问HN：人工智能编程的商业现状
我建议一位创始人使用人工智能进行编码，并向他介绍了replit.com。我们的初步理解是，如果人工成本为100美元，我们可以通过replit以1美元的成本更快地完成编码。
在原型开发后，我意识到它的后端模型有些浅薄，并且总是回到它想要做的事情上。如果在使用React Native时遇到挑战，它并不能解决问题，而是在几次尝试后回归到“更简单的HTML/CSS”方法。
我想稍微放宽视角，问一下是否有同样简单但更好的编码助手。理想情况下，我希望能有一个以移动优先为理念的工具，鼓励我不写任何代码并快速推进。
我之所以提这个问题，是因为我现在意识到，我们可能需要在replit上花费的不仅仅是1美元，而是10美元。只希望这不是我开始的一段“剃羊毛”的旅程。
是否可以通过程序员使用的优秀母提示来挽救这个问题？我在想，因为这些前端最终都必须连接到类似deepseek、gemini或chatgpt的API，对吧？如果你有其他有效的建议，也欢迎分享。
查看原文
Advising a founder to use AI to code, I introduced him to replit.com<p>Our initial understanding was that if humans will cost $100 we can code this via replit in $1 and faster.<p>After prototyping I have realised that it has a somewhat shallow model at backend and keeps reverting to what it wants to do. If there is a challenge with react native it is not being able to fix but reverts to "a simpler htmlcss" approach after few attempts.<p>I wanted to zoom out a little and ask if there is an equally simple, but better coding assistant than this. Ideally I hope there is something that makes this mobile first and encourages me to not write any code and move really fast.<p>Asking because I now realise we might have to spend not $1 but $10 with replit. Just hoping this isn't a yak shaving journey I've embarked on.<p>Could this be salvaged by a good mother prompt that programmers use? Wondering because most of these frontends must be finally connected to similar deepseek/gemini/chatgpt apis no? Any other suggestions that worked for you are also welcome.