Unverified50% confidenceOpinionExact time
Blow proposes that the system that truly succeeds may not be today's large language model approach, but rather an AlphaZero-style system trained on rules and experience, with an LLM used as a conversational interface to bridge it.
1
Sources
50%
Confidence
Long-term
Relevance
11/16/2026
First Seen
Sources
Jonathan Blow: The AI Coding Myth and the Truth About Software Quality
bilibili一摩尔炸鸡翅11/15/2026
Related Entities
Related Claims
UnverifiedFor tools that go from natural language to executable models, feasible approaches usually involve adding a validation layer after LLM parsing, or providing a manual correction interface.68% similarUnverifiedCombining the conversational capabilities of LLMs with traditional automation tools is often more practical than purely pursuing model intelligence.68% similarUnverifiedLinguaArray's success demonstrates one possible path for combining LLMs with formal methods: using the LLM as a semantic guide that collaborates with the solver through a verifiable interface.68% similarUnverified模型的工具调用能力来自专门的微调训练,而非通用语言模型的自然涌现67% similarUnverifiedEarly LLM applications were mostly single-turn input-output patterns where prompt tuning could solve most problems, but Agentic workflows have made application logic increasingly complex66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1173731API
curl https://kongchang.com/api/v1/knowledge/claims/1173731MCP
get_claim(id=1173731)