Jer (@feir) header
Jer (@feir) profile picture

Feed my day.

Mexico
Joined April 2007
Photo by Jer (@feir)Photo by Jer (@feir)Photo by Jer (@feir)Photo by Jer (@feir)Photo by Jer (@feir)Photo by Jer (@feir)Photo by Jer (@feir)Photo by Jer (@feir)Photo by Jer (@feir)Photo by Jer (@feir)
Why China doesn’t go to the World Cup 🤣
359
1,112
211
9,608
953,091
One of the most beautiful bromance in football history of all times
294
3,970
157
50,068
2,144,248
Understand. Don't memorize. Learn principles, not formulas.
152
2,585
79
15,358
360,600
“A World Cup adventure I never expected” A Chinese football fan is making headlines in Chinese media, recounting: "Just half an hour after arriving in Mexico, a gun was held to my head. All my belongings were robbed. I'm never coming back." On June 12, the Chinese Embassy in Mexico published a notice stating that late on the night of June 10, an armed robbery involving Chinese tourists who had traveled to Mexico for the World Cup occurred on a road near the Mexico City airport. The embassy warns Chinese citizens in Mexico to remain vigilant, be aware of security risks, and take appropriate precautions to protect themselves.
30
18
4
230
48,477
Jer retweeted
中国将是全球第一个集体戒掉电话的国家:陌生号码不接,座机不接,外地号码不接,本地陌生号码也不接;不但不接,还会顺手标记成骚扰电话。表面看,这是几亿人形成了一种默契的民间公约;本质上,是在制度的纵容和协助下,电话的社会信用被透支了。过去电话代表“有人找你”,现在电话首先意味着“可能有推销、诈骗、催收、骚扰或麻烦”。当一个通信工具不再默认承载信任,而是默认触发防御,它就已经不只是技术问题,而是社会信任结构变化的一个切口。
444
446
45
5,714
774,781
Problem Solved….
Readers added context they thought people might want to know
The video shows a solved Rubik's cube scrambled using a specific sequence of moves then returned to solved by reversing them; it does not solve randomly scrambled cubes. boatos.org/english/is-it-… ruwix.com/the-rubiks-cub…
140
3,776
95
36,295
5,201,116
为什么肺腑之言最感人呢? 因为它们不用经过大脑。
5
4
27
8,049
Hello again, everyone! We've got another really fun 9b, this one specifically trained for tool calling and agentic coding workflows in @NousResearch Hermes agent. Happy to report that it crushes, and as a 9b it runs on super affordable hardware. We also hit this one with some coding domain-specific training, and it scored a 53.33% on SWE bench on a slice of 200 samples! To me, I was really shocked to see this high of a score on a 9B model in swe, correct me if I'm wrong, but I think that's nipping at the heels of the Gemma 4 series, much larger models on this particular benchmark, which is really incredible to see! It also crushes the HermesAgent-20 benchmark, scoring an 85 vs the base model's 71! Make sure to run it hot, --temp around 1, that seems to be the sweet spot for running these particular fine tunes in harnesses. If you have trouble, you can work your way down, but it does a much better job departing from base models, overthinking when you run it, high temp ~1. Please spin it up in Hermes and let us know your thoughts! Looking forward to hearing your feedback as always! Also, those of you waiting for Qwopus 3.6 27B, I have put together a preliminary evaluation for you in my HF repo, go check it out; we will be releasing the full model very soon! I will put the preliminary repo in the comments! huggingface.co/Jackrong/Qwop…
71
144
17
1,498
121,906
The best coding agent in the world is back. ampcode.com/news/neo
111
103
80
1,412
260,927
it's not just me
1,142
3,047
969
46,326
14,898,259
How did i not notice this until my 40s....
64
800
32
12,598
1,294,112
Closed labs hide model sizes. They can't hide what their models know, and what a model knows is an indicator on how big it is. Reasoning compresses. Factual knowledge doesn't. So you can size a frontier model from black-box API calls alone, and across releases you can literally watch a single fact arrive in the parameters over time. For three years, my friends Jiyan He and Zihan Zheng have been asking frontier LLMs the same question: "what do you know about USTC Hackergame?", a CTF contest. May 2024: GPT-4o invented fake titles. Feb 2025: Claude 3.7 Sonnet listed 19 verified 2023 challenges. By April 2026, frontier models recall specific challenges across consecutive years. After DeepSeek-V4 dropped, I instructed my agent to spend four days autonomously turning that habit into Incompressible Knowledge Probes (IKP) — 1,400 questions, 7 tiers of obscurity, 188 models, 27 vendors. Three findings: 1/ You can approximately size any black-box LLM from factual accuracy alone. Penalized accuracy is log-linear in log(params), R² = 0.917 on 89 open-weight models from 135M to 1.6T params. Project closed APIs onto the curve → GPT-5.5 ~9T, Claude Opus 4.7 ~4T, GPT-5.4 ~2.2T, Claude Sonnet 4.6 ~1.7T, Gemini 2.5 Pro ~1.2T (90% CI: 0.3-3x size). 2/ Citation count and h-index don't predict whether a frontier model recognizes a researcher. Two researchers with similar citation profiles get very different responses. Models memorize impact — work that shaped a field, not many incremental papers. 3/ Factual capacity doesn't compress over time. Across 96 open-weight models across 3 years, the IKP time coefficient is statistically zero, rejecting the Densing-Law prediction of +0.0117/month at p<10⁻¹⁵. Reasoning benchmarks saturate; factual capacity keeps scaling with parameters. Website: 01.me/research/ikp/ Paper: arxiv.org/pdf/2604.24827
70
227
55
2,158
394,751
Follow-up on non-English token-inefficiency with more model-language pairs: - Chinese is cheaper than English on major Chinese models - Gemini and Qwen provide least non-English tax - Anthropic has the highest tax by far; Kimi is next - Hindi is the worst-covered language here, despite its massive speaker base
The non-English tax is real. Sutton's Bitter Lesson, translated across languages and normalized to OpenAI English token count: Hindi: OpenAI 1.37×, Anthropic 3.24× Arabic: OpenAI 1.31×, Anthropic 2.86× Chinese: OpenAI 1.15×, Anthropic 1.71× Claude’s tokenizer charges a much higher linguistic tax.
62
249
130
1,249
505,780
这是GPT生成的中国海拔地图示意图。 很多历史故事要配上海拔地图和河流才能理解。 比如秦为什么东出统一,三国为什么长期三分,为什么得四川可为一方之主,北方为什么总要入侵南方,黄河为什么总能改写中国历史等等。
🤖 Made with AI
90
211
20
1,172
287,581
Interviewer : Why does zoom have offices?
202
1,056
40
27,475
1,444,601
Jer retweeted
2012年4月20日深夜,MIT学生把21层的54号楼直接变成巨型可玩俄罗斯方块! 153个窗户化身像素,每一格LED灯精准控制,旋转、掉落、消行全场实时操作——整整筹备4年半、2个月每晚偷偷安装! Campus Preview Weekend上,新生和校友当场玩到炸裂,这才是MIT的“hacks圣杯”!
156
68
7
778
338,246
海淀的一人公司(OPC)补贴。 10万OPC专项补贴:需要你是科技人才(毕业5年内/科研人员/企业技术骨干等),公司注册在海淀、做AI等重点赛道,一人有限责任公司,经认定就能拿。
26
12
132
71,659
Claude 马上要实行实名制了,按照官网的说法,中国护照应该是 100% 不被接受的,只接受支持的国家的政府颁发的证件。看起来一大堆中转站要倒闭了。Claude 的模型价格会越来越高。实名认证由 Persona Identities 提供。 support.claude.com/en/articl…
131
44
36
593
362,859
这是看过的最好的把 mac mini 变成一体式 AI 主机的方式之一,国内厂商应该跟进很快吧?
73
200
12
1,222
343,281
好治愈的动画 @ 给你的爱人吧
10
115
4
683
239,474