Qwen Goes Personal

"Do Something Human for Once"

The evil Mark Zuckerberg has made Personal Agent go mainstream.

Meta's new product, Muse, connects to email, calendar, and other services to help users schedule things, handle tedious tasks, and even remember personal details mentioned in conversation.

The user says one sentence, and the AI goes and finds the information it needs, calling tools on its own.

A week after launch, it shot straight to #1 on the App Store. What does that tell you? It tells you Americans all want to cosplay as a boss with an assistant. By extension, the China market has even bigger potential.

Now looking at China, the hottest thing in primary markets is Qijunyuan's Today.ai — they've mobilized everyone on Jike to flood the zone with press releases for the China market.

Then there's Qwen. All along, they've been focused on continuously integrating tools and services.

I used to think Qwen's app was piling on too many Agent features, that the bar was too high. Most people just chatting with AI is enough; throw in a video call with AI and you're unstoppable.

Turns out I was being shallow. Qwen had already laid groundwork for the "Personal Agent" mega-release ahead of time.

What is the "Personal Agent" version? It means the model can do anything for an individual, becoming a life/work assistant.

Damn, actually ordering milk tea was the right call — it's just that besides milk tea, you gotta do other stuff too: shopping, hailing rides, booking restaurants. That was the early release at the start of the year.

Now that Agent capabilities have leveled up, all these features Qwen had been accumulating can finally connect together better, helping users complete more complex tasks.

This is corroborated by the data too: "complex tasks on Qwen have grown 5x, user conversation rounds have reached 5-8 turns, and most questions can be answered better by invoking medium-to-long-term context."

Of course more features is good — the key is whether the AI can actually put these features to use without torturing the user.

Everyone's operating on the same logic: users should be able to solve their needs as conveniently as possible within one entry point. Chatbot, Agent — these are specific entry points that progress as model capabilities rise.

Qwen was a connector all along. So Qwen had been waiting at the Agent stage this whole time.

The more specific approach has two aspects. One is increasing Context, so the AI knows you better.

The other is orchestrating the backend functions well — once it knows what the user wants to do, it can actually get it done.

Useful Context is what changes the answers the AI gives you.

What work you do, what habits you have, how your health is, what you're trying to accomplish lately — all of these affect how the same question should be answered. Just remembering what you asked last time isn't enough.

Qwen has already integrated health, fitness, and other personal data with user authorization, and supports some users in connecting their investment holdings.

So the information it can draw on extends from a few lines in the chat box to real physical states and life circumstances. Beyond app, PC, and glasses, the entry point can extend to earbuds, smart cockpits — touching more everyday scenarios.

Say you're training for a half-marathon. The AI can arrange training based on continuously logged workout data, then adjust recommendations as conditions change. The goal hasn't changed, but how much you can run today, how well you've recovered — all of this affects how you should train tomorrow.

The hard part is that this information needs to be both separated and connected. Your health records can't get mixed up with your parents', but when arranging your training, it also has to account for your work and schedule.

So, same as always: there's no vertical Agent, only general Agent. Only a general Agent has a shot at becoming a "Personal Agent."

Specialized tools can be domain-specific. Human life can't be separated. Health, learning, work — these are just entry points.

AI ultimately serves the same person. It has to be able to cross these scenarios to have a chance at truly understanding the individual.

Qwen's advantage is inheriting from Quark: cloud storage, question banks, scanning, college entrance exam counseling tools, and more. Now these capabilities are being packaged as Skills that the Agent can call upon.

The longer a user uses Qwen, the more context accumulates, and the more the AI understands what this person is about. The same question — the recommendations it gives you, the functions it calls — should be different from someone else.

To sum up: from day one of Qwen's app launch last November, the story has been about how to put AI to personal use. That problem was once hard to solve. Now it looks solvable.

With Memory, proactivity, and context, you can finally move from ChatBot to Personal Agent.

(Cover image generated by ChatGPT; purely human-written text)

⬇️

Subscribe to our Substack: funeralai.substack.com