shows how one model handles state, motion and input.
- 作品Pieces
- 86
- 用途Uses
- 互动 57 · 数据 24 · 内容 2 · 工具 3
- 运行Run
- 外链 86
- 许可Licence
- 未声明 86
为什么值得看Why it matters互动类占比最高,能对照「同一句 brief 下,哪些作品真的可操作,哪些只是看上去会动」。Interaction dominates, making it possible to separate pieces that are genuinely operable from those that only look animated.
边界 / 它没主张什么Boundary / what it does not claim「可操作」是本站的实际打开结果,不是作者自述;未声明许可的 86 件都不复制文件。“Operable” is this site’s own open result, not the author’s claim; none of the 86 unlicensed pieces are copied.
来源Sources:原仓库Source repository · 补充来源Supplementary source
2026同题对照 / 1 件Controlled pair / 1 piece
Claude Opus 5.5:B6 代码动画的单件样本Claude Opus 5.5: one B6 code-animation sample
同一份 B6 提示词的 Claude Opus 5.5 版本。和 DeepSeek V4.1-Flash 版并排看实现差异;原仓库明确说明这不是受控模型对比。The Claude Opus 5.5 take on the same B6 prompt. Compare it with the DeepSeek V4.1-Flash version; the source repository states this is not a controlled model comparison.
- 作品Pieces
- 1
- 用途Uses
- 内容与视觉 1
- 运行Run
- 站内 1
- 许可Licence
- free-share 1
为什么值得看Why it matters它的价值只在「同题并排」:同一句 prompt 下,两个模型的代码动画选择可以直接对照。Its value is the same-brief pair: under one prompt, the two models’ code-animation choices can be compared directly.
边界 / 它没主张什么Boundary / what it does not claim单件、单题、非受控;不能据此判断两个模型的整体能力,也不把并排当作评测。One piece, one brief, uncontrolled; it cannot establish overall model capability, and the pair is not an evaluation.
来源Sources:原仓库Source repository
2026同题对照 / 2 件Controlled pair / 2 pieces
DeepSeek V4.1-Flash:B6 与 X7 两件代码动画DeepSeek V4.1-Flash: B6 and X7 code animations
B6 与 Opus 共用同一份提示词;X7 只给一句话。两件都按 free-share 许可在站内直接运行。B6 shares its prompt with Opus; X7 came from a single line. Both run here under a free-share licence.
- 作品Pieces
- 2
- 用途Uses
- 内容与视觉 2
- 运行Run
- 站内 2
- 许可Licence
- free-share 2
为什么值得看Why it matters两件都可在站内运行,适合直接看 B6 的同题差异与 X7 的「一句话 brief」能走多远。Both run here, so the B6 difference and how far a one-line X7 brief goes can be inspected directly.
边界 / 它没主张什么Boundary / what it does not claimB6 的同题对照仍不是受控评测;X7 只有一件,不能代表模型的一般水平。The B6 pair is still not a controlled evaluation; X7 has one piece and cannot represent the model’s general level.
来源Sources:原仓库Source repository