Tools are being used to slow down AI chatbots to a "snail's pace" in order to reduce dependence on them.

Slow LLM forces people to question their reliance on tools like Claude and ChatGPT by making their use extremely frustrating.

Báo Khoa học và Đời sống•26/03/2026

Witnessing people entrusting their critical thinking, emotions, and composure to flawed "AI" chatbots has become one of the most frightening aspects of modern life in recent years.

Wealthy "tech evangelists" like Sam Altman are constantly making bold claims about how large language models (LLMs) will replace our jobs and parenting.

Slow LLM will make chatbots run at a "snail's pace," creating a frustrating experience.

Critics, however, compare the reliance on chatbots to an attempt to convince people that they are actually better off letting machines think, act, and create for them.

Now, there's a new way to prevent friends, family, and even strangers from switching to chatbots like Claude and ChatGPT: use a tool called "Slow LLM" to make them extremely, extremely, extremely slow.

“Are you worried that you or a loved one is participating in a mass skill-loss event? Experiencing LLM-induced psychotic states? Resigning cognitive and emotional functions to auto-completion? Install SLOW LLM on your computer, or a loved one’s computer, today!”—a description on the tool’s website states.

Created by artist Sam Lavigne, Slow LLM causes anyone accessing an AI chatbot on a computer or network to experience mysteriously slow and frustrating response times.

SlowLLM was created because the creator was fed up with students and their relatives becoming increasingly dependent on AI chatbots.

This tool works by exploiting a feature of the Javascript language to rewrite the "Fetch" function that returns data to the browser. When a user accesses a chatbot domain and enters a question, the modified Fetch function will extend the response for an extremely long time.

As a result, users perceive the LLM as slow, when in reality it's simply Lavigne's code that's deliberately slowing it down.

Lavigne said the idea for the project came after witnessing some of his students and acquaintances becoming increasingly reliant on generative tools to perform basic tasks.

"Many people begin using these tools to neglect their cognitive and emotional functions, and in the process, they forget the basics they once knew how to do," Lavigne shared.

Slow LLM can be installed as a Chrome browser extension, but can also be deployed across an entire network via the “Enterprise Edition”. The DNS service will cause everyone on the home, school, or company network to experience slow chatbot responses.

This simply requires changing the DNS server on the router to Lavigne's custom domain name, which he has released for free on Github.

The browser extension currently only affects Claude and ChatGPT, while the DNS version also slows down Grok and Google Gemini.

In theory, someone could activate Slow LLM without anyone noticing. Most people would probably assume that chatbot providers like Google and OpenAI are experiencing technical issues, which occasionally happen without external intervention.

Lavigne said that so far, he hasn't heard of anyone successfully implementing Slow LLM on a company or school network. But he also doesn't discourage people from experimenting.

"I haven't tested it on any unknown subjects yet, but I'm considering it," Lavigne said playfully, suggesting it would be an interesting experiment to see how people react to an artificially slowed-down chatbot.

“They might get angry and stop using LLM.” Slow LLM is the latest addition to the series of quirky tech projects that Lavigne is known for.

During the Covid pandemic, he released "Zoom Escaper"—a tool that disrupted your Zoom audio with echoes, distortion, and annoying interruptions to the point where others couldn't stand your presence.

Sam Lavigne and his colleagues often create quirky tools to help people become wary of their dependence on machines.

In 2018, he caused controversy when he collected publicly available LinkedIn profile data to build a large database of ICE agents, which was later removed from platforms like Github and Medium.

Lavigne's frequent collaborator, Tega Brain, also released browser tools like "Slop Evader," which filtered out AI-generated content by removing all search results after November 2022, the date ChatGPT was first released to the public.

However, Lavigne wasn't an extremist. He admitted to using Claude to help write some of the code for Slow LLM until Slow LLM became operational and forced him to complete the project alone.

Instead, Lavigne said he wanted people to question the habits formed from the frequent use of chatbots, tools that make it easy for us to entrust all our knowledge, decisions, and emotions to giant corporations run by tech billionaires like Altman and Elon Musk.