System Over Model, Tested: Reproducing Mythos’s FreeBSD Find on Local Open-Weight Models¶
📊 Level ⭐⭐ | 2.7KB
相关实体¶
- unexpected lessons from an ai-assisted prototyping experimen → 原文存档
- ai gpus probably live longer than three years
- ddosing software delivery pipelines
深度分析¶
System Over Model, Tested: Reproducing Mythos’s FreeBSD Find on Local Open-Weight Models
核心观点¶
- A week later, Stanislav Fort at AISLE published a counter-thesis and reproduced the same find with `gpt-5.
- 4-nano
using their publishednano-analyzer` pipeline for under $100. - I wanted to see whether that reproduction works further down the cost curve.
- So I ran the pipeline at full sub-system scope (~50 files) using two open-weight models,
openai/gpt-oss-20bandgoogle/gemma-4-31b-it. - Out of the box it looked like both missed.
关联实体¶
- 存之有序治之有矩Agent 记忆系统的工程实践与演进
- 两万字详解Claude Code源码核心机制
- 你不知道的 Agent原理架构与工程实践 V2
- 龙虾装上了可以用来干啥分享下我的 Openclaw 多智能体团队搭建经验 V2
- Karpathy 最新访谈从 Vibe Coding 到 Agentic Engineering
- Karpathy Vibe Coding Agentic Engineering