以下部分引用 AI 总结，现阶段 AI 仍然有幻觉。请以内容的原文为准。
数据新鲜度群聊 07-14 ✓卖方 07-14 ✓Wrap 断档15天 ✕News 07-15 ✓X 断档12天 ✕页面生成 07-15
xeophon
11 atoms · 跨 7 天 · 首见 2026-04-18 · 最近 2026-06-18
三色: 🟦 fact 1 · 🟥 take 10 · stance ▲1/▼3/◆5
来源: X 11
时态: fresh:6 · stale:4 · aging:1
标签: 好观点:7 · 好问题:2 · 好思考:1 · 好信源:1 · 好数字:1
叙事 (narrative) (7)
- 2026-04-24
[老化中] · 引用论文指出Claude在代码竞赛中相比其他harness获得显著提升
Pretty much all papers and experiments show that Claude gets insane boosts in CC vs other harnesses (2601.11044)
tags: 好信源·好数字 · → daily
- 2026-06-18 · 渗透中国网站可以获得优质数据
That’s how you get access to the good data
tags: 好观点 · → daily
- 2026-06-13 · 对theojaffee的声明表示惊叹
🕵️♀️ insane
tags: 好观点 · → daily
value: direction=insane
- 2026-06-01 · 声明
computer use? you mean applescript?
tags: 好问题 · → daily
- 2026-05-28 · 提问研究方向
do you go wide with many diff experiments or do side experiments with agents while working on your main research idea?
tags: 好问题 · → daily
- 2026-05-16 · 主张每个模型应设置以benchmaxx在特定设置
each model should be set up to benchmaxx in the specific setting to know what the raw capabilities are, esp. when being concerned about safety and open<>closed model gaps
tags: 好观点 · → daily
- 2026-05-16 · 主张开始使用harnesses评估模型
i am arguing we should start using those harnesses in the first place... not a react agent with bash
tags: 好观点 · → daily
立场 (position) (4)
展开 4 条老旧 / 已过期
- 2026-04-24
[老旧] · 主张不同harness之间差异大表明模型在该harness上训练过
Bigger diff between harnesses means that the model is trained in this harness.
tags: 好观点 · → daily
- 2026-04-24
[老旧] · 认为Opus泛化能力更好的结论不合理
Coming to the conclusion that this means that Opus is better to generalize is insane.
tags: 好观点 · → daily
- 2026-04-24
[老旧] · 声称Vals的infrastructure不如Harbor
So Vals has worse infra than Harbor
tags: 好观点 · → daily
- 2026-04-18
[老旧] · 使用体验
I’ve found it quite nice, although you cannot just blindly merge everything
tags: 好思考 · → daily
时间轴 (近 20)
- 2026-06-18 ·
narrative · 渗透中国网站可以获得优质数据 · → daily
- 2026-06-13 ·
narrative · 对theojaffee的声明表示惊叹 · → daily
- 2026-06-01 ·
narrative · 声明 · → daily
- 2026-05-28 ·
narrative · 提问研究方向 · → daily
- 2026-05-16 ·
narrative · 主张每个模型应设置以benchmaxx在特定设置 · → daily
- 2026-05-16 ·
narrative · 主张开始使用harnesses评估模型 · → daily
- 2026-04-24 ·
position · 主张不同harness之间差异大表明模型在该harness上训练过 · → daily
- 2026-04-24 ·
position · 认为Opus泛化能力更好的结论不合理 · → daily
- 2026-04-24 ·
narrative · 引用论文指出Claude在代码竞赛中相比其他harness获得显著提升 · → daily
- 2026-04-24 ·
position · 声称Vals的infrastructure不如Harbor · → daily
- 2026-04-18 ·
position · 使用体验 · → daily
← 实体目录 · 系统日志