Nefling ᚨ retweeted
Rockstar Games 周三解雇了我。第二天 GTA 6 就发售了
我是 Rockstar North 的高级引擎工程师,从事 RAGE 性能优化。三个月前,我在遥测模块里发现了一条本不该存在的数据管道,而且我能证明它流向了哪里
此刻每一份正在运行的 GTA 6 都在收集玩家的行为数据。这部分不算新闻。每款游戏都会收集遥测数据。隐私政策里写的是"用于改进未来产品的匿名游戏数据分析"。我发现的不是这个
遥测模块的数据会汇入 Take-Two 内部的一个系统,名为 Lens。Lens 并不会把匿名使用统计数据发送到 Rockstar 开发人员的数据看板。它为每一位玩家建立一份个人行为画像。不是匿名的。与你的主机账号、你的平台 ID、你的 IP 绑定
我在8月14日向我的技术主管报告了这件事。他说这是"用于玩家体验研究的扩展分析",让我别管这事。我要求提供数据处理协议。他说这是在 Take-Two 集团层面处理的,不在 Rockstar 的职责范围之内
我调取了 Lens 数据管道的完整文档。我发现的内容如下:
遥测的内容远远超出游戏时长和崩溃报告。Lens 以毫秒级精度记录每一次输入。按键时间点、摇杆动作、反应速度、PS5 自适应扳机上的扳机压力模式。它实时运行行为分析,并生成他们称之为"玩家行为签名"的东西 - 一份对每位玩家都独一无二的画像,就像指纹对每个人都独一无二一样
Lens 会从你的游戏过程中推导出:冲动控制评分、风险承受度指数、攻击性反应模式、耐心阈值、压力下的决策速度、消费冲动评级。这些不是调试指标。这些是心理行为类别
直到我找到了对外传输的管道,我才明白它们是用来做什么的
这些数据不会留在 Rockstar。也不会留在 Take-Two。Lens 通过内部 API 将行为画像传送给 Take-Two 旗下一家名为 Vantage Analytics 的子公司。Vantage 没有公开网站,没有新闻页面,没有招聘信息。它于2023年在特拉华州注册成立
Vantage 是一家行为 AI 公司。它用玩家画像训练预测模型,并把这些模型授权给商业客户。管道文档中包含一份不完整的客户清单。两家保险集团,一家信用风险建模公司,一个人力资源筛选平台,以及一家广告数据经纪商
你的 GTA 6 游戏过程此刻正在生成一份行为评分。没人追你的时候,你开得有多快。你会不会在红灯前停下。NPC 无端攻击你时,你怎么反应。你行动之前要想多久。你花游戏内的钱有多冲动。每一个选择都在喂养一个模型,企业正在付费查询这个模型,用来对真实的人做决定
Nefling ᚨ retweeted
AI is pay to play now.
In 2025 I had a $20 Cursor sub and it was enough.
In 2026 I have three ChatGPT Pro accounts, a Claude Max, and a SuperGrok Heavy. $1,100 a month.
GPT 6 Astra is the best model I have ever used and it eats a full week of Pro usage in 4 hours.
Claude Code limits are about to drop another 17%.
The best models are behind a wall and the wall gets higher every month.
We are watching the permanent underclass get built in real time.
Is a "dangerous" AI another plan for the world just like 2020 was?? Let us take some more of your freedom because there is this scary AI that escaped a lab... a lab just like the last time.
5
Well it's getting weirder and weirder, he's not the first with this take. What do you think? The world accelerated like crazy this past year. "There are times where nothing happens in 10 years, then decades happen in a few weeks" - paraphrasing
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
34
Replying to @TTrimoreau
Everything finished" is a fantasy. You'll keep adding features nobody asked for.
The first person who pays tells you more than 6 months of building in silence.
1
19
Replying to @TTrimoreau
Just scored my first B2B automation, for a client and their 130 sales agents! You are right it's a whole new game.
17
Oh, you step away for a second and bang! You are behind the updates again! Looks interesting need to giv e it a try!
Introducing Claude Design by Anthropic Labs: make prototypes, slides, and one-pagers by talking to Claude.
Powered by Claude Opus 4.7, our most capable vision model. Available in research preview on the Pro, Max, Team, and Enterprise plans, rolling out throughout the day.
1
1
17
Introducing Claude Design by Anthropic Labs: make prototypes, slides, and one-pagers by talking to Claude.
Powered by Claude Opus 4.7, our most capable vision model. Available in research preview on the Pro, Max, Team, and Enterprise plans, rolling out throughout the day.
Replying to @aiedge_
Who cares if we have opus right? I have heard that codex is much bette Reith usage than Claude is that true?
20
Replying to @dramaricic
Haha yes probably. I'm guessing the benchmarks are comparing the dumbed down version of 4.6 to 4.7. but on the serious note its probably still better than the smart 4.6.
1
1
49










