hello hell

[hello hell @hellohellsc]

@elonmusk @TheAliceSmith I bet you’re missing Jack Ma deep down.

[hello hell @hellohellsc]

@0xTraderYi 自如被当成了抖加

[hello hell @hellohellsc]

@0xLogicrw autoresearch啊

[hello hell @hellohellsc]

@caiyue5 敢

[hello hell @hellohellsc]

@tualatrix $200 也一样🙂

[hello hell @hellohellsc]

@Lonely__MH gemini oauth限流非常严格

[hello hell @hellohellsc]

@sitinme 128g有教育？

[hello hell @hellohellsc]

opus 4.7 有些许叛逆

[hello hell @hellohellsc]

claude opus 4.7 web和app都有了

[hello hell @hellohellsc]

@luoleiorg 收到银行扣款短信了🥹

[hello hell @hellohellsc]

@jesselaunz 只要跑的时间够长就会要求login，昨天是在API 500错误之后出现的401

[hello hell @hellohellsc]

@ScottYeung @shenhuang 把降智说成重大升级，真是稳中向好

[hello hell @hellohellsc]

@lxfater 有没有可能出个满血版opus作为新模型…

[hello hell @hellohellsc]

TikTok-style algorithmic feeds basically lock your brain into a closed loop and converge you toward stupidity.

[hello hell @hellohellsc]

@GenAI_is_real I don’t think context engineering and harness engineering are the same thing. Harness engineering optimizes the agent’s objectives, while context engineering focuses on the LLM’s output.

[hello hell @hellohellsc]

@yucheng 我不觉得是一件事，harness针对的是优化问题空间，为agent目标服务，不是对模型做的

[hello hell @hellohellsc]

@cloudwu @realPureNomad 大模型也只是人类意识映射到文字后的再次映射，离全能神差远了

[hello hell @hellohellsc]

@tison1096 认同，可度量才是关键，只要能度量，问题最优解总是可通过迭代逼近的

[hello hell @hellohellsc]

@SaitoWu 我倒是非常同意这些观点，语言只是人类意识的映射，不能代表这个世界的通用规则

[hello hell @hellohellsc]

@runes_leo 通过垂类沉淀隐性领域知识，最后才能做到解通用问题