Curline Zephirin (@Curline1222) profile picture

LLM enthusiast

Joined January 2025
Photo by Curline Zephirin (@Curline1222)Photo by Curline Zephirin (@Curline1222)Photo by Curline Zephirin (@Curline1222)Photo by Curline Zephirin (@Curline1222)Photo by Curline Zephirin (@Curline1222)Photo by Curline Zephirin (@Curline1222)Photo by Curline Zephirin (@Curline1222)Photo by Curline Zephirin (@Curline1222)Photo by Curline Zephirin (@Curline1222)Photo by Curline Zephirin (@Curline1222)
Replying to @Vincent_AINotes
Hitting the gym, brb on Tuesday
3
5
1
433
31,207
Curline Zephirin retweeted
Well I guess the distillation process for the new Gemini models already started and they're paying premium for it, the demand increase for Gemini is very clear IF THIS IS TRUE, maybe @grok can confirm and translate this. Wouldn't be surprised Neon and Carbon are worth distilling.
15
3
56
10,550
Today: Yesterday
When will the Riemann Hypothesis be resolved? November 2021: 2057 November 2023: 2053 November 2025: 2044 November 2026: 2035 metaculus.com/questions/6494…
42
GPT-6.2 Sol Ultrafast is basically the new GPT-6.1 Terra Fast, but at 12x the usage cost.
Need speed when building with GPT-6.2 Astra or Sol? On ChatGPT Pro 200, you can use Ultrafast with Sol. Here’s how @dkundel uses Ultrafast to rapidly iterate and how he decides when the extra speed is worth it:
39
If you're in math or physics, pivot to political science or political marketing.
If you're in maths, pivot to physics.
3
227
Curline Zephirin retweeted
Anthropic 从来没有把「反华」当作公司叙事。我看到所有把 Anthropic 和反华放在一起的,都是中文内容,都是中国的叙事。 Dario 的意识形态和逻辑思路向来清晰,他在地上画了三个圈: 第一个圈,非民主国家。一方面,民主国家和非民主国家的竞争会延续,他站边美国利益就必须反对非民主国家。另一方面,独裁政府获得更强大 AI 会对本国人民实施更强大的压制。 第二个圈,蒸馏。 第三个圈,缺乏安全机制的前沿模型。 画第一个圈时,其实俄罗斯、朝鲜、伊朗等国家都被画进去了。但画完三个圈之后,有资格留下的就只有一个国家了。
unavailable
114
35
12
642
65,136
Curline Zephirin retweeted
Mathematicians did not ask for this work to be done.
JUST IN: Riemann Hypothesis has been solved. alpo.ge/remark.pdf
2
3
107
2,495
Curline Zephirin retweeted
Distillation can reproduce powerful AI capabilities without the safeguards built around them. We’re sharing new research on our work to combat distillation of our models, how these capabilities transfer, and how we detect and disrupt coordinated attempts. These findings are helping us sharpen our defenses and support safeguards across the industry. openai.com/index/combatting-…
752
297
453
3,801
1,148,269
In 1993, mathematician Ronald Graham imagined a computer answering the Riemann Hypothesis: “Yes, it is true, but you won’t be able to understand the proof.” 33 years later, Claude and Levent Alpöge claim to have proved it, with only a massive Lean formalization released so far. What a prediction.
4
1,289
RIP Greg
I bet @YafahEdelman at 5:1 that no Millennium Prize problem would be solved within a year (i.e., by 2027-08-12), regardless of AI involvement. I'm tapped out on this one, but arrange your own bets in the replies!
25
Curline Zephirin retweeted
astra 6.2 so bad that openai staff literally had to lie to promote their models
5
1
1
79
41,353
even $1 got blocked
Fun experiment for the week! Text Poke, pitch your idea and earn a Devin Bot to help you build it. 10,000 accounts available, each up to $10,000 (sadly Poke doesn’t think I need the credits) poke.com/bots
40
That's amazing!!!!
Meet Poki: a 100% free, open-source native iOS phone-use agent. No running Mac. No jailbreak. No external hardware. Just bring your own ChatGPT subscription or API. Summoned via the Action Button, Poki controls your iPhone just like you—hailing rides, ordering food, booking flights, and beyond. Most personal agents stay trapped in a sandbox with their hands tied. Poki actually executes everything you need on mobile—including complex, long-horizon workflows running continuously for up to 30 minutes. And of course, Poki always asks for your permission before any sensitive action. Enjoy 👇
1
45
BIG IF TRUE
👀 Kath Korevec (@simpsoka) has left OpenAI.
1
53
Curline Zephirin retweeted
🚨 Gemini users !! Google just launched Computer Use in the Gemini app for Mac — and it does NOT hijack your mouse. Gemini runs clicks, typing, local files, and desktop apps inside an isolated background window with a live preview in chat. Your main screen and cursor stay free while it works. blog.google/innovation-and-a…
24
6
4
86
5,784
Curline Zephirin retweeted
1,996
9,884
3,259
118,480
12,500,643
Curline Zephirin retweeted
GEWINNSPIEL 2.0!!! 🔥 Zum Anlass dieses großen und besonderen Tages, hauen wir einfach noch mal ein paar fette Gewinne raus! 😍 Zu gewinnen: 10x 100€ PSN Guthaben 1x PS5 Pro Das müsst ihr tun: - Uns folgen - Beitrag liken und teilen - Kommentieren Teilnahmeschluss: 23.11.26, 13:30 Teilnahmebedingungen: magentasport.de/files/AGBsGe…
GEWINNSPIEL!! 🔥 Es kommt ja bald das ein oder andere interessante Game raus. Darum haben wir uns was für euch überlegt! Zu gewinnen: 10x 100€ PSN Guthaben 1x PS5 Pro Das müsst ihr tun: - Uns folgen - Beitrag liken und teilen - Kommentieren Teilnahmeschluss: 17.11.26, 23:59

9,345
8,105
145
10,444
490,620
Bro, do you remember what you posted yesterday 💀💀💀
🚨 Gemini 4 Xenon evals leaked 🔥 Neon just drop yesterday and this demon already sitting at 71 AA Index 🤯 > AA Index : 71 (Neon 61 , Fable 5.2 65) > HLE-Diamond : 86.2% 💀 > Terminal-Bench : 74.8% real flagship above Carbon , as always with proof 👀
1
3
48
Curline Zephirin retweeted
Anthropic researchers believe they're getting a bigger speedup from their Claude models than OpenAI researchers believe they're getting from their GPT models.
We're sharing new data on how AI is changing research at OpenAI. We find researchers' AI use continues to rise exponentially, and agents are completing longer tasks, but human involvement remains central in research. Researchers are also spending more time on safety and alignment, with human effort in these areas growing faster than in any other individual research area. This report aims to increase transparency around research acceleration and inform the governance of increasingly capable AI. We believe that AI labs should be required to track and publicly report indicators of recursive self-improvement (RSI): the process by which AI accelerates the development of future AI systems. Our hope is that this expanded report will help sustain and advance shared measurement norms across the industry. openai.com/index/research-ac…
11
18
4
309
31,128
Curline Zephirin retweeted
We’ve priced our IPO at $83 per share, with trading expected to begin on Nasdaq tomorrow. anthropic.com/news/anthropic… Legal disclaimer: A registration statement relating to the Class A common stock was declared effective by the Securities and Exchange Commission on November 18, 2026. The offering is being made only by means of a prospectus. Copies of the final prospectus may be obtained, when available, from: Goldman Sachs & Co. LLC, Attention: Prospectus Department, 200 West Street, New York, NY 10282, by telephone at (866) 471-2526, or by email at [email protected]; or Morgan Stanley & Co. LLC, Attention: Prospectus Department, 180 Varick Street, 2nd Floor, New York, NY 10014, or by email at [email protected]
1,304
2,190
1,385
16,952
3,412,131
Curline Zephirin retweeted
GPT-6.2 Astra is the highest scoring model on RoboDojo-RC Tier 1, a benchmark of real robot tasks. It beats Dream-1, the best robotics model we've tested.
30
42
18
542
63,008