ByteDance has launched a new AI product named Trea

official website: https://www.trae.ai/

Ship Faster with Trae
Trae is an adaptive AI IDE that transforms how you work, collaborating with you to run faster.

ByteDance has launched a new AI product named Trea.

This is an IDE tool, which in Chinese means “Integrated Development Environment”. Technical readers should be quite familiar with this term.
Simply put, it is a workspace that programmers use to write code and develop software.
Trea is different from traditional IDEs. It is an AI-based development workspace that allows you to have AI write programs and build a runnable product through natural conversation.
As long as you have a product idea and can describe it completely, you can ask Trea to bring it to life.
For example, I simply wrote a prompt for generating a side hustle list application. Next, the AI will understand and break down my intention, and then construct its own tasks.

Next, the AI will automatically create a project library, install dependency libraries, and write code.

After the project is built, it supports real-time preview and will also help you summarize the specific implementation list for this requirement.

Finally, run it directly, and you can see a usable “side hustle list” application.

How about it? Don’t you think it’s very powerful?
Not only that, Trea can also automatically develop front-end pages based on the product prototype diagrams you upload, and it can also analyze and optimize existing code.
If you have a basic understanding of technical knowledge, you can completely use Trea to develop H5 applications, WeChat mini-programs, as well as Android and iOS versions of apps.
Overall, the experience with Trea has brought me some pleasant surprises.
First of all, it replicates the entire technical development process and completely displays every step of the operation in the conversation area, including every technical instruction.
As the operator, you only need to observe its operations and click confirm or run at the corresponding nodes to observe the entire workflow.
This process is very similar to the steps I used when writing programs myself back then, except that now it’s all done by AI.

Secondly, the code written by Trea is very standardized, and some of it even includes comments. Technical readers know that the constraints and implementation of code standardization have always been pain points in development.
In terms of code quality, it can already be compared to that written by a proficient programmer.
You might ask, can such an AI IDE tool be used in actual production development?
I think it’s still quite difficult to completely replace existing development tools, but it can already assist programmers in significantly improving their efficiency.
Take front-end development as an example. In the past, programmers needed to write code based on high-fidelity prototype diagrams. They first had to build the page structure, then write the interaction logic, and finally add the visual materials.
A seemingly simple page requires at least three steps of code work for a programmer.
If an AI product like Trea is used, you can directly upload the high-fidelity prototype diagram to it, and then obtain the written front-end code.
Not to mention that it can fully meet the requirements, but being able to restore about 70% of them is already a significant improvement in efficiency.
In addition to front-end product development, there are also back-end logic and database processing. Although Trea can also be integrated into other workflows and data processing, it still has difficulties in matching with existing actual production processes.
Therefore, it’s still very difficult for a product like Trea to completely replace existing development tools. It will at least take a long time.
However, this is a trend. With the enhancement of AI capabilities and the iteration of product development, there will be more and more products developed purely by AI in the future.
Just a few years ago, many product managers were learning programming languages like Python or PHP, because having some knowledge of programming technology was an advantage at that time.
Now, the situation has changed.
For product managers who don’t understand technology, products like Trea offer some new options, because you can complete the development of a product without writing any code at all.
For some simple applications, Trea can actually handle them completely.
A bit more advanced, you can implement the back-end logic using a workflow, and then apply for a cloud server and a database, and an online product is completed.
As long as you have an idea, AI may turn it into reality.
I introduced ByteDance’s Coze at the beginning of last year. It can complete the construction of workflow and can be used in conjunction with Trea.
It’s not just Trea. There are also products like Cursor and Windsurf abroad that can achieve the same functions.
Trea is built-in with the Claude-3.5-Sonnet and GPT-4o models. Currently, it’s free to use, and the key is that it supports Chinese, which is very user-friendly for domestic users.
I’ve been playing with Trea these two days. As a product manager with a programming background, I’m increasingly feeling that the boundary between these two roles is shrinking.
If this trend continues, the super product manager might really become a reality.
The so-called super product manager is someone who can handle the entire process of business, product, design, development, operation and maintenance, and customer service all by themselves.
With the assistance of AI, the entire process can be connected.
Once again, the primary driving force of the Internet is the iteration of technology, and product opportunities lie within it.
················· Produced by Tang Ren ·················
Encore Moment
I don’t know if you’ve noticed, but in the AI era, there is a very important ability: writing.
Words are the medium through which we interact with AI. Writing prompts (prompts) requires words. The more precise and clear the expression, the higher the quality of the results we can get from AI.
The same goes for using Trea. Whether you can accurately describe your needs and ideas through words directly determines the quality of the results given by AI.
Therefore, a lack of writing ability may become an obstacle to entering the AI era.