# devex does not matter. it is now ai-devex that matters.

Source: https://x.com/skylar_eth/status/1902064253051568428

## Summary

The post argues that developer experience now depends on how easily AI can understand documentation and produce working apps from it, which the author calls "ai-devex." The author suggests that every library, framework and tool add an llms.txt file addressing common mistakes AI models make. The suggested process is to attempt to build an app from the docs, note what the AI consistently gets wrong, and then fix the docs or add AI-specific context, such as a snippet developers can paste into Cursor at the start of each project. The post names Solidity, scaffold-eth, Hardhat, Vyper and Foundry as projects that should do this.

## Article

hot take: devex does not matters. it is now ai-devex that matters. 

how easy is it for AI to understand your docs and correctly oneshot apps that work?

one idea: all libraries/frameworks/tools should add an llms.txt that address common gotchas that AIs get wrong. try to oneshot an app using your docs, see what it consistently gets wrong, fix the docs or clarify more AI-specific context in llms.txt or an easily copy-paste snippet that developers can copy into cursor at the start of each project.  

solidity should do this
scaffold-eth should do this
hardhat
vyper
foundry 
etc etc