Expired50% confidenceEventExact time
Google DeepMind engineer Philip pointed out at the AI Engineer conference that almost everyone is using Skills for coding Agents, yet almost no one is writing evaluations for those Skills.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/19/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/17/2026(expired)
Sources
Related Claims
UnverifiedPractices at institutions such as OpenAI, Anthropic, and Google DeepMind indicate that when agent tasks span tens to hundreds of steps, traditional prompt engineering can no longer meet system-level reliability requirements.68% similarUnverified即便是最强的大模型面对陌生私有代码库依然无法凭空理解原作者设计意图，这是信息缺失问题而非智能不足问题68% similarVerified大多数 AI 编程智能体无法有效完成复杂任务的主要原因是上下文缺失，而非模型能力不足66% similarUnverifiedA meta-Skill that transforms everything is unachievable in the face of the Loop architecture; this is a boundary at the level of computability, analogous to the halting problem.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/563159API
curl https://kongchang.com/api/v1/knowledge/claims/563159MCP
get_claim(id=563159)