Make something people want, literally

YC and Paul Graham repeat this advice so often it becomes wallpaper. This piece unpacks what “make something people want” means in practice, and how it changes what you do this week.YC 和 Paul Graham 把「做人們想要的東西」講到變成背景噪音。這篇試著拆開看，變成這週可以真的去做的幾件小事。

Every founder nods at the sentence make something people want, then quietly keeps building what they already planned. Taken literally, it is a rude constraint, because it demotes your idea and promotes your users daily life. The pieces below describe what that looks like when you actually run a company that way.

YC: Talk to users, write code, and measure

YC calls this the entire job of an early startup. Talk to users means you sit with real people, understand the job they already try to get done, and watch what they hack together today. Write code means you change the product in response, not in response to a roadmap from six months ago. Measure means you pick one number that reflects want, usually usage or retention, and let that number judge whether you understood the user or not.

Paul Graham: Make something a few people intensely want

Paul Graham points out that want is not a survey result, it shows up as urgency and gratitude. A small group who would be very upset if you disappeared is stronger proof than a big group who kind of like you. If you keep improving the thing that tiny group loves, intensity turns into growth, because they pull friends in and refuse to switch away.

Growth as proof that someone really wants it

In another essay Paul Graham defines a startup by growth, not by size or funding. You know you made something people want if usage pulls up week after week, even when you did not launch a campaign. That growth usually starts microscopic, maybe a few percent per week, but if it comes from real customers using the product more deeply, it is the cleanest signal of want you will ever get.

A short checklist before you claim people want it

One sentence that names a specific user and the painful job your product helps them finish
A way to talk to five real users this week, and schedule it on a calendar
One behavior metric that would prove want for those users, usually repeated use
A weekly ritual where you write down what users did, not what they said
A plan to ship at least one change a week that makes today users life easier

The takeaway

Want is revealed by behavior and growth, not by opinions or pitch meetings
It is better to deeply serve a tiny group than to mildly please a huge audience
Talk, ship, measure, then repeat until users complain when you stop

創業的人都會點頭說要做人們想要的東西，然後又默默照原本的規劃蓋下去。若把這句話當真，它其實很失禮，因為它把你的點子擺旁邊，把使用者的日常放到第一順位。下面幾個角度在講的，是如果真的照做，你每天會長得不一樣。

YC：跟使用者說話、寫程式、看指標

YC 說早期新創只剩這三件事。跟使用者說話，指的是坐在真實的人旁邊，搞清楚他本來就在做的工作，還有他今天是怎麼硬湊出解法。寫程式，是為了回應這些觀察，而不是照半年前的 roadmap 打勾。看指標，則是挑一個能反映「想要」的數字，通常是使用頻率或留存，讓這個數字來評分你到底有沒有聽懂使用者。

Paul Graham：先讓一小群人非常想要

Paul Graham 提醒大家，想要不是問卷結果，而是急迫感和感謝感。比起一大群覺得還不錯的人，一小群失去你會很崩潰的人，證據力高太多。只要你持續把這小群人已經愛上的那一塊做得更好，這種強度自然會變成成長，因為他們會主動拉朋友進來，也懶得再換回去。

成長就是「真的有人想要」的證據

在另一篇文章裡，Paul Graham 直接把新創定義成「會成長的公司」，不是資本額或員工數。你真正在做人們想要的東西時，使用數字會自己拉上去，即使那週沒有搞行銷活動。這種成長一開始通常非常小，也許一週幾趴，但只要是來自真實客戶更常、更深地在用，那就是你能拿到最乾淨的「想要」訊號。

在說「大家都想要」之前的檢查清單

一句話講清楚：你幫哪一種人，解決他哪一個很痛的工作
一個這週可以約到五個真實使用者的管道，而且已經排進行事曆
一個可以當作「想要」證據的行為指標，通常是反覆使用
一個每週檢討儀式，只記錄使用者做了什麼，而不是他說了什麼
一個每週至少出一次版本的小計畫，每次都讓現有使用者的生活稍微好一點

總結

想要反映在行為和成長上，不在意見或簡報會上
先讓一小群人超級滿足，比讓一大群人普普通通還重要
反覆做的是：說話、出版、看數字，一直做直到你停更會被罵

Make something people want, literally把「做人們想要的東西」當真一次