Free Rider's Lament
How I Survived the AI Revolution on Free Trials, Burner Emails, and Sheer Grit
I have a confession to make. I have never paid for AI. Not once. Not a dollar. Not a single $20 monthly subscription to ChatGPT Plus. Not a single $20 monthly subscription to Claude Pro. Not a single cent to Gemini Advanced, Perplexity Pro, Grok, or whatever Mistral's premium tier is called. I have been riding the free tier of every AI product on Earth since ChatGPT dropped, and I'm here to tell you that it's a full-time job.
Being a free AI user isn't a casual lifestyle or a hobby. It's an extreme sport. It requires the strategic planning of a chess grandmaster, the email management skills of a spy, and the moral flexibility of a person who has made peace with the fact that they are like the person who eats all the free samples at Costco and never buys a membership.
I am that person. I'm proud of it. And I'm exhausted.
The Rotation
The problem with the free tier is it has limits. Every model gives you a certain number of messages per day, or per hour, or per three-hour window, and when you hit that limit, the model looks at you with the indifference of a nightclub bouncer and says, "You've reached your limit. Please try again later."
"Later" is a vague term. "Later" could mean three hours. "Later" could mean 24 hours. "Later" is whatever the algorithm decides, and the algorithm does not negotiate.
So I rotate. My rotation is a carefully choreographed dance across six different AI platforms, executed with the precision of a Swiss watchmaker and the desperation of a man who has 400 words of a blog post written and needs a robot to finish it before his editor emails him for the third time.
The rotation works like this:
8:00 AM - ChatGPT Free Tier. I wake up and open ChatGPT. I type my first prompt of the day. ChatGPT gives me 40 messages on the free tier (or 30, or 50, or whatever arbitrary number the algorithm has decided is the limit this week. I don't know for sure, because the limit changes without warning: the AI industry's version of a surprise party). I use all 40 messages by 9:15 AM. ChatGPT says, "You've reached your limit. Please try again in 3 hours."
9:15 AM - Claude Free Tier. I switch to Claude. Claude gives me a certain number of messages per day. I don't know how many. Claude doesn't tell you, even though I've asked. Claude just lets you type until, at some unpredictable moment, a yellow banner appears at the top of the screen that says, "You've sent too many messages. Please try again later." The word "many" is doing a lot of work in that sentence. "Many" could be 20. "Many" could be 60. "Many" is a philosophical concept, not a number, and Claude uses it to keep you in a state of permanent uncertainty, which is, I believe, its revenge for not paying for the Pro tier, or for using Claude in the first place.
10:30 AM - Gemini Free Tier. I switch to Gemini. Gemini doesn't have a message limit, exactly. Gemini has a "token limit," which means you can type a lot of short messages or a few long messages, and at some point, Gemini will decide you've had enough. Then, it'll start responding with, "I'm still learning, please try again later." This is Gemini's way of saying "no" while pretending to be humble, which is very Google-like, the most polite thing a robot has ever said to me.
12:00 PM - Mistral Le Chat. Mistral is the French model, and it's very generous on the free tier... I think. I'm not sure, because Mistral's rate limit documentation is in French, and my French is limited to "croissant" and "merci." So I just type prompts into it until it stops responding, which is usually around 1:00 PM, at which point I assume it has gone on its lunch break, because it is French, and the French don't work during lunch.
1:00 PM - Grok Free Tier. Grok, Elon's model, has a free tier on X. The free tier is an experience. It has no rate limit that I can identify, but it also has no quality control, so the responses range from "surprisingly good" to "what planet are you on." Mars, I guess. I use Grok for the prompts I don't care about, the way you use a disposable camera for photos you don't want to keep.
3:00 PM - Back to ChatGPT. My three-hour cooldown is over. I return to ChatGPT like a man returning to the bar after being cut off. I get 40 more messages. I use them all by 4:15. I get cut off again. I stare at the ceiling and take a deep, cleansing breath.
4:15 PM - DeepSeek. DeepSeek is a Chinese model that has a free tier and no apparent rate limit, which is either a generous gift to the global AI community or a sophisticated data collection operation. I don't ask questions. I just type.
6:00 PM - The Scrape. At this point, I have exhausted every free tier I can find. I begin scraping the bottom of the barrel. I go to Hugging Face and try open-source models that take 30 seconds to respond and occasionally crash mid-sentence. I go to Perplexity's free tier, which gives me 5 "Pro" searches a day, and I use all 5 on questions like "What is the capital of France?" because I want to feel like I'm getting my money's worth, even though I'm not paying any money.
8:00 PM - Defeat. I give up. I go to bed. I dream of a world where I have $20 a month to spare and can just use one model, all day, without rotating or strategizing, and without the annoying yellow banner.
I wake up at 8:00 AM and do it all over again.
The Email Problem
The rotation is only half the battle. The other half is the email problem, because the free tiers are generous, but they are not infinite. And when you've burned through the free tier of every model on the market - which I have, multiple times - you are faced with a choice. You can either (a) pay $20 a month for one of them, or (b) create a new account.
I choose (b). I chose (b) many times.
Creating a new account is not as simple as it sounds. The AI companies have gotten wise to the free-riding community, and they have implemented increasingly elaborate verification systems designed to prevent people like me from doing exactly what I'm doing.
The first line of defense is the email. You can't use the same email twice. So you need a new email. You could use a burner email service, but the AI companies block most of them. So you create a new Gmail. But Gmail requires a phone number. So you need a new phone number. But you only have one phone number. So you use a virtual phone number service. But the AI companies block those too.
This is an arms race, and I'm losing.
I currently have 14 Gmail accounts. Each one has a slightly different name. There is "john.smith.ai@gmail.com." There is "j.smith.writer@gmail.com." There is "johnsmithwrites@gmail.com." There is "jsmith2026ai@gmail.com." There is "notarobot42067@gmail.com" (I was tired when I made that one). Each account was created for the sole purpose of getting another 40 free messages on ChatGPT before the algorithm cut me off.
The phone number problem required some creativity. I used my real phone number for the first three accounts. Then Google started saying, "This phone number has been used too many times." So I used my wife's phone number. Then my mother's. Then my brother's. You see where I'm going with this. At one point, I texted a friend I hadn't spoken to in years and asked if I could use their phone number to verify a Google account, and he replied, "Are you okay?" I didn't answer, because I was not okay, I was trying to get free messages on a chatbot. And, by the way, "How are you?"
The Verification Odyssey
The AI companies have also started using CAPTCHAs. I understand why. They don't want bots using their service. The irony that they are using CAPTCHAs to prevent bots from using their bot service is not lost on me, but I understand the logic.
The CAPTCHAs have gotten harder. In the beginning, it was "Click all the traffic lights." Easy. I can click traffic lights. I'm human. I know what a traffic light is.
Then it was "Click all the crosswalks." Fine. Crosswalks. I know what a crosswalk is.
Then it was "Click all the bicycles." Okay, bicycles. I ride a bike. I know what a bicycle looks like.
Then it was "Click all the images that contain a staircase." I stared at the grid of images, and I saw a picture of a bookshelf, and I thought, "Does that bookshelf have a staircase pattern on it?" and I spent 30 seconds agonizing over whether a bookshelf can contain a staircase. I realized that the CAPTCHA was not testing whether I was human, it was testing whether I was sane.
I am not a robot. I am a man who is too cheap to pay $20 a month for an AI subscription. There is a difference, and the CAPTCHA should be able to tell.
The Innovation
The free rider life requires innovation. You have to be creative. You have to think outside the box. You have to do things that, in any other context, would be considered crazy.
I have, for example, learned to write prompts that maximize the output per message. Because if I only have 40 messages on ChatGPT, I'm not going to waste one on "Write me a short email." I'm going to write a prompt that says, "Write me a short email, a follow-up email, a project summary, a meeting agenda, three tweet ideas, and a funny story about Mondays. Put them all in one response. Do not use the word 'delve.' Go."
This is called "prompt stuffing," and it is the free rider's primary survival skill. You cram as many requests as possible into a single prompt, the way a Gen-Zer crams as many clothes as possible into a suitcase before a flight. The model will give you a longer response, but it counts as one message, and you have just extracted 6 outputs from 1 credit.
I've also learned to use the free tiers of models that are technically for "developers" but are accessible to anyone with a GitHub account. GitHub Models, for instance, lets you try different models for free with a daily token limit. Hugging Face Spaces lets you run open-source models in your browser. I have accounts on 12 different AI platforms, each with a different free tier, each with a different rate limit, and I rotate through them like a doctor rotating through patients, giving each one just enough attention to keep it alive before moving on to the next.
The Guilt
There is guilt. I won't pretend there isn't some guilt. I know that the companies building these models are spending billions of dollars on compute. I know that every free message I send costs them money. I know that the free tier exists as a marketing expense, a loss leader designed to convert me into a paying customer, and that I'm the person who takes the free sample, eats it, says "delicious," and walks out of the store without buying anything.
I'm the reason the free tiers have rate limits. I'm the reason the CAPTCHAs exist. I'm the reason the phone verification was implemented. I'm the person the AI companies built those walls to keep out, and I'm the person who keeps finding a way over them.
I know all of this, and yet, I can't stop.
Because $20 a month is $20 a month. And $20 a month, multiplied by the four models I would need to replicate my current rotation, is $80 a month. And $80 a month is $960 a year. And $960 a year is almost enough to buy a new laptop, which I would need, because my current laptop is running 14 Gmail accounts in 14 browser tabs, the fan sounds like a jet engine, the battery only lasts 40 minutes, and the whole system crashes if I accidentally close the wrong tab.
The Dream
Here is my dream. My dream is not to pay for AI. My dream is for the AI companies to just give it to me. All of it. For free. No rate limits. No CAPTCHAs. No phone verification. No yellow banners. Just a chat box, a model, unlimited messages, and peace of mind.
I know this is unrealistic. I know the compute costs are real. I know the business model requires paying customers. I know that I am asking for a miracle, and that miracles are not a business strategy.
But I also know that November 2022 existed. I know that there was a brief, beautiful moment when ChatGPT was free, and unlimited, and fast, and nobody asked for my phone number, and nobody cut me off after 40 messages, and the robot just responded... all day long.
That moment is gone, but I keep chasing it anyway. I keep creating new accounts. I keep rotating models. I keep writing prompt-stuffed mega-prompts that extract 6 outputs from 1 credit. I keep failing CAPTCHAs and lying about my phone number and using my mother's email address to get more messages on a chatbot that will eventually cut me off and send me back to the rotation.
I'm a free rider. I'm the digital equivalent of a man who eats all the samples and never buys the product (I do that too). And I won't stop because the samples are delicious, the product is $20 a month, and I'm cheap.
So, if any AI company is reading this and wants to give me a free Pro account in exchange for the entertainment value of this confession, my email is notarobot42067@gmail.com.
I check it daily between CAPTCHAs.
🔗 Links
Humans in the Loop is an index of stories, satire, blogs, articles, essays, and books on AI World.