Simon Willison(@simonw)

I think loops were a short-lived patch for models that couldn't reliably keep working on long proble...

6.5内容质量
I think loops were a short-lived patch for models that couldn't reliably keep working on long proble...

TL;DR · AI 摘要

当前大模型已能无需循环处理长期任务，GPT-5.6等模型表现突出。

核心要点

  • GPT-5.6和Kimi K3可直接处理长期任务，无需人工循环干预
  • Simon Willison认为循环是临时解决方案
  • Wes McKinney公开批评循环技术为'bullshit

结构提纲

按章节快速跳转。

  1. §引言

    指出循环机制是模型处理长期任务的临时解决方案

  2. ·模型能力对比

    Fable和GPT-5.6等新型模型已具备原生处理长期任务能力

  3. ›专家观点

    Simon Willison和Wes McKinney对循环技术的否定态度

  4. ·技术趋势

    大模型正在摆脱对人工循环干预的依赖

思维导图

用一张图看清主题之间的关系。

查看大纲文本（无障碍 / 无 JS 友好）
  • 大模型处理长期任务能力
    • 技术演进
      • 循环机制（临时方案）
      • 原生处理能力（GPT-5.6等）
    • 专家观点
      • Simon Willison否定循环
      • Wes McKinney批评循环

金句 / Highlights

值得收藏与分享的关键句。

#大模型#AI#GPT#技术趋势
打开原文

Simon Willison on X: "I think loops were a short-lived patch for models that couldn't reliably keep working on long problems until they hit a defined goal Fable and GPT-5.6 (and probably Kimi K3 as well) can just do that out of the box, unassisted" / X

Simon Willison

@simonw

I think loops were a short-lived patch for models that couldn't reliably keep working on long problems until they hit a defined goal Fable and GPT-5.6 (and probably Kimi K3 as well) can just do that out of the box, unassisted

Wes McKinney

@wesmckinn

Jul 22

I think loops are bullshit

1:29 AM · Jul 23, 2026

152.9K

Views

139

59

1.3K

216