[Actual Test] MiniMax Agent has covered front-end + back-end + model + UI all at once, and I’m completely stunned

I remember when I just recommended Bolt.new to my colleague last week, we were still sighing: This thing can really replace most front-end developers. And now, I have to change my tune:

The MiniMax Agent directly eliminates the entire “full stack development”.

Yes, no exaggeration—I’ve actually tested two requirements, and it **can not only write front-end code, but also interface with databases, deploy back-end logic, integrate large models, automatically generate UI pages, and even call its own model to generate audio & video.** It forms a complete closed loop, as if collaborating with a silent, efficient, and technically comprehensive developer.

💩 Project 1: AI helped me generate, deploy, and dub my son’s toilet picture book in one go!

Our family has recently been training our son to use the toilet independently, but the progress is worrying.

Coincidentally, he really likes the picture book series “Posy and Pip”. A sudden inspiration struck me:Could I ask AI to generate a “toilet-themed” picture book in this style to make it easier for him to accept?

So I threw this requirement (the prompt is very simple) to the MiniMax Agent:

Referring to “Posy and Pip”, create a picture book suitable for toilet training of 2-3 year-old babies, with Posy and Pip as the main characters; the number of images should not exceed 20; each page is accompanied by illustrations and supports click-to-read, with a gentle motherly voice, and will ultimately be presented in web page form.

After the Agent and I confirmed the requirements, we immediately got to work, and the final results are as follows👇:

👉 Preview URL (recommended to open on mobile for better experience):https://3bjg9crjj1a2.space.minimax.io

• The content is complete, full of childlike fun, and logically coherent
• Each page is illustrated with a unified style and an ultra-gentle painting style
• The point-and-read and voice-over function has also been implemented, and a single click can play the female voice narration
• The web version has been directly deployed and is ready to use upon opening!

⚠️ It is worth emphasizing that this is not a static page, but a fully functional delivery with dynamic generation, deployment, interface connectivity, and comprehensive supporting features .

What shocked me the most was that:Originally, I thought it would require two steps of development – first creating the text and images, then adding voiceover. But as it turned out, the Agent took care of everything in one go and directly presented a “usable product”.

🧑🏻💼 Project 2: Workplace Narrative Generator Mini Program, completed “20 person-days” of work in 2 hours

The second item is something that I, as a Product Manager, have a strong personal need for:

A Mini Program that can automatically generate workplace communication narratives, helping workers handle those “tedious but necessary” conversations.

For example, notifying about team building, politely declining invitations, thanking colleagues, accepting offers, etc., all need to be written every day, but we’re too lazy to repeatedly deliberate on the tone.

My prompt is as follows:

Create a website with backend capabilities, with the following features:

1. After the user inputs requirements and clicks Generate, a narrative that fits the workplace context will be output (calling the large model)
2. Optional tones: formal, relaxed, humorous, sincere
3. Supports text & voice input
4. Save historical records for each generation (including the original input + output narrative)
5. Mobile End Adaptation
6. The UI style is simple and consistent with the usage habits of office workers.
7. There is a menu to view historical records and generation time

Preview URL:https://ehm7dqnj7n7c.space.minimax.io/

I thought this requirement might only allow Agent to generate a front-end template, but unexpectedly it did everything step by step for me:

UI page is set up, and interaction is smooth

The backend integrates Supabase, and data can be saved

Automatic API interface docking (this time I used Deepseek’s API, which has free quota on Volcengine)

Tone adjustment + voice input module are both implemented

It also allows users to view historical records, and the page style is very “worker-friendly”.

Among them, the only two steps I took were authorizing Supabase (it is recommended to create a new project) and filling in my Deepseek API. For the remaining processes, they are fully automated .

I sent the finished product to our technical colleagues for review, and their reactions were as follows👇:

Normal development takes 20 person-days. And I myself only took 2 hours.

🤖 The logical ability of MiniMax Agent really scares me a bit

It is not like GPT, which simply “writes as you say”, but an Agent that fully understands the requirements, proactively connects the whole-link, and executes .

The strongest part is not generation, but “the ability to understand and act”.

For example, when I mention “save history”, it doesn’t just leave a placeholder for you; instead, it actually implements complete data writing, querying, and display functions. When I write “mobile end adaptation”, it truly optimizes mobile end interaction, page layout, and button size.

🧠 When comparing with traditional AI tool flows, you will find the differences are very obvious