CASE LAB / CONTROLLED COMPARISONS
Same task. Why are the results different?
Keep the input the same, record the model, Agent Harness, time and cost, then open each generated result and judge for yourself.
- 03
- Cases
- 10
- Runs
- 07
- Models
PUBLISHED CASES / 01
Start from the results
Each Case keeps the shared input, run conditions, cost data, the full web pages and the review conclusions.
01Image to HTML / 2026-08-16
Image to HTML: a ramen ordering site
Gemini 3.7 Flash runs in Pi Harness and in DeepSeek Harness with the same reference image and the same task, generating an interactive ramen ordering web page in each.
- Models
- Gemini 3.7 Flash
- Runs
- 2
- Winner
- Pi Harness
02Text to Three.js / 2026-08-18
Text to Three.js: Airbus H145 helicopter
Using exactly the same Prompt in Pi Harness, Grok 4.5, Grok 4.6, Kimi K3 and GLM 5.2 each generate an interactive Three.js helicopter HTML.
- Models
- 4 models
- Runs
- 4
- Harness
- Pi Harness
03SVG Portrait Animation / 2026-08-20
SVG Animation: a portrait of Lin Yueji
With Pi Harness as the single execution environment and Qwen Max, Kimi K3, GLM 5.2 and DeepSeek V4 Pro connected through EvoLink, the same Prompt generates a single-file, replayable SVG hand-drawn portrait animation.
- Models
- 4 models
- Runs
- 4
- Harness
- Pi Harness