Shifan Zhao

1.1K posts

Shifan Zhao banner
Shifan Zhao

Shifan Zhao

@llm_kv

Build something useful. Built @lessmore_llm, clawSpark. Prev @MorganStanley

New York, NY Katılım Eylül 2019
359 Takip Edilen240 Takipçiler
Sabitlenmiş Tweet
Shifan Zhao
Shifan Zhao@llm_kv·
We’re happy to announce that @lessmore_llm is accepting external beta testing. Over the past two months, we’ve been building LessMore, a personal knowledge base agent. 2025 has been a very exciting year for agents. As agents become more capable, most problems can be solved given the right context. However, providing context goes against human nature, especially since you have to do it over and over again. With powerful agentic capabilities, persistent memory, and our differential summary, we believe the long-dreamed time to have an agent that distills information into knowledge, synthesizes knowledge into insights, and generates actions from insights is coming true. We are so excited about this idea. Thus, we built LessMore.
Shifan Zhao tweet mediaShifan Zhao tweet media
English
3
1
8
1.9K
Shifan Zhao
Shifan Zhao@llm_kv·
Kimi paused their subscription plans for new retail customers as expected
Shifan Zhao tweet media
English
0
0
0
35
（二狗子）Dogquant 小狗量化
（二狗子）Dogquant 小狗量化@Dogquant0·
币安高频做市盈利第二天，数学模型参数还在找甜饼区，等收集足够的数据📊就可以慢慢叠加资金上去狂奔了，跑的usdc 0free。
中文
5
2
8
1.2K
Shifan Zhao
Shifan Zhao@llm_kv·
@jasondeanlee I think gpt 5.6 sol pro do adaptive thinking, you probably can try gpt 5.6 sol ultra in work/codex. It's relentless in pursuing goals.
English
0
0
1
207
Jason Lee
Jason Lee@jasondeanlee·
My chatgpt 5.6 sol pro queries are coming back instantly. This means I am getting routed to 5.5 instant right?
Kensington, CA 🇺🇸 English
20
0
76
15.7K
Shifan Zhao
Shifan Zhao@llm_kv·
@thsottiaux It’s really slow each time I open my phone to connect to codex. It would be super nice if you can make it faster. Also the pet is great, wonder if it’s possible to have the pet on phone
English
0
0
0
302
Tibo
Tibo@thsottiaux·
How can we make ChatGPT better. It’s already on your phone, but what do you wish would happen more or less of when you open the app. How can it bring you more joy?
English
4.3K
120
4.8K
548.9K
Thariq
Thariq@trq212·
This was due to a heroic effort by many people at Anthropic working sometimes literally around the clock It was not at all clear that we'd be able to do this in time, and so proud of everyone who made it happen. Enjoy Fable.
Claude@claudeai

Beginning July 20, Claude Fable 5 will be included in all Max and Team Premium plans, at 50% of limits. Pro and Team Standard users will continue to have access to Fable via usage credits, and will receive a one-time $100 credit. Demand for Fable has been challenging to predict, which is why we rolled it out to subscription plans in stages, extending access several times as we secured additional capacity.

English
1K
184
7.9K
1M
Shifan Zhao
Shifan Zhao@llm_kv·
@Dan4444L 没办法，算力跟不上，资金跟不上，用户只能靠真爱充值，但是要留住用户还是难。但是为了支持kimi的开源精神，我还是充了一年
中文
1
0
2
1.7K
有的没的
有的没的@Dan4444L·
我一直都很喜欢 Kimi，也是 Kimi 的粉丝。可是 Kimi 的模型想要正经开始用，必须得订阅 199 的套餐。这个 199 的套餐比 ChatGPT Plus 的基础套餐和 Claude 的基础套餐都要贵，我实在是有点找不到理由，我为什么非要订这个 199 的套餐。
中文
133
3
325
132.6K
Shifan Zhao retweetledi
Yifan Zhang
Yifan Zhang@yifanzhang_·
Heard that some frontier models are basically a 48-layer transformer looped twice (48L x 2). Now we are introducing DeepLoop: Depth Scaling for Looped Transformers (arxiv.org/abs/2607.13491), making the loop transformer stable and scalable!
Yifan Zhang tweet mediaYifan Zhang tweet media
English
18
125
1.2K
341.5K
Shifan Zhao
Shifan Zhao@llm_kv·
@miantiao_me 好奇用gemini来做啥呢，干啥啥都不行啊，白充了一年会员🤣
中文
0
0
2
635
面条
面条@miantiao_me·
谷大善人真的耐薅
面条 tweet media面条 tweet media
日本語
12
0
50
31.3K
Shifan Zhao
Shifan Zhao@llm_kv·
@elonmusk Hey Elon, noticing Grok Heavy seems to default to a 4-agent team. Is the full 16-agent scaling still available on SuperGrok Heavy for complex tasks, or has that changed from the original branding? Could you add highter thinking efforts like claude max or gpt pro/agent swarm like kimi? Currently heavy users have basically no control of the thinking efforts. I have two accounts, super grok and grok heavy they basically have same thinking efforts, I have no way to get better answer from heavy even I pick the heavy mode.
English
0
0
2
68
Woooooow
Woooooow@woooooow_w·
@ShanghaoJin @BabybusFL 迅速去杠杆后，动量资金是不是可能来第二次对半导体扫货，市场找不到更好叙事的题材。
中文
2
1
11
7.4K
Babybus
Babybus@BabybusFL·
不用金老师睡醒了回答你。我回答你就可以了 主要是你太嫩了 没见过世面。sk海力士 三星 美光 都是泡沫 泡沫的点不在于你一直强调的周期 和 eps pe ps这些。存储的泡沫在于股权结构。所有的股权结构都是杠杠买入的，存储公司的股票都是Leverage Bubble。Leverage Bubble泡沫破裂的一个特征就是margin call。就是现在这个情况。属于无论你如何呐喊也没用 你不是很懂金融 今天一次给你上一课 希望你可以铭记终身。多米诺骨牌一定会塌的 最后一定是平的。
Pike时政财经分析@BigbirdflyChan

金老师认为存储PE倍数很低，存在泡沫。 泡沫是股价长期脱离企业未来现金流和盈利能力，而不是PE高低。例如2000年的互联网泡沫，很多公司PE几十倍、几百倍，甚至没有盈利，这当然是泡沫。 2026年年的美光，PE只有5倍左右，而且2027年pe倍数可能更低。 一个未来可能创造近800b现金流的企业估值给了900b，我真看不出来哪里来的泡沫。

中文
67
21
205
111.5K
Shifan Zhao
Shifan Zhao@llm_kv·
@interjc 直接送reset成本更高了,9m用户重置一次要烧不少钱
中文
0
0
1
649
Justin
Justin@interjc·
我其实不太建议 codex 动不动就重置额度，因为你重置的时候也许我都还没开始用或者刚刚重置过 应该每次想重置就送一次重置次数，这样皆大欢喜
中文
98
2
299
66.6K
Xinyu Yang
Xinyu Yang@Xinyu2ML·
Why can Kimi ship K3? Let me tell my story. Earlier this year, I left academia for industry. I talked to a lot of companies along the way. Here's what I saw: 1⃣Arrogance. They believe the AI war is over, and they won. No hunger for the future, and no hunger for talent. 2⃣Restlessness. Young labs short on foundation, either rushing to catch the frontier or pivoting away from the competition. 3⃣Fear. Strong teams with real experience, but from the second tier, they can't quite bring themselves to aim for #1. 4⃣Misalignment. Everyone is optimizing for their own credit, but nobody really cares whether the company can reach AGI. Kimi was different. Over many conversations with the founders, the same thing came through every time: a raw, genuine hunger for AGI. I joined. The hunger was real. We shipped K3. This is only the beginning.
Xinyu Yang tweet media
English
80
195
1.6K
714.4K

Keşfet

@Dogquant0 @jasondeanlee @thsottiaux @trq212 @Dan4444L @miantiao_me @elonmusk @BarackObama