A 3D Tech Radar Built on AIHOT: Curated News, Multi-Platform Drafts, and an Agent-Automation Future
A general-purpose open-source aggregator can be bent into a niche monitoring and publishing tool without building the whole pipeline from scratch. The project also illustrates a common agent-design constraint: human attention, not model capability, sets the ceiling on how many tasks a single operator can supervise.
AIHOT, an open-source hot-topic aggregator from Chinese developer Card God (卡神), already handles collection, deduplication, scoring, and event grouping. One developer remixed it into a 3D tech radar by narrowing the sources to Three.js, Shader, digital twin, visualization dashboards, and AI+3D, then tightening the filters to items with source code, runnable demos, tutorials, or notable version changes. Retrieval went from a pile-ahead policy to a six-hour cycle that prioritizes the latest 48 hours of material, keeping model-call quotas from drowning in backlog.
The remix added a content generation workbench that turns a single hot item into drafts for Xiaohongshu, Douyin, Bilibili, Zhihu, and WeChat, each with platform-specific formats: Xiaohongshu gets image cards plus captions, Douyin a 60-90 second voiceover script, Bilibili a full script with demo steps, Zhihu a problem-led technical explanation, and WeChat a long-form piece with personal analysis. The system also connects to Notion for archiving and to 墨排, the developer's own editor, for layout, and it supports daily, weekly, and monthly reports with PDF or platform-specific image exports.
The author says many features were added iteratively after hitting friction points, and admits that the performance-tracking loop still needs real published data to validate. The stated next step is to move from click-triggered execution to signal-triggered automation, following the principle that human attention, not model capability, is the real bottleneck in multi-task agent systems.
The iterative feature additions read like a priority queue of user pain: each new capability appeared only after the previous step became annoying, which is a sane way to build personal tools.
Quoting the principle that human attention is the scarcest resource in agent systems, the article points toward a design where automation escalates only at decision points; this is a stance worth adopting beyond this niche.
The radar's real value lies not in AI-generated drafts but in the surrounding infrastructure: source filtering, retrieval cadence, archiving, and formatting that keep the human in the loop.