Slow LLM: Deliberately Slows AI Chatbots for Reflection

Imagine typing a question into ChatGPT and waiting over 40 seconds for a response. Frustrating? Absolutely. But for artist Sam Lavigne, this is the point. His web tool, Slow LLM, intentionally slows down AI chatbots like ChatGPT and Claude to make users rethink their reliance on automation. In a world where AI answers everything from homework questions to grocery lists, Lavigne’s project challenges us to pause and ask: Are we outsourcing too much to machines?

How Slow LLM Works

Slow LLM operates by modifying JavaScript’s fetch function in browsers—a core component chatbots use to communicate with servers. While the AI model itself processes data at normal speed, the tool delays the delivery of responses to users. From your perspective, the chatbot feels agonizingly slow, forcing you to wait for answers that would otherwise appear instantly.

Lavigne offers two versions of the tool:

• Chrome Extension: Users install the extension to slow chatbots on their own devices.
• Enterprise Edition: Network administrators can deploy it via DNS settings, slowing AI interactions for everyone on a shared network.

Why Slow AI Matters

Lavigne calls the rise of AI chatbots a “massive de-skilling event.” When we rely on AI for basic tasks, we risk losing the ability to solve problems independently. For example:

• Students might skip learning math formulas if they can just ask an AI.
• Professionals might outsource critical thinking to chatbots instead of developing expertise.

By making AI interactions inconvenient, Slow LLM nudges users to reflect on their habits. It’s not about rejecting AI but fostering a healthier relationship with it.

Practical Use Cases for Slow LLM

While the tool may seem extreme, it has real-world applications:

1. Education: Teachers can use it to encourage students to think through problems rather than relying on AI.
2. Workplace: Companies might deploy the Enterprise Edition to reduce over-dependence on chatbots for simple tasks.
3. Personal Growth: Individuals can test their ability to solve problems without AI assistance.

Open Source and Security

Lavigne has open-sourced the code for Slow LLM, allowing developers to audit and customize it. For security-conscious users, he recommends hosting the DNS component locally rather than trusting third-party servers. This transparency aligns with the project’s goal of empowering users to take control of their digital tools.

Try Slow LLM—and Reflect

If you’re curious about your AI dependency, give Slow LLM a try. Install the Chrome extension or explore the Enterprise Edition. The frustration might just lead to a valuable realization: Sometimes, slowing down is the best way to move forward.

Ready to rethink your AI habits? Visit the Slow LLM website and experience the intentional lag for yourself. Your next question might take a while to answer—but the insight could be worth the wait.