Anthropic has demonstrated through 400k conversations that: programming ability is depreciating, and industry knowledge is the moat in the AI era.

The report shows that the success rate of non-programmers using intelligent agents to write code has narrowed to within 7% of professional software engineers, with management roles even slightly surpassing them.
Users lead 70% of planning decisions, while intelligent agents handle 80% of execution.
Expert-level conversations produce 3,200 words per session, more than five times that of beginners;
beginners have a 19% frustration and dropout rate, while intermediate and advanced users only 5%-7%.
What does this mean for the crypto industry?
In the past, the development threshold for DeFi protocols and smart contracts was being leveled by AI tools.
Business understanding—such as knowledge of AMM mechanisms, tokenomics, and cross-chain bridge risks—is more critical than Solidity proficiency.
Product managers, analysts, and even community operators without programming skills may directly participate in protocol development.
Meanwhile, the use cases for intelligent agents are shifting from debugging to end-to-end creation.
Debugging sessions accounted for 33% seven months ago, dropping to 19%, while deployment, operation, and data analysis have doubled in share.
The average estimated value of tasks has increased by about 25%.
This indicates that AI is transforming from an auxiliary tool into a core productivity driver.
But risks also exist: low barriers mean an influx of low-quality code.
Audit demands will not decrease; they may even increase due to the inscrutability of AI-generated code.
Teams relying on AI development need stricter security reviews and testing processes.
The competition in the crypto industry is shifting from technical prowess to industry knowledge.
Who understands the business better will be able to build products faster with AI.
$ai #amm #defi #链上数据 #ai