Worst of all, a new feature—an AI tutor that Lin Wei had reluctantly added to compete with ChatGPT—began answering questions instantly. And while it was efficient, something was lost. Users stopped explaining why an answer worked. They just pasted the AI’s output and moved on.
Lin Wei still codes on weekends. Maya runs community workshops on "digital kindness." Jun built an open-source version of the whiteboard for rural schools with no internet access. studykaki
Within 20 minutes, three different users had annotated the whiteboard. A student in Surabaya drew the contour. A teacher in Manila corrected a sign error. And a man in Taipei—who had once been that lonely student—added a final note in the margin: Worst of all, a new feature—an AI tutor
They rolled back the AI tutor to a "last resort" button. They removed seed farming by capping daily reputation gains. They introduced a "slow lane" for the whiteboard—answers took at least one hour to appear, forcing users to think before typing. They just pasted the AI’s output and moved on
He shared the link in three small Facebook groups: "Taipei Engineering Students," "Self-Study Physics," and "Late Night Coders." For two weeks, nothing happened. Then, one Tuesday morning, he woke up to 14 notifications. A student in Kaohsiung had answered his fluid mechanics question—not with a text reply, but by uploading a step-by-step diagram, annotated in red, with arrows showing the flow separation point.
But Lin Wei saw a problem. The platform was becoming… noisy.