Suddenly Everyone Is a 'Builder,' Whatever That Means --- Creating apps and websites used to be the rarefied world of tech developers. No more.

“SAN FRANCISCO -- During the tech boom of the 2010s, coding teams here spent months building apps that changed the way we live.

 

Now, that's just a Saturday night for an 11-year-old.

 

San Francisco has long been a city of builders, a label once largely reserved for developers in the tech industry. But thanks to advancements in artificial intelligence, a few well-crafted prompts in vibe-coding programs like Claude Code, Replit, DeepSeek V3/R1 (leading open-weight model), Tongyi Lingma (Alibaba’s AI coder), and Baidu Comate  are all you need to create your own apps, websites and AI agents that can automate aspects of your workflow.

 

Poof! Now everyone's a builder: marketers, product managers, people working in the construction industry, tweens. The term is appearing in profiles all over LinkedIn and in social-media posts. Claude and Replit have seen their global monthly average user counts surge.

 

"Every single person and their mama is a builder," says Willis Clayton-Stankowski, 25, who lives in Oakland. "I'm still not entirely sure what that means."

 

Clayton-Stankowski was until recently a project engineer at a Bay Area construction company. Surrounded by friends who work in tech -- many using automated coding programs -- he started thinking about what he could automate at work. He also wondered why he was paying $10 a month for a habit-tracking app when he could build one himself. So he did.

 

He was hooked. Soon he was identifying ways to save time and money using AI at the office. He even pitched a new role for himself: AI engineer. He started last month.

 

The building bug has been catching on across the U.S. but San Francisco is, unsurprisingly, the heart of the boom. Talk to enough builders here and a journalist might start to wonder if she should be marketing herself not as a writer but as a builder of stories.

 

Madina Gbotoe, 35, has overlapped with engineers for years in her work as a product manager at a healthcare company. But she'd never coded before. After joining an organization called Women Defining AI, she started using OpenAI's ChatGPT and Anthropic's Claude Code.

 

Now, she has a freelance side business building websites; AI fields new inquiries and creates an account for viable clients. A Claude-based agent books her airline tickets and hotels, and she recently built another to help with her new German shepherd puppy.

 

"I consider myself a builder," she says. "From zero to building."

 

To build, as she sees it, is to create an AI solution for your work or your home life. When she's out mingling, Gbotoe likes to ask: Who is building? She says everyone raises their hands.

 

But when she digs deeper, around a fifth of them actually meet her definition. Some are using ChatGPT as a glorified search engine.

 

"That is not building," she says.

 

South of San Francisco, the children of Silicon Valley builders are taking up the family trade. Will Raybould, 11, learned how to use Claude Code from his dad, James, who spent 12 years at LinkedIn, including as a director of product. Will created both a videogame and an interactive geography website to teach the location of every country in the world, along with fun facts about each.

 

While Will can understand how using Claude Code might be daunting to a newbie, he assures those curious about building that it's not so hard. He recommends giving the program simple prompts at first; your skills will improve as you go.

 

"At times it will seem like it's going to be really annoying," he says. "But in the end it will work out."

 

Omar Maniya, a doctor operating several urgent-care practices in New Jersey, got into building after tinkering with ChatGPT for a year. One night he was captivated by a podcast about what was possible with automated code.

 

"It took a few weeks but I was obsessed," he says. "I'm in bed at midnight with my laptop building apps."

 

Building is still a nascent practice in the tri-state area. Maniya suspects that if he told anyone in his circles he was building, they'd think he was renovating his kitchen.

 

"People on the East Coast don't know what it is to be a builder," he says.

 

Those at his offices do.

 

Maniya built an agent with Replit that can log in to diagnostic platforms, pull patient test results and add them to their charts. He told his staff to suggest ideas they'd like to see come to fruition, then he and his partner built them. They've churned out 10 apps so far, and three more are coming.

 

Saving lives is great, he says, but there's a different gratification that comes from having an idea in your head and then seeing it come to life.

 

"I'm feeling this intoxication of what engineers have probably felt for decades," he says.” [1]

 

 

Yes, you can develop an AI project in Replit and then transfer it to your local machine for training.

 

The core of this process involves using version control to move your code out of the Replit cloud environment and setting up your local machine to run the project.

 

Here's how you can do it:

 

    Version Control: The most reliable way to move your project is by using GitHub for synchronization.

        Connect your Replit project to a GitHub repository.

        Clone that repository onto your local machine using a local IDE or your preferred command-line tools.

    Set up your Local Environment: Your local machine will need the necessary software and dependencies to run the code.

        Install the appropriate runtime (e.g., Node.js or Python) and project-specific packages.

        Replicate the environment configuration, including setting up local .env files for environment variables and secrets (like database credentials or API keys).

    Run and Train Locally: Once the environment is configured, you can run and train your AI model using your local machine's resources.

        You may need to debug a few errors related to environment differences during this process. Copy and paste any errors into new chats in your local IDE with an AI agent to help you debug.

 

Important Considerations:

 

    The "smarts" of Replit's own AI Agent are proprietary and run on their servers, so you can't run that specific tool locally.

    Training large AI models locally may require a powerful computer with significant resources (e.g., high-end GPUs).

 

Can you train AI you build with any Chinese AI model locally not using Chinese companies' servers?

 

Yes, you can train and run Chinese AI models (such as DeepSeek-R1 or Qwen) locally on your own hardware without using Chinese companies' servers. By downloading open-weights models from platforms like Hugging Face and running them via tools like Ollama or Jan.ai, you ensure data privacy, as the models operate completely offline and on-premise.

 

How to Run/Train Chinese Models Locally:

 

    Select a Model: Download open-weights Chinese models like DeepSeek-R1 (available in various sizes) or Qwen from Hugging Face.

    Local Setup: Use tools such as Ollama, Jan.ai, or LM Studio to run these models on your machine's GPU/CPU.

    Hardware Requirements: A powerful Nvidia GPU is ideal for training and running these models effectively.

    Data Privacy: By running these models locally, your data never leaves your environment, ensuring compliance and security.

 

While some Chinese companies train their latest models abroad to access specific hardware, you do not need to connect to their servers to utilize their open-weights models, which are often considered superior for local, open-source applications.

 

1. Suddenly Everyone Is a 'Builder,' Whatever That Means --- Creating apps and websites used to be the rarefied world of tech developers. No more. Bindley, Katherine.  Wall Street Journal, Eastern edition; New York, N.Y.. 20 Mar 2026: A1.