返回首页
最新
我在田纳西州的乡村创建了这个项目,没有计算机科学背景,仅凭借密集的人工智能协作和模式识别。<p>问题在于:陪审团选择依赖直觉,没有任何文档支持。当上诉发生时,律师无法证明他们的过程是合理的。<p>MA'AT 创建了哈希验证的审计轨迹。它分析陈述的一致性,不依赖心理学,而是通过连贯性评分、矛盾计数和证据缺口来进行分析。<p>关键限制:一个语言防火墙在报告到达之前阻止了诊断术语(“诊断”、“原型”、“预测”)。这是经过设计的,能够在法庭上生存。<p>技术栈:Python/FastAPI 后端,SQLite,原生 JS 前端。使用 Docker 部署。<p>免费层是永久的。我们的理念是“强者为基础设施买单,弱者获得机会。”<p>希望能获得关于技术方案的反馈,以及法律限制是否合理的意见。
我从11岁开始编程,现在已经38岁了。在这27年里,我曾在EA DICE工作过游戏引擎,曾在OTOY和Embark进行渲染研究,现在经营一家为AMD、Intel和Meta提供服务的图形研发公司。我见证了许多新工具的兴起与消亡。
当AI编码工具开始流行时,我需要弄清楚:这是我公司应该投资的领域,还是可以安全忽视几年的事情?
我决定彻底解决这个问题:全力以赴几周,构建一些真实的东西,超出我们现有的代码库,使用自主智能AI——不仅仅是自动补全,而是让Claude Opus实际编写文件、运行命令和提交代码。
三周后,我得到了答案。
Sonars在三周内从零到上线。一个开发者。纯自主智能AI。一旦核心功能正常工作,我就用Sonars来构建Sonars。
我在其他工具上遇到的一个问题是:Claude直接在我的工作目录中进行更改。一次实验性的提示就让我分支变得一团糟——未完成的功能、破损的测试、与自己产生的合并冲突。
回过头来看,解决方案显而易见:Git工作树。Sonars中的每个AI会话都在完全隔离的环境中运行。Claude可以尽情发挥——读取文件、编写代码、运行测试、提交更改——而你的主分支在你明确合并之前保持不变。
功能特点:
- Claude可以读取文件、编写代码、运行终端命令、提交更改
- 每个会话在Git工作树中隔离
- 实时流式处理,具有扩展思考模式
- 带有语法高亮的可视化差异查看器
- 原生Rust + Dioxus应用
技术栈:Rust + Dioxus(前端),Axum(后端),SQLite + PostgreSQL,gix(纯Rust Git)。
提供免费层:[https://sonars.dev](https://sonars.dev)
经过27年并发布游戏引擎,我并不是轻易说出这句话:这是我有史以来最具生产力的时刻。对我最初问题的答案?这并不是可以忽视的事情。
欢迎提问有关架构、工作树方法或使用自主智能AI构建Rust的实际体验。
人类是否能够征服如此巨大的规模?
Hi HN!<p>My name's Robert Escriva. I got my PhD from Cornell's Computer Science department back in 2017. And three years ago I had a psychotic episode that irreversibly shook up my world.<p>Since then I've been applying a skill I learned in grad school---namely, debugging distributed and complex systems---to my own mind.<p>What I've found I've put into a [book (pdf)](<a href="https://rescrv.net/engineering-schizophrenia.pdf" rel="nofollow">https://rescrv.net/engineering-schizophrenia.pdf</a>) on engineering, my particular schizophrenic delusions, and how people who suffer as I once did can find a way through the fog to the other side.<p>This is not a healing memoir; it is a guide and a warning for all those who never stopped to ask,
"What happens if my brain begins to fail me?"<p>I am writing because what I've found is not a destination, but a process. It is an ongoing process for me and for people like me. I also believe it is automate-able using the same techniques we apply to machine-based systems.<p>I am looking for others who recognize the stakes of the human mind to engage in discussion on the
topic.<p>Happy hacking,
Robert
Simply paste a product URL, answer a few personalized questions about your needs and preferences, and our AI analyzes your responses to provide a calm, reasoned recommendation.
You can paste links to products from any e-commerce site including Amazon, eBay, Best Buy, and most online retailers. Our AI analyzes the product to generate tailored questions.