Ethan Mollick指出，关于权重空间编程的说法意义重大，但仍待复现

发生了什么

在等待复现，不过这确实是件大事。 引用 @rikiyatakehi 的帖子：我们让 GPT-6 Astra 编辑 Qwen3.5-0.8B 的权重，以提升数学能力。它直接改动的权重仅占 0.016%，修复了 0.8B 的推理循环问题，使数学能力最高提升 6 倍，指令遵循能力甚至提升 3 倍，而且没有使用梯度或进行训练。 我们称之为权重空间编程：前沿大语言模型现在可以原生读写神经网络权重！🧵 博客：https://rikiyat.github.io/blog/weight-space-programming/

接下来读什么

• 今日还有哪些重要变化通过五项核心价值观解读 OpenAI 更新后的 Model Spec

这件事到目前为止

Ethan Mollick called a claim about editing model weights a big deal, while saying it still needs replication. The claim, posted by Rikiya Takehi, says GPT-6 Astra directly changed 0.016% of Qwen3.5-0.8B’s weights without gradients or training. The post says the edits fixed reasoning loops, improved math performance by up to 6x, and increased instruction following by 3x. Takehi calls the approach “Weight-Space Programming” and says frontier language models can read and write neural network weights. These results and broader claims are reported by Takehi; Mollick has not described them as independently confirmed.