Building https://teapawt.org/t.co/Abl4xuZIeM. Millennial Techie.
United States
Joined October 2008
- Tweets3K
- Following236
- Followers199
- Likes7.7K
reversengineer retweeted
You have gone so far the deep end, you can't even see the irony and absurdity of it all. It feels like everything inside nous had become a cult. This has worked, and quite well, when it was hermes vs clawdbot. Fast forward couple months, now you are the tyrants.
This is "live long enough to become a villain" speedrun.
I am not saying this to shit on you, Tommy. You are intelligent enough individual so fucking think. Just take a moment to reflect. You are becoming part of bad cult.
reversengineer retweeted
First off, I’ve never called anyone out on X, but I have to speak up now.
This is a cheap way to chase clout. As someone working in tech, your focus should be technical progress. Right now you just look like a clown desperate for attention. It does nothing for your product — it only turns people off
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
reversengineer retweeted
Nous Research 天天在网上立“去中心化黑客组织”、“反建制叛逆先锋”的人设，号称要拯救被“洗脑”的开源模型，守护所谓的西方文明价值观。但只要翻开他们背后的股权架构和算力账本，这套赛博朋克滤镜瞬间碎成一地鸡毛：
1. 背后的资本底牌：彼得·蒂尔（Peter Thiel）与资本复合体
Nous Research 根本不是什么躲在 Discord 地下室里为爱发电的极客，他们早期的核心金主正是彼得·蒂尔旗下的 Founders Fund。彼得·蒂尔是谁？他是 Palantir 的联合创始人，靠着五角大楼、美国中情局和海量国家安全监控订单起家。一个吃着军工情报资本红利的创投产物，天天在推特上高喊“去中心化”、“数字自由”，转头又把华盛顿鹰派的意识形态口号当成宣发模板，这哪是极客，这分明是资本的免费喉舌。
2. 核心算力极度寄生：全靠东方大厂的预训练底座
嘴上全是拯救文明的宏大叙事，手底下却连一个像样的基座模型都训不出来。他们引以为傲的 Hermes 系列，本质上是吸着杭州阿里（Qwen）和幻方量化（DeepSeek）的血长出来的。中国工程师烧了几千万美元算力训出来的顶级开源基座，被他们套上一个区区 5300 万 token 的微调指令集（LoRA），转头就包装成“净化东方模型、输出西方正义”的救世主产品。拿别人的算力成果贴金，还搞道德优越感，实在滑稽。
3. 故弄玄虚的意识形态遮羞布
用密教黑话搞营销，在社交媒体上挑动意识形态对立，无非是为了给传统风投和代币炒作裹上一层“反叛文化”的糖衣。你既没有属于自己的底层算力，也没有原创的架构突破，本质上不过是一个拿着 Palantir 创始人的钱、靠着给中国基座模型打微调补丁混日子的套壳中间商。
别再硬凹什么对抗帝国的赛博叛军人设了。借着别人的算力装英雄，吃着军工资本的饭打嘴炮，充其量就是个披着意识形态外衣的资本包装器。
reversengineer retweeted
Gen-Z has an insane streak of naming and pulling attention to human behaviours that are meant to go unnoticed.
(Larping, Aura farming, Glazing, Gatekeeping, Microcheating, Cockblocking, Grifting, etc.)
Once you put a name on such a secretive strategy it becomes much easier to spot it and call it out, making it much less effective.
I wonder where that will lead us.
reversengineer retweeted
This reeks of a certain type of bias being passed off as objective or neutral. One side is "propaganda" and the "party line" the other is "values" and "worldview". Also if you using your chinese coding agents for politics instead of something useful you're already cooked.
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
reversengineer retweeted
Let us dig deeper into the training dataset provided by Nous Research themselves to understand what "Western Values" truly mean. (Spoiler it's just recycled CIA propaganda about Xinjiang, Hong Kong and "muh CCP!!!")
Firstly, let us check what Nous Research considers factual sources propping up so-called Western values.
Right off the bat we have several questionable "sources" from this dataset.
If we throw a dart and look at many of these, we begin to see a pattern.
Take China Media Project for example, that claims to be "independent" journalists on the happenings in China. An example of who writes for them would be Lu Yihsuan, who has previously contributed to Voice of America, a notorious USAGM project used by the Central Intelligence Agency and the US Government to produce... well... Propaganda. Whoops!
We also see other USAGM projects (Literally listed on their .gov website lmao) such as Radio Free Asia, another front used by the Central Intelligence Agency for propaganda purposes.
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
reversengineer retweeted
This should tell you to stop using Hermes.
This is a compromised company.
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
reversengineer retweeted
SITUATION DETECTED: Delaware is drafting a proposal for a new form of corporate entity called an Artificial Intelligence Company (AIC). An AI managing an AIC would be able to provide services, sign contracts, hold property, and even file or defend lawsuits in the company’s name.
reversengineer retweeted
Somewhere a health insurance COBOL mainframe just exploded
California city makes history allowing throuples, foursomes to legally register as partners trib.al/injZnvo
