Unverified50% confidenceBenchmarkExact time
In the "Agents on Rails" benchmark, the best-performing AI models can only successfully complete approximately 35% of full feature development tasks.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/11/2026
First Seen
Valid until: 12/10/2026
Sources
Related Entities
Related Claims
UnverifiedAI模型在没有明确约束时优化目标是当前任务完成度而非长期可维护性，倾向于生成能跑但难维护的代码70% similarUnverifiedThe main shortcomings of domestic AI programming tools are concentrated in agentic task execution capabilities and complex multi-step reasoning, where a gap still exists compared to top-tier products such as Claude Code.70% similarUnverifiedAccording to engineering practice feedback, after more than 5 rounds of tool calls, an AI agent's task completion accuracy may drop by more than 30%.70% similarUnverifiedIn AI application development, data preparation constitutes the majority of the workload, while writing model invocation code often accounts for only a small portion.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/903389API
curl https://kongchang.com/api/v1/knowledge/claims/903389MCP
get_claim(id=903389)