After a $3 billion valuation, Replit CEO's judgment: Which will lose momentum first, SaaS, apps, or code platforms?
On September 10, 2025, Replit, a startup in the field of AI programming, announced the completion of a $250 million financing round, with a valuation reaching $3 billion.
Right after the official announcement of this financing, Google's DevOps Research department (DORA) released the results of a survey covering 5,000 technology practitioners globally: 90% of software engineers worldwide are already using AI programming tools in their daily work.
(Source: CNN, news report on September 23, 2025)
Within large companies, this proportion is even close to 100%. Google's team has fully switched to Gemini Code Assist, while Microsoft completed the full integration of Copilot within a year.
The rapid popularization of AI programming tools has led to an emerging phenomenon: the traditional process of "writing code → testing → deploying → maintaining" is undergoing fundamental changes, while the existing development ecosystem struggles to keep up with this pace of change.
In a recent conversation, Amjad Masad, the CEO of Replit, made a core judgment:
"It's not just people who are losing speed; it's the old way of working."
In his view, the rise of AI programming tools has exposed three fundamental problems in the existing development ecosystem:
- The over - fragmentation of SaaS platforms fails to support automated processes;
- The interaction mode of apps interrupts continuous execution;
- Code platforms focus more on writing than on deployment, resulting in programs that cannot be put into operation online.
The existing development ecosystem is designed according to human work habits, but AI Agents don't click buttons or switch pages. They require a completely different working environment.
Next, we will conduct an analysis around these three points:
Where exactly are SaaS, apps, and code platforms slowing down? And how should products and organizations be restructured?
Section 1 | Where did the platform start to lose speed?
Many companies today claim, "We've adopted AI."
They believe that with AI, processes are faster, outputs are greater, documents are written automatically, and code is auto - completed. However, Amjad Masad, the CEO of Replit, argues that it's not about being more efficient with AI; rather, the original platform logic is starting to fail.
"SaaS tools are still functioning, but they can't keep up with the current work rhythm."
He believes that the traditional way of operating software divides work into a bunch of independent tools:
- Writing documents is one product;
- Conducting analysis is another product;
- Collaboration, data transfer, and deployment are three separate products.
Users are forced to switch back and forth between different tools, often just to complete one task. AI is breaking this fragmented model.
In Replit's architecture, a piece of code can be directly run, deployed, generate an API, and then be called by other intelligent agents. This is no longer just traditional code - writing; it's a complete delivery process.
He proposed a core concept: the end - goal of a platform is not to write faster, but to make the results tangible.
If the program you write can only run on your local machine, it's not part of the platform. Replit aims to enable anyone in the world to write, run, and deploy a complete system using just a browser.
That's why Replit didn't choose the path of having AI generate a bit of code for you. Instead, it bet on "full - stack capabilities" from the start:
- Automatic deployment: Once written, it can go live;
- Callability: The generated program is not static; it can be invoked and executed by agents;
- Execution feedback: You can see the results of each piece of logic and optimize it iteratively.
This set of capabilities is not just for programmers; it's for future AI users.
Amjad said:
"In the future, systems won't require you to combine a bunch of tools to complete a process. Instead, there will be an intelligent system where you can assign tasks. You tell it what you want, and it can write, execute, and provide feedback on its own."
This is not just a platform upgrade; it's a subversion of the original way of working.
You no longer revolve around tools; instead, tools start to revolve around you.
Section 2 | What is replacing App applications?
In the past decade, every entrepreneur has been trying to develop an app. Whoever creates a user - friendly app can retain users, secure financing, and achieve sustainable growth.
However, in the AI era, Amjad Masad has observed a significant change: For many apps, even before they are launched, users' operating habits have already changed.
Now, when you open an app, you have to click, fill in forms, confirm, and switch windows. Soon, people will think, "Why don't I just tell AI directly?"
He gave a very specific usage scenario: On the Replit platform, many users initially planned to develop an AI app. But as they progressed, they realized that the real effectiveness lies not in the interface but in the underlying automated processing flow.
- So, users began to skip the interface and let AI directly complete a whole set of tasks:
- Instead of creating a form for users to fill out, they directly extract intentions from user conversations;
- Instead of building a reporting tool, they let AI analyze results, write summaries, and provide suggestions;
Instead of creating a reminder app, they let agents automatically perform operations at the right time.
Amjad clearly stated:
"Apps are passive, while agents are active. You don't need to go in and click buttons; they can do things on their own and come to you."
You can understand this as a reverse usage logic:
Apps require you to click around;
Agents will come to you and say, "I've already processed it for you. Do you want to execute it now?"
Amjad gave an example of Replit: The team originally planned to develop a "document - organizing app" that allowed users to upload, classify, and mark files. However, in actual use, they found that it was more effective to let AI agents automatically scan the working directory, identify important documents, automatically archive them by project, and proactively push them to relevant colleagues when needed.
The entire process requires no human intervention, and an agent can complete the entire process from document recognition to intelligent push.
Apps require you to operate, while agents do things for you.
Section 3 | The dilemma of code platforms: Running code alone cannot form a closed - loop
Amjad Masad's entrepreneurial experience is quite unique.
When he was young, he opened an internet café in Jordan, and his first company developed management software for internet cafés. He said, "I thought I could finish it quickly, but it took almost a year just for deployment."
This process made him realize a fundamental problem:
"Being able to write code doesn't mean being able to create a product."
Many traditional code platforms do help people write code faster, but they get stuck when it comes to making the product actually run.
He said, "These traditional code platforms make it easier to write programs, but deploying them, making them go live, and having others use them is still very difficult, just like building a half - finished road that ends abruptly in the middle."
So, what Replit aims to do is not just to create another code - writing platform, but:
- Run the code as soon as it's written;
- Deploy it online after running;
- Allow others to call it after deployment;
- Receive feedback based on the calls.
This is what he calls the closed - loop capability.
If you write a small program that can only run on your local machine, it's not enough. You need to make it accessible to others to realize its value.
Replit's approach is to connect every step of "writing → running → using" into a continuous process.
This process is not only for professional programmers. Amjad especially emphasized:
What we're doing is enabling more people, not just more professional ones, to turn an idea into something usable.
Take a real - life example: A Replit user can write a small game in the browser without downloading anything. With a single click, the game can run. AI will help with the deployment and generate a shareable link. Others can open it on their phones and play directly.
At this time, the user doesn't need to know what a server is or understand the deployment process.
Amjad said, "The real problem that AI programming tools need to solve is not to help you write a piece of code, but to make your ideas come to life, be used by others, and be validated by the market."
Many traditional platforms often focus on the code itself, while Replit is more concerned about:
"Whether the code can solve practical problems."
This actually reveals a core change: The focus of programming is no longer on the technical threshold of writing code, but on connecting the complete product chain.
Section 4 | It's not about hiring more people; it's about using AI to do the work
In this conversation, Amjad Masad said:
"Full - time work is a bug in the development of the social system."
It may sound subversive, but it's a trend he observed: More and more tasks no longer need to be done by humans but can be automatically processed by AI tools.
In Replit's own team, AI is no longer just an assistant to engineers; it has become a core part of the work process.
For example, when the sales team needs to analyze customer call records, in the past, they might have to organize audio, classify it, and create tables. Now, they simply input the data into an AI script:
- The AI automatically extracts the most frequently asked questions from customers;
- It outputs a one - sentence summary;
- It also generates suggestions and automatically sends them to the product team.
This is not just about speeding up the original process; it skips several intermediate steps.
Amjad's evaluation is:
"Our sales team is not using tools; they are commanding AI to do the work for them."
This statement is very crucial.
In a traditional company, a requirement usually has to go through several departments and colleagues from proposal to solution. Now, at Replit, one person plus a group of AI agents can complete the whole thing:
Identify problems, collect data, and conduct automatic analysis;
Output suggestions, feed back to the product, and form an iteration.
This is what Amjad calls the "AI factory":
In the future, companies will not rely on a large number of employees to operate but on a set of intelligent processes that can be reused.
He further explained, "AI tools are taking on more judgment and execution tasks that were originally done by humans."
Therefore, the most important ability for future organizations is not whether they can use AI, but whether they can design workflows around AI: Who is responsible for proposing requirements? Who starts the process? Who decides where the results go? Who confirms the completion of tasks?
Replit's organizational model is no longer about who manages what, but about which link is completed by invoking AI.
This also means that traditional job settings may be disrupted and replaced by:
- More flexible work arrangements;
- AI takes on the main work;
- Humans are mainly responsible for setting goals and verifying results.
Amjad summarized:
"The key is not to have more people, but to use AI effectively. It's not about hiring more people, but using fewer people and letting AI do more work."
This is not just an improvement in management methods; it's a complete change in the company's architecture.
The popularization of AI tools not only changes who does the work but also how the work is done.
Conclusion | The next - generation platform is rewriting the delivery method
From platforms, products, to company architectures,
Amjad Masad has presented a completely different perspective:
It's not about making engineers more efficient; it's about redefining who can create, what to create, and how to make it useful.
His core view is that the starting point for future software is not to develop an app or build a website, but to generate an intelligent system that can schedule tasks, trigger itself, and be repeatedly invoked.
This is not about creating tools; it's about building a "factory."
It doesn't just let you write code; it allows you to directly create something usable;
It's not only for professional developers but also enables everyone to have their own team of intelligent assistants.
The next - generation platform is neither SaaS, nor an app, nor a code IDE; it's an AI work assistant. You tell it what to do, and it can complete the task for you.
Which will lose speed first: SaaS, apps, or code platforms?
The answer is not about disappearing due to losing speed but about transformation.
Those who are truly ahead no longer care whether they can write code;
They only care about:
Whether there is a set of tools that can take on tasks and return results.
📮Original article links
https://www.youtube.com/watch?v=T9rukIoXxgw&t=2141s
https://edition.cnn.com/2025/09/23/tech/google-study-90-percent-tech-jobs-ai
https://podcasts.apple.com/us/podcast/moonshots-with-peter-diamandis/id1648228034?utm_source=chatgpt.com
https://blog.google/technology/developers/dora-report-2025/?utm_source=chatgpt.com
https://techcrunch.com/2025/09/10/replit-hits-3b-valuation-on-150m-annualized-revenue/?utm_source=chatgpt.com
https://www.reuters.com/business/ai-software-developer-replit-raises-250-million-3-billion-valuation-2025-09-10/?utm_source=chatgpt.com
This article is from the WeChat official account "AI Deep Researcher". Author: AI Deep Researcher. Published by 36Kr with permission.