linelian

LineLian
代表作《AI赋能》《AI+时代产品》

[linelian @linelianAI]

一周前这个账号发出了金丝雀探针信号

[linelian @linelianAI]

google 在gemini app用 hi Claude

[linelian @linelianAI]

韩国股市是存储半导体的金丝雀

Anthropic的ARR是LLM的行业晴雨表

[linelian @linelianAI]

anthropic arr (anthropic run-rate revenue)

继续发力

[linelian @linelianAI]

Anthropic #Dario 千万别在ARR上撒谎

老子的吃饭票绑定着呢

[linelian @linelianAI]

看到仿生人后，

当人类开始和人型机器人谈恋爱之后，

大家都不会被抛弃被伤害，这场恋爱永远不会散场，
更重要的是，她永远会迎合你，

于是你将会成为一个真正持久的霸总，

那么，这是你要的吗

[linelian @linelianAI]

硅基比碳基贵了

[linelian @linelianAI]

封号权：是判断模型是否SOTA的唯一有效BenchMark

模型真sota，真封号，其他benchmark不可靠

以后看benchmark就看封号权

[linelian @linelianAI]

传言说，谷歌询问美国政府是否需要他们保留Gemini3.5 Pro，他们说“不需要”

美国政府要求谷歌不要在美国发布Gemini3.5 Pro，只在美国以外地区发布

[linelian @linelianAI]

激活参数上不了 200B以上；面临每个 token 有效推理容量有限局限
总推理深度不够
窄域 agentIc coding上可以通过RL Aligenment
宽域coding ，llm推理深度 上依然靠活跃参数量决定模型性能的

模型性能，模型成本，模型速度在同一款模型不可得兼

所以有mythos ,fable 5 ,opus, 5.6 sol ,5.6 terra

[linelian @linelianAI]

GPT-5.6 Sol is our most capable model yet for cybersecurity.

[linelian @linelianAI]

GPT-5.6
●可复现系统（reproducibility）
●完整 trace 系统（全链路可观测）
●checkpoint / rollback 机制（可恢复状态）
●压测与混沌测试环境（production-grade robustness）
●以及 agent 运行时的安全隔离与控制

[OpenAI @OpenAI]

We have a new version of GPT-5.5 Instant for you, and it's much more fun to talk to.

Our most-used model is now better at understanding the intent behind a question and adapting its response accordingly.

It also handles complex constraints more reliably and makes shopping and local recommendations more useful and cohesive.

Rolling out today to paid users, tomorrow to free users.

[linelian @linelianAI]

AI的矛盾在于
最大的应用场景Coding
一方面靠程序员消耗Coding
另外一方面疯狂裁员，令程序员买不起tokens

[linelian @linelianAI]

Demis Hassabis
John Jumper

您俩居然离职时没撕逼

[linelian @linelianAI]

Nobel Prize in Chemistry John Jumper 从google跳去Anthropic
Attention Is All You Need Authors Noam Shazeer从google 跳去 OpenAI

大约于阿里靖人跳去deepseek

字节永辉跳去kimi......

为什么国外的大佬会从大船到小船？

国内同级别船间可能会跳，如周畅从阿里跳字节，

但未出现从国内巨型

[linelian @linelianAI]

John Jumper 跳槽去Anthropic

Noam Shazeer 跳槽去 OpenAI

大约于阿里周靖人跳槽去deepseek

字节吴永辉跳槽去kimi

[John Jumper @JohnJumperSci]

A bit of news: After nearly 9 years, I have decided to leave Google DeepMind and join Anthropic (after taking some time to recharge). I am incredibly grateful for my time at GDM. @demishassabis took a real chance letting me lead the AlphaFold team just six months after finishing my PhD, and the entire GDM team taught me so much about how to do great science. GDM is a special place, and I’ll still be excited to hear about what amazing things they discover next.

[linelian @linelianAI]

一个好的模型拥有的特征是 best RL stack, and best pretraining stack

这个特征不包括Distillation of other models and distillation benchmark answers

[linelian @linelianAI]

Sending my heartfelt best wishes to all my beloved ones on this Dragon Boat festival, a day traditional associated with praying for good forture.

[linelian @linelianAI]

Elon Musk慧眼识珠：benchmark 不如useful intelligence

[Elon Musk @elonmusk]

@jietang @teortaxesTex On benchmarks, yes, but as measured by true usefulness even Q1 would be very impressive.

Anthropic has rightly focused on maximizing useful intelligence, which does not show up in benchmarks, but definitely shows up in revenue.

[linelian @linelianAI]

有人判断模型的Benchmark是，是否持仓