#web-agents

HN did not push Browser Harness because it was another browser wrapper. It took off because the repo lets an LLM patch its own browser helpers in the middle of a task, trading safety rails for raw flexibility.

MM-WebAgent tackles a real flaw in AI-made webpages: models can generate pieces, but the page often loses visual coherence. The paper adds hierarchical planning, self-reflection, a benchmark, and released code/data so builders can test multimodal webpage agents beyond code-only output.