前言
这 AI,进化到什么时候是个头。
做 PPT、文字润色、优化 Prompt、做总结,随便抓一个工具来,给出的结果都不算差。
但是“写”只是入场券,我现在挑 AI 产品只看一件事:它要等我走到哪一步,才开始真正帮忙。
现在像 Claude Code、Codex 这类桌面 Agent,你给一个方向,它们就能自己去项目里找文件、读上下文、拆任务。不需要你把材料一份份喂进去。
但本质上,它们还是属于“发包模式”的工具。你要先意识到“这件事可以交给 AI”,再把它叫出来,告诉它大概要去哪儿看、希望最后做成什么样。
不过我最近试的 Vida,很有点不一样!它减少的,是每次使用 AI 前那段“重新交代背景”的过程。
你让它帮忙时,它能从当前场景、历史记录和正在发生的工作里,自己拿到一部分上下文。
换句话说,Vida 在尝试一件更难的事:让 AI 提前看懂工作现场,知道你想干嘛。
Vida
大家光看我这么说,可能会觉得有点云里雾里。没关系,一个工具到底有没有用,还是要放到具体场景里看。
我这几天测试的场景,基本都是些大家会遇到的小问题。它们不一定复杂,但很适合看清 Vida 到底不一样在哪里。
① 一条消息
先从一条很普通的微信开始。
当时我接到了一条同事的消息,是问活动页今天还能不能发的。
这种消息我相信很多人都见过。它不难回,但也不能随便回。
因为对方要的不是一句“我确认一下”,而是一个能拿去同步客户的说法。这里面要有当前进度,要有今天为什么不适合直接发,也要有一个更稳的下一步。
麻烦就在这里。
这些信息并不都在微信里。它们散在排期文档、昨晚的修改记录,还有团队群里几句零碎的进度同步里。
我当然能想起来,但每次都重新捡一遍、拼一遍,再组织成一条分寸合适的回复,确实挺消耗。
当我把这件事交给 Vida 后,是这样的 👇
不知道大伙看懂了没有,我全程只是双击 Option 呼出 Vida,让它帮我回复。没有再把排期粘贴给它,也没有把昨晚的修改记录重新讲一遍,更没有手动整理一段背景说明。
但它最后给出的草稿里,已经带上了这几个关键点:
这就是我觉得它和普通 Agent 不太一样的地方。
普通 AI 也能写回复,但你要先把故事讲给它。Vida 在这个场景里省掉的,正是讲故事前那一轮整理上下文。
② 一整天
如果说前面那条消息,展示的是 Vida 怎么在一个具体场景里接上上下文,那另一个让我觉得有意思的地方,是它对“一整天”的处理。
这个场景更日常。
一天结束的时候,你可能并不是完全不知道自己干了什么。只是工作被切得太碎了。上午看了几个文档,下午改了一轮页面,中间回了几波消息,又顺手处理了几个临时需求。
每件事你都记得一点,但要把它们整理成一份清楚的工作总结,还是得重新回想一遍。
这时,Vida 就能给出一张今日战报卡。
前面是临时拿上下文。它要知道这条消息背后发生了什么。
这里是持续记上下文。它知道这一天里你到底推进了哪些事。
所以它整理出来的不只是“今天工作很忙,完成了若干事项”。它能把一天里的关键输出、时间分布、今日关键词、明天待办放到同一张卡片里。
③ 更多现场
这几天用下来,我能明显感觉到,Vida 不是只想在某一个任务里做得更细。
它更像是在把“少交代一点背景”这件事,放到更多工作现场里。
比如帮你整理 Prompt,根据 JD 改简历,或者先给杂乱桌面生成整理预览。
这些场景单看都不大,但放在一起,就能看出它们背后是同一个方向。少让你从零解释,少让你从零整理。
写在最后
当然,主动不等于可以完全放手。
从我这几天的体验看,Vida 能够帮我把下一步先准备好,而且不会替我跳过判断。
这恰恰是我觉得最合理的地方。AI 可以更早进入工作现场,但最后那一下,还是应该留给人。
他们现在在做一个 100 SOTA use case challenge,想持续挑战 100 个真实工作场景。
比如回一条消息、总结一整天、写一个 Prompt、筛简历、整理乱掉的桌面。这些事都不大,但它们确实是很多人每天会遇到的场景。
如果你也想试一下,现在注册 Vida,还能免费领取 7 天 Pro 版 👇
官网:https://vida.app
我的建议是,就从一个你每天会重复处理的小场景开始,看它能不能少让你解释一次、少让你整理一遍。
本文发表于公众号【莫理】
关注我们,阅读更多精彩内容
▽▽▽
Introduction
At what point does this AI reach its limit?
For PPT creation, text refinement, prompt optimization, and summarization, any tool you grab will produce results that aren’t too bad.
But “writing” is just your ticket into the game. Now, when picking an AI product, I only care about one thing: where do I have to reach before it actually starts helping?
Nowadays, desktop agents like Claude Code and Codex can take a direction you provide and search for files in your project, read context, and break down tasks. You don’t need to feed them material piece by piece.
However, fundamentally, they still operate in a “job assignment” mode. You need to first recognize that “this can be handled by AI”, then call it over, and tell it approximately where to look and what you want in the end.
But Vida, which I tried recently, is quite different! It reduces the process of “re-explaining background” each time you use AI.
When you ask for help, it can already capture part of the context from the current situation, historical records, and ongoing work.
In other words, Vida is attempting something more difficult: allowing the AI to understand the work environment in advance so it knows what you’re trying to do.
Vida
People might find this a bit vague just from what I’ve said. That’s okay — the true value of a tool always lies in the specific scenario.
The scenarios I’ve been testing these past few days are all small issues that everyone encounters. They aren’t necessarily complex, but they clearly show how Vida stands out.
① A Single Message
Let’s start with a very ordinary WeChat message.
I received a message from a colleague asking if we could still push the campaign page today.
I’m sure many of you have seen this kind of message. It’s not hard to reply, but you can’t just reply casually.
The other person isn’t looking for a simple “I’ll confirm”, but rather a statement they can use to sync with clients. This requires the current progress, why it’s not suitable to push immediately today, and a more reliable next step.
The tricky part is that this information isn’t all contained in the WeChat chat. It’s scattered across scheduling documents, last night’s revision logs, and scattered progress updates in team groups.
Of course, I can remember it all, but having to reassemble everything, formulate a response with the right tone, and then reply — that’s quite draining.
Followed
Follow
Replay Share Like
Close
Watch more
More
Exit full screen
Mo Li Followed
Share video
, duration 00:28
0/0
00:00/00:28
Switch to landscape mode
Continue playing
Progress bar, 0%
Play
00:00
00:28
00:28
Speed
Full screen
Playback speed
0.5x 0.75x 1.0x 1.5x 2.0x
HD Smooth
Continue watching
After testing the buzzworthy Vida, I finally understood — agents can be this proactive
Watch more
Original
,
After testing the buzzworthy Vida, I finally understood — agents can be this proactive
Mo Li Followed
Share Like Read
Already synced to Moments Write your comment
Video details
Did everyone get what I meant? Throughout the process, I only double-clicked Option to bring up Vida and asked it to reply. I didn’t paste the schedule, re-explain last night’s revisions, or manually provide background information.
Yet, the draft it produced already included these key points:
Regular AI can write replies, but you have to tell it the story first. Vida saved me the step of gathering context before explaining anything.
② A Full Day
If the first example showed how Vida connects to context in a specific scenario, then another feature that impressed me is its handling of “a full day”.
This is even more everyday.
At the end of the day, you might not know *nothing* about what you did — but your work is fragmented: reading documents in the morning, editing pages in the afternoon, answering messages, handling last-minute requests.
You remember bits and pieces, but organizing them into a clear work summary usually requires a manual recap.
Here, Vida offers a “daily battle report” card.
Followed
Follow
Replay Share Like
Close
Watch more
More
Exit full screen
Mo Li Followed
Share video
, duration 00:10
0/0
00:00/00:10
Switch to landscape mode
Continue playing
Progress bar, 0%
Play
00:00
00:10
00:10
Speed
Full screen
Playback speed
0.5x 0.75x 1.0x 1.5x 2.0x
HD Smooth
Continue watching
After testing the buzzworthy Vida, I finally understood — agents can be this proactive
Watch more
Original
,
After testing the buzzworthy Vida, I finally understood — agents can be this proactive
Mo Li Followed
Share Like Read
Already synced to Moments Write your comment
Video details
So its output isn’t just “I was busy today”. It breaks down key outputs, time distribution, daily keywords, and tomorrow’s to-do list — all on one card.
③ More Fieldwork
These days, I’ve noticed that Vida isn’t trying to dominate every single task.
Instead, it seems to be applying the principle of “require less background explanation” across more common work scenarios.
For example: organizing prompts, tailoring resumes based on job descriptions, or generating previews for messy desktops.
Followed
Follow
Replay Share Like
Close
Watch more
More
Exit full screen
Mo Li Followed
Share video
, duration 00:11
0/0
00:00/00:11
Switch to landscape mode
Continue playing
Progress bar, 0%
Play
00:00
00:11
00:11
Speed
Full screen
Playback speed
0.5x 0.75x 1.0x 1.5x 2.0x
HD Smooth
Continue watching
After testing the buzzworthy Vida, I finally understood — agents can be this proactive
Watch more
Original
,
After testing the buzzworthy Vida, I finally understood — agents can be this proactive
Mo Li Followed
Share Like Read
Already synced to Moments Write your comment
Video details
Followed
Follow
Replay Share Like
Close
Watch more
More
Exit full screen
Mo Li Followed
Share video
, duration 00:11
0/0
00:00/00:11
Switch to landscape mode
Continue playing
Progress bar, 0%
Play
00:00
00:11
00:11
Speed
Full screen
Playback speed
0.5x 0.75x 1.0x 1.5x 2.0x
HD Smooth
Continue watching
After testing the buzzworthy Vida, I finally understood — agents can be this proactive
Watch more
Original
,
After testing the buzzworthy Vida, I finally understood — agents can be this proactive
Mo Li Followed
Share Like Read
Already synced to Moments Write your comment
Video details
These examples may seem small, but taken together, they show a clear trend: minimize the need for manual explanation or organization.
Final Thoughts
Of course, proactivity doesn’t mean you can abandon all control.
From my experience, Vida helps me prepare the next step without skipping the judgment phase.
That’s actually the most reasonable approach — the AI enters the scene early, but the final decision still rests with the user.
They are currently running a “100 SOTA use case challenge”, aiming to test 100 real-world work scenarios.
These include replying to messages, summarizing days, writing prompts, filtering resumes, and tidying desktops — all common, everyday tasks.
Official website: https://vida.app
My advice is to start with a scenario you repeat daily — see if it can reduce your need to explain or organize things from scratch.
官网:https://vida.app
My suggestion is to start with a small, daily-repeating scenario and see if it can reduce the need for explanation or re-sorting.
This article was published on the WeChat public account [莫理].
Follow us for more insightful content.
▽▽▽












