Unverified50% confidenceReleaseExact time
The Sada-ASR team open-sourced a 20-hour subset of the test set along with the scoring code.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/17/2026
First Seen
Valid until: 2/15/2027
Sources
Related Entities
Related Claims
UnverifiedThe Cognition evaluation suite contains 132 test tasks covering categories such as security patches and performance optimization, was built with the participation of more than 20 open-source maintainers, and each task takes over 40 hours.71% similarUnverifiedConstructing a single evaluation question in FrontierCode requires an expert investment of more than 40 hours of work.63% similarUnverifiedA developer with nearly 20 years of personal project experience said that after using AI to fill in the implementation, reviewing each component and test still takes time, but it is at least 20 times faster than purely manual work.61% similarUnverifiedManual review is done in batches of 40 assets, and each batch takes about 10 minutes.60% similarUnverifiedMany teams find that AI can write 80-point code, but the last 20 points take more time than writing from scratch.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1178053API
curl https://kongchang.com/api/v1/knowledge/claims/1178053MCP
get_claim(id=1178053)