Hi HackerNews,I ve been working with multimodal agentic systems ever since the ImageNet and DQN days.When I started vibecoding last year, I came up with new ways to set up projects to avoid known issues in long-horizon task execution. I thought it would be great if I had one simple file that a coding agent could read when I start a new project to address these known issues while improving efficiency and autonomy long-term.OpenMetaHarness is a cognitive architecture that enables coding agents to develop and execute on a long-horizon project vision from 0 to 1 with minimal human intervention and maximum token efficiency across multiple sessions and context windows.You will be able to direct a persistent engineering operation whose memory, judgment, and discipline compound with the project. As the project accumulates verified decisions, operating lessons, and category-specific trust, any compatible coding agent can resume with more context, repeat fewer mistakes, and work autonomously for longer stretches.Would love for everyone to try it and contribute! It’s mind blowing seeing it in action.