@ElorianAI Agent 视觉推理：从像素生成代码解迷宫
Could you find the way out of a 32x32 maze?
清晰拆解了 Agent 从视觉输入到代码执行再到结果验证的具体工作流，技术细节扎实且结构完整，适合创作者参考如何展示 AI 产品的核心能力。
Could you find the way out of a 32x32 maze?
你能找到32x32迷宫的出口吗？
@ElorianAI gave their agent brand new mazes, from a tiny 8x8 up to that 32x32 one
@ElorianAI 给他们的智能体提供了全新的迷宫，从极小的8x8到那个32x32的都有
The coolest part is how it works → it writes code to read the walls right from the pixels, searches for the way out, then checks the route against the image before answering
最酷的部分在于其工作原理 → 它编写代码直接从像素中读取墙壁信息，搜索出口路径，然后在回答前将路线与图像进行比对验证
Thats real visual thinking, same skill matters for robots, factory lines and game replays, where AI has to really see what's going on
这是真正的视觉思维能力，这项技能对机器人、生产线和游戏回放同样重要，在这些场景中AI必须真正看清正在发生的情况
Made by the team that led Gemini and Apple's MM-1
由主导Gemini和Apple MM-1项目的团队打造
Glimpse is free to try → https://platform.elorian.ai
Glimpse 可免费试用 → https://platform.elorian.ai
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力