Tokenmaxxer: AI should cost as much as your rent
Add Axios as your preferred source to
see more of our stories on Google.
Sigrid Jin and Madison Mills at Web Summit in May 2026. Photo: Shauna Clinton/Web Summit via Sportsfile
If you're worried you're falling behind on AI, developer Sigrid Jin has some advice: use AI so much that your monthly bill rivals your rent.
Why it matters: Jin argues that tokenmaxxing, the practice of using as many AI tokens as possible, is the best way to understand the value of AI. He would know: Jin said he used 50 billion tokens in a single year.
What they're saying: The "majority of people" aren't experiencing how much AI has to offer because they're using free or $20 subscription plans, Jin told Axios on stage at the opening ceremony of Web Summit Vancouver, a major tech conference.
- Those using basic versions of AI are missing out on the "higher intelligence" available through $200 plans, which he said produce clearer returns on the initial investment.
- "If you want to know the future of AI, try to do tokenmaxxing," Jin said, adding that he tells friends "to try to spend [on] your AI pricing as much as you pay for your own house rent" to get "return on investment."
- That ROI can look like "running multiple businesses" or delegating common life tasks to agents.
Yes, but: At some point, revenue from AI will need to outpace the cost of tokens in order to justify all the spending. That's especially so for enterprise users.
- Companies are increasingly spending more on their AI bills than they do on the salaries of their human employees.
- Jin said there's no universal way to measure the cost effectiveness of AI — every company or individual uses the technology differently and needs to develop their own benchmark for measuring returns.
Flashback: Jin went viral on March 31 after he recreated Claude's codebase when Anthropic accidentally leaked its own source code. He rewrote it in Python to avoid a copyright takedown from the AI lab.
- The tokenmaxxing paid off: Jin created the fastest growing GitHub repository in history, called Claw Code.
- Since then, Jin has received job offers from several AI labs, but has decided to focus on personal projects instead. He plans to launch a startup next month.
The bottom line: The pressure to increase AI usage isn't ebbing any time soon.
- That has led to what some developers, including OpenAI co-founder Andrej Karpathy, call "AI psychosis" — an obsession with pushing AI to its limits.
- Jin has found genuine joy in his usage, describing code as a "public good." He says he doesn't want to solely rely on AI, preferring to treat agents more like human collaborators than machines.
- When asked if he feels pressure to use even more tokens, Jin said, "Yes, of course."
