DeepSeek 智能体结合视觉模型，耗时 1 小时迭代生成动画
PandaBearFred · reddit · 2026-08-13
A developer shared an experiment using a PI Agent where the DeepSeek model wrote HTML animations and used the Muse-Glimmer vision model for evaluation.
- Workflow: After DeepSeek generated code, it called Chrome to take a screenshot and sent it to the vision model for review, then iteratively modified the code based on feedback until the vision model was satisfied.
- Results: This visual feedback loop took about 30-60 minutes. Compared to the DeepSeek-only version, the version with vision feedback better matched the prompt details, though the standalone DeepSeek version spontaneously added a fade-in intro, showing different creative flair.
「编程与Agent」频道最新
- 找不到 Mac 版童年游戏，让 ChatGPT 动手 15 分钟复活到 Mac 和 iPhone — Shleemy_Pants · 2026-10-16
- 开源 Agentflow 用 FastMCP+pgvector 打造跨智能体共享记忆网络 — agentflow0328 · 2026-10-16
- 产品要做「两次销售」：先让 Agent 选中，再让人买单 — FinanceYF5 · 2026-10-16
- Sentry CEO 晒图：AI 实习生在 code review 里反驳人类同事 — zeeg · 2026-10-16
- SWE-Duel:让编码 agent 互相出题埋 bug 的抗污染竞技场 — Consistent_Fox_2990 · 2026-10-16
- 把 Claude Code 等 agent 轨迹攒成数据集，微调开源小模型 — Rasmic · 2026-10-16