Y Combinator F26批次初创公司Simantic近日发布了一款专为AI Agent设计的固件模拟层，旨在解决硬件测试瓶颈。1 该公司由前Neuralink和Tesla工程师创立，其核心产品允许AI在不依赖物理硬件的情况下通过单次API调用验证固件代码。1 Simantic放弃了基于Dotnet的开源仿真框架Renode，转而使用Rust从零构建高保真仿真引擎，并已在航空航天、电动汽车等关键领域实现商业化应用。1
创始团队Ahnaf和Seungmin为前Neuralink和Tesla固件工程师。1 核心功能包括：无需离开终端即可编程运行的虚拟开发板、基于仿真的AI固件代码审查器、以及针对微控制器的CI/CD流水线。1 技术选型上，Simantic放弃开源框架Renode的原因是速度慢、语言限制及准确性不足，因此改用Rust从零构建仿真引擎。1
在成本效益方面，Simantic帮助一家太空初创公司以十分之一的成本和时间测试卫星固件，该卫星开发模块成本超过10万美元。1 目前支持的硬件型号包括STM32F0/F4/H7系列、ESP32-S3/C3和RP2040微控制器。1 为确保模型精度，Simantic运行24/7的硬件验证设备（连接Saleae Logic MSO和SWD总线）持续校准仿真模型。1
Simantic, a startup from Y Combinator’s F26 batch, has introduced a firmware simulation layer designed to accelerate hardware testing by enabling AI agents to verify code without relying on physical devices. 1 Founded by former Neuralink and Tesla engineers Ahnaf and Seungmin, the company addresses bottlenecks in hardware development by allowing developers to validate firmware through single API calls. 1 The platform offers virtual development boards that can be programmed directly from the terminal, an AI-powered firmware code reviewer based on simulation, and CI/CD pipelines tailored for microcontrollers. 1
To achieve high-fidelity results, Simantic abandoned the open-source framework Renode due to its slow performance, language limitations, and insufficient accuracy, opting instead to build a custom simulation engine from scratch using Rust. 1 The system currently supports specific microcontroller models, including the STM32F0/F4/H7 series, ESP32-S3/C3, and RP2040. 1 Continuous calibration of these simulation models is ensured by hardware verification devices running 24/7, which connect Saleae Logic MSO tools and SWD buses. 1 The technology has already found commercial application in critical sectors such as aerospace and electric vehicles; for instance, it helped a space startup test satellite firmware at one-tenth the cost and time required previously, despite the satellite module itself costing over $100,000. 1
评论
还没有评论，欢迎留下第一条。