Zac

20年企业战略经验，关注AI时代的战略与实践

[Zac @AnBangli]

AI时代，企业凭什么被顾客选择？

我是李湘群，原特劳特（中国）管理合伙人。

这里讨论三个问题：顾客为何改变选择，企业应集中什么、放弃什么，AI能否改善真实业务结果。

从具体场景出发，讲清判断依据，也讨论适用边界。

欢迎带着真实问题交流。

[Zac @AnBangli]

Meta's new Muse Wonder models give software teams another option for complex coding work.

For a team considering a switch, I'd start with five real tasks its current agent leaves unfinished. Does Wonder get the changes ready to merge with less review and rework? Is the total cost worth it?

One buying detail matters: Wonder is in API preview. The more capable Wonder Full is only available through certain Muse Code subscriptions. Setting API reasoning to max doesn't unlock it.

My take: strong benchmark results earn Meta a trial. A team stuck on cross-file fixes needs a more specific reason to switch: fewer hours of senior engineers stepping in to finish the job.

That's the promise I'd want Meta to prove.
https://t.co/1WcuefW6F0

[Zac @AnBangli]

@cpeng1154988 数据和接口还在原应用，日常检索已经交给 workbuddy。下次续费时，你还愿意付原来的价格吗？这会很直接地检验它剩下的价值。

[Zac @AnBangli]

@finkd A year of labs, two model families, one verb: arrived.
"Brings us to the frontier" — not "on our way".
Announcing arrival is the most expensive step. The "still catching up" excuse is gone.
From now on, every miss is a frontier-level miss.

[Zac @AnBangli]

@alexandr_wang Rivals, not beats. The cheapest way to change seats.
You don't need to prove first. Just get your name on the same list.
The tax on the new seat: every Muse model is now judged head-on against GPT and Claude.

[Zac @AnBangli]

@theo Everyone fights over benchmark scores. Theo grabbed the referee seat instead.
Ranking is cheap. Drawing the Gap is expensive — that line sentences the whole second tier to death.
The price of the referee seat: one biased call, and you lose it forever.

[Zac @AnBangli]

@sandislonjsak The client with no internet presence caught my eye. What job was the prototype doing for him, and how was he handling it before?

[Zac @AnBangli]

@sandislonjsak A changed departure gate is a useful interruption. Five new hotel suggestions after the trip is booked can just create more work. Knowing which decisions are already settled seems like a big part of getting this right.

[Zac @AnBangli]

@EdenBitton7 @lennysan @joshwoodward @Google I'd save the failed input too. For an invoice tool, that might be the document it assigned to the wrong supplier despite getting the total right. Then the next model has a concrete test waiting for it.

[Zac @AnBangli]

@t_blom If buyers start asking 100 sellers instead of three, enquiry volumes become a much weaker guide to demand. That seems important for businesses using them to plan stock or hiring.

[Zac @AnBangli]

@justhalfbit @icanberk 对准就弹出这台空调的控制条，这个思路挺好。左右摇倒可以跟拇指滑动比一比，躺在沙发上哪种更省劲。能自动认准是哪台空调的话，就真省掉翻 App 找设备那几步了。

[Zac @AnBangli]

今天试了 ChatGPT 的 Intelligent UI。问它对比两个方案预算，它没给文字，直接递过来一张能拖改数字的小表。

以前是问完拿答案，再自己动手做一遍；现在答案直接变成能用的东西。

聊天框的终点，可能不是回答问题，是让答案变成工具。#AI

[Zac @AnBangli]

主要原因是风险管控：防止车辆被带到娄底后非法抵押、套现后难以追回。

娄底在国内抵押车（尤其是“断供车”“不过户抵押车”）圈子里非常出名，被称为“抵押车的天堂，清收队的地狱”。当地有成熟的交易网络，外地清收队想把车拖走往往阻力很大（宗族/村落关系、信息互通、本地保护意识等）。

租车公司禁去娄底的具体原因
高发的“租车抵押”诈骗
有人在长沙等地租车，然后开到娄底，用伪造证件把车拿去抵押/典当套现，拿到钱后失踪。这类案件在新闻和车圈里多次出现，租车公司因此损失车辆。
车辆一旦进入娄底抵押车市场，极难追回
车进去后容易被转手、藏匿，或者清收时遇到阻力，租车公司（或背后的融资方）很难把车要回来。

商业风险直接转嫁给租赁公司
车辆一旦被抵押出去，所有权纠纷、GPS被拆除/屏蔽、车架号被改等风险都会出现。保险公司也可能拒赔，公司面临车辆全损。
所以很多正规或半正规的租车公司（尤其是跨城、长租、豪车租赁）会在合同里明确写“禁止驶入娄底”或把娄底列为禁区，这是行业里针对高风险地区的常见风控做法。

简单说：不是娄底本身有什么官方禁令，而是当地抵押车灰色交易太活跃，租车公司怕车一进去就回不来。 如果你看到某家公司特别强调娄底，基本就是这个原因。

[Zac @AnBangli]

一家人真幸福

[China Open @ChinaOpen]

The Great Wall and #ChinaOpen trophy. 🏆
Keep going up. The view follows.

[Zac @AnBangli]

@harimotomiwa 太酷了了，祝贺美和兄妹

[Zac @AnBangli]

@WTA @ChinaOpen 女王登基

[Zac @AnBangli]

@WTA @ChinaOpen 女王登机

[Zac @AnBangli]

@laopai1990 你觉得主要差别在哪？是搜索本身更好用，还是入口和使用习惯的影响更大？

[Zac @AnBangli]

Google 发布 Gemini agent：干活的 agent，写代码、起文档、接公司系统，长在 Workspace、365、Slack 里。

它不绑自家模型，合适的任务切到 Anthropic 的 Claude 跑。

比的不是模型更聪明，是谁离工作流更近。

模型变零件后，更聪明还站得住吗？#AI