Show HN: SandBox – AI agents simulating possible futures (github.com)

SandBox is a trio of experimental projects demonstrating agentic AI workflows that simulate and reason about possible futures. AIBlog is an autonomous researcher that crawls arXiv, OpenAI, DeepMind, HuggingFace and other sources, synthesizes multi-source research via a ReAct agent wrapped in a LangGraph state machine, and publishes fully styled HTML articles complete with citations, code snippets and a DALL·E 3 banner hosted on Azure Blob. TomorrowNews uses LangChain agents plus Azure OpenAI (GPT‑4) to ingest real-time news (hourly), generate speculative next‑day headlines and detailed articles, and produce a responsive, image‑rich HTML newspaper using an Image Generation tool. GenBox runs a structured prompt–response loop to produce daily high‑level governance decisions for a simulated world (economy, society, environment, politics). Technically significant features include LangGraph’s stateful, multi‑actor workflows supporting cycles/branching, persistence, streaming outputs and optional human‑in‑the‑loop interruption, plus integration with LangChain/LangSmith. The stack leverages recursive agent architectures, streaming node outputs, and Azure OpenAI image/text models to automate multi‑step analysis, synthesis and publication. The projects highlight both capabilities—scalable autonomous content creation, iterative forecasting and simulated decision‑making—and risks: potential misinformation, verification challenges, and governance/ethics concerns. As open experiments, they offer a practical sandbox for researching agent orchestration, stateful reasoning, and the policy implications of increasingly autonomous AI systems.