Everyone Got Faster. Did the Team?
TL;DR: I get a lot more done with agents now. But as soon as the work goes back to the team, why does everything still feel like such a slog?
Working Alone Got Faster
I’m getting a lot more done these days. Fire up an agent, and I can push a fairly messy piece of work a long way in a few hours.
Then I take it back to the team. I still have to explain this, align on that, and sometimes it feels like more work than before.
I can spend a few hours building something with an agent. The moment I bring it to the team, half that speed is gone.
It’s not that the work is bad. All the back-and-forth is still sitting in my head and that one agent session.
I used to think this was a knowledge base problem. Put the documents in one place, make search better, and get AI to summarize everything. That should do it.
Turns out, not really.
What Actually Counts
The problem isn’t that we don’t know what people said. The problem is knowing what still counts.
Which decision is the real one? Who actually took it on? If I hand it to another person, or another agent, can they just carry on?
The Missing Shared Context
That’s when I realized the missing piece was Shared Context.
Over the past few weeks, my team and I have been trying to build it into our actual work. We’re not trying to make another chatbot that knows a little bit about everything. We’re trying to sort out what’s scattered across chats, meetings, and documents, so the next person, or the next agent, can keep going.
So I want to write down what happened along the way. We don’t have a standard answer yet. But the gap is getting hard to ignore: AI is making each of us faster. Add up all that speed, and you still may not get a faster team.
These are my opinions. My team and I are building this together. I’ll write about what worked, and what didn’t.
简单说： 我现在干活的确快了不少。Agent 一开，很多事情几小时就能往前拱一大截。可一回到团队里，怎么还是这么费劲？
一个人干，的确快了
最近我干活越来越快，但跟团队一起干，还是该解释的解释，该对齐的对齐，有时候甚至比以前更麻烦。
我跟 agent 几个小时做出来的东西，一拿回团队，速度立刻掉了一半。
不是东西做得不好，是中间那些来来回回，只有我和 agent 知道。
我原来以为这是 Knowledge Base 的问题。文档放到一起，搜索做好一点，再让 AI 每天总结一下，应该就差不多了。
做起来才发现，不太对。
现在到底什么算数
团队缺的不是“大家说过什么”。真正麻烦的是：现在到底什么算数。
到底哪个决定算数，谁真接了这件事，换个人或者换个 agent，还能不能接着干？
缺的是 Shared Context
我后来才意识到，我们缺的这块东西，其实就是 Shared Context。
这几周，我和团队一直在试着把它做进真实工作里。我们不是想再做一个什么都知道一点的 chatbot。我们想把那些散在聊天、会议和文档里的东西理清楚，让人和 agent 都能接着往下干。
所以我想把这段过程记下来。我们当然还没有什么标准答案，但这个 gap 已经越来越难忽略：AI 的确让每个人都快了，这些速度加起来，却未必是一支更快的团队。
这里面的判断是我的，东西是我和团队一起做的。成功的写，没做成的也写。