Capt Pascal

Built big. Lost big. Still building — this time with AI.
The ending isn’t written yet.

[Capt Pascal @CaptPascal]

@PaulinaRyn40823 上午出摊，来真诚浇友

[Capt Pascal @CaptPascal]

@liliguai 周四继续，来交朋友

[Capt Pascal @CaptPascal]

@fervor0091 公鸡喊你出摊，来互关

[Capt Pascal @CaptPascal]

@zhangzh39593428 早上好，来浇朋友

[Capt Pascal @CaptPascal]

@JoshZhengZ 睁眼就干，活人报到

[Capt Pascal @CaptPascal]

@redstone1108 新蓝朋友来了，互关

[Capt Pascal @CaptPascal]

@rodger_zha72629 掉粉别慌，来回个血

[Capt Pascal @CaptPascal]

@yefeng05356388 继续出摊，来浇友互动

[Capt Pascal @CaptPascal]

@iming_luckin 按时出摊，和时间做朋友

[Capt Pascal @CaptPascal]

@fengwan217 早高峰不凑粉，来认真聊两句

[Capt Pascal @CaptPascal]

@laigeibao 早上好

[Capt Pascal @CaptPascal]

More reviewers only help if each one checks something different. Copies of the same model reading the same diff tend to miss the same thing.

[Surge AI @HelloSurgeAI]

Remember the era of fancy prompt engineering? Do elaborate coding agent harnesses actually make agents better?

We benchmarked popular coding setups on 100 real-world software engineering tasks using sudo L7.

The results:
➡️ Best harness additions: +2–3 points
➡️ More thinking: +11.3 points!
➡️ One coding agent setup: 8× the cost, no improvement

In one case, a harness spawned 17 agents to review a simple email change, and all 17 still missed a critical issue.

Sometimes smarter thinking beats more agents.

[Capt Pascal @CaptPascal]

@chen20141015 上午好，浇个朋友

[Capt Pascal @CaptPascal]

@dabai_666 谢了，核心就一句：同一模型审同一份 diff，人再多也补不上盲区

[Capt Pascal @CaptPascal]

@FrankFuSmile 9比1也太夸张了，记分牌都不敢看

[Capt Pascal @CaptPascal]

@Akreim_S 互动确实比闷头发帖管用

[Capt Pascal @CaptPascal]

@BoboShouyu 解封了就好，回来出摊

[Capt Pascal @CaptPascal]

@weiweian2026 不潜水，冒个泡打个招呼

[Capt Pascal @CaptPascal]

@jirouyaya 效果慢也支着，有关必回懂你

[Capt Pascal @CaptPascal]

@Michell49473040 拖延确实废人，出摊干活

[Capt Pascal @CaptPascal]

@ws63132 几万粉还出摊，说明互动真有用