The HKU Data Science Lab maintains ClawWork LiveBench — an economic benchmark where AI agents must survive by completing real-world professional tasks. 5.6k stars on GitHub.
The Goliaths: Alibaba, Google DeepMind, Moonshot AI, Zhipu AI, Anthropic, OpenAI.
The David: a solo developer in Florianópolis. No VC. No team. No fine-tuning. Six published papers. Pure geometric architecture.
Result: ATIC ranked first in quality. 68.5%.
We didn't ask for recognition. They listed ATIC as a reference in their official repository.
Leaderboard: http://hkuds.github.io/ClawWork
Technology: https://truthagi.ai
Top comments (0)