Everything is the model, like that’s the thing that wins in the end is when it’s really the model has the ability to reason over context and talk about it in a way you find helpful.
—Boris Cherny (@bcherny) Creator of Claude Code
Everything is the model, like that’s the thing that wins in the end is when it’s really the model has the ability to reason over context and talk about it in a way you find helpful.
—Boris Cherny (@bcherny) Creator of Claude Code
近期工具开发的感悟
We suffer more often in imagination than in reality.
— Seneca in Epistulae Morales ad Lucilium
工程上让我着迷的两种范式
自己能够改善自己 要有平台属性,如使用 C 编译器优化自己(编译器)、使用 Linux 开发 Linux。
全局能力强大,局部简单 简单的规则进行巨量的重复,如 GPT-like 模型、集成电路。
当有一个好的环境,可以聚焦于解决问题时,这会激发纯粹的解题乐趣。 感觉很微妙,非常纯粹,非常充实。
There’s lots of ways to be, as a person.
And some people express their deep appreciation in different ways.
But one of the ways that I believe people express their appreciation to the rest of humanity is to make something wonderful and put it out there.
— Steve, 2007
它会让你:
LLM 拥有近似无限的能力,本质上也是通过「指令微调」的方式「教会」了模型。 算是一种 scaling law,对于新技能的 scaling law。对于新增一个领域的业务,扩展他的成本极其低。 甚至是自动化的事情。
LLM 区别于其他 AI 技术点在于,它的 scaling 能力太强大了。 模型本身的「智力」、模型解决特定领域问题的能力等等。
它的训练方式(模型+数据)以及与世界的交互方式(文字),奠定了它的能力天花板非常非常高。
富士山的枫叶