# A cyclist made with Pi and Mercury

LotusDecoder identifies Pi + Mercury 3 medium. Inception identifies Mercury 3 as a text diffusion LLM; Pi documents its coding-agent workflow. Together with the original attribution and rendered animation, this supports code generation rather than direct text-to-video. Exact code, prompt and iteration count remain unpublished. Model attribution is source-reported and unverified. Full recording and source are retained; unpublished code is not mirrored and no onsite Play demo is offered.

- Recorded date: 2026-10-20
- Date precision: Day
- Date basis: source-publication
- Model (source label): Mercury 3 Medium
- Model attribution: Source-reported, not independently authenticated
- Creator / publisher: LotusDecoder (@LotusDecoder)
- Format: Animation
- Prompt category: Source-documented test
- Original prompt: mercury-3 测试
内测中  diffusion model  ，

来自 @_inception_ai 
作为市面上少见的商业闭源非 transformer 模型。
跑分现在说对标 gpt-6-luna ， haiku-5.5 。
优势是速度超快，1000 tokens/s。
测了一下，鹈鹕骑车。

Pi + Mercury 3  medium
- Last updated: 2026-10-22
- caseRole: case
- caseNumber: 1110
- timelineVisible: True
- datePrecision: day
- dateBasis: source-publication
- Original source: https://x.com/LotusDecoder/status/2112467354630418609
- Archive record: https://pelicanmap.aveniqa.com/en/specimens/x-mercury-pelican-bicycle-2026-10-20/
- Rights: Works belong to their original creators. Archiving does not change their licenses.
- Media: https://pelicanmap.aveniqa.com/media/collected/2026-10-22-derivative-collection/x-mercury-pelican-bicycle-2026-10-20.mp4
  Source: https://video.twimg.com/amplify_video/2112467089634316288/vid/avc1/1920x1080/9yTIN2B-M5HRCRp9.mp4?tag=29