Original Summary

Hi guys We’re a small team of electronics engineers trying to build our dream EDA<p>Context: Astra computer-use building PCBs in KiCad was a gpt-3 kind of moment for us. Seeing an agent doing work that’s always required a dedicated professional was both exciting and unsettling. Results are still coarse, but we all saw how it went with coding — if the first take looks like this, we can expect it to be fast, cheap, and reliable in a year.<p>Our vision: Most engineers using human tools ← we&#x27;re here Most engineers using machines using human tools ← one year from now Most engineers using machines using tools built specifically for machines ← future<p>So we basically try to build a tool straight from the future<p>First, we built the EDA itself - Browser-based - KiCad&#x2F;Altium imports and exports supported - Shared workspace with real-time Figma-like collaboration<p>But it&#x27;s built around MCP workflows - MCP covers all features; anything you can do in traditional EDA, your agent can do in Nomoreda - MCP is optimized for product design work - built-in project logic, workflows, and checks - better outcomes with less trial and error compared to bare LLM - You can get decent outcomes starting from around DeepSeek 4 - level. Frontier models are obviously superior, but not a must. - Direct tool use cuts token spend compared to computer use - We don’t resell tokens - you use the harness subscription you already have

  • 情报分类：服务器与云资源
  • 分类依据：内容涉及服务器、云资源或网络线路
  • 信息来源：Hacker News 新项目
  • 发布时间：2026/9/23 01:24:57