At Anthropic, a single utilizer of the company’s AI coding system, Claude Code, racked up a bill of more than $150,000 in a month.
And at tech companies like Meta and Shopify, managers have started to factor AI utilize into performance reviews, rewarding workers who build heavy utilize of AI tools and chastening those who don’t.
This is the new reality for coders, some of the first white-collar workers to feel the effects of AI as it sweeps through the economy. AI was supposed to support tech companies boost productivity and cut costs. But it has also created an expensive new status game, known as “tokenmaxxing,” among AI-obsessed workers who are desperate to prove how productive they are.
At some tech companies, including Meta and OpenAI, employees compete on internal leaderboards that reveal how many tokens — the atomic unit of AI utilize, roughly equivalent to a word fragment — each worker consumes, two people familiar with those companies’ practices declared. Generous “token budobtains” are becoming a job perk for coders, like dental insurance or free lunch, and some are spconcludeing thousands of dollars a month testing to automate as much of their own work as possible.
“I probably spconclude more than my salary on Claude,” declared Max Linder, a software engineer in Stockholm. (Linder’s employer pays for his tokens.)
Until recently, power utilizers might have consumed thousands of tokens a day applying an AI tool like ChatGPT, Claude or Gemini. A student writing an esdeclare, for example, may go through 10,000 tokens — roughly equivalent to 7,500 words — including several rounds of revisions. Using millions of tokens would require hours in front of a computer, doing nothing but typing, and applying billions of tokens was virtually impossible.
Story continues below this ad
But the advent of so-called agentic coding tools has upped the ante. These systems can work unsupervised for hours at a time, reviewing and editing large code bases and writing entire software programs from a single prompt. Each agent can spawn subagents to handle different parts of a tquestion, generating thousands of tokens at each step. Some AI systems, like the popular open-source AI assistant OpenClaw, are designed to run 24/7, churning through tokens while their human utilizers sleep.
“If you have some continuously running agents, you’ll do 700 million tokens a week from a single full-time agent,” declared Ege Erdil, a co-founder of Mechanize, an AI startup, who estimated his own token consumption at between 1 billion and 10 billion a week. “It doesn’t really take that much.”
All of that adds up for the AI companies selling the tokens. Anthropic more than doubled its revenue projections in two months this year, largely becautilize of the breakneck growth of its agentic coding tools. OpenAI recently declared that its agentic coding tool, Codex, had tripled its weekly active utilizers since the start of the year, and that overall Codex utilize, measured in tokens, had increased fivefold. Last year, Google declared its AI models processed more than 1.3 quadrillion tokens a month.
Even for the most dedicated programmers, applying billions of tokens isn’t straightforward. For comparison: I went through a period of heavy Claude Code utilize earlier this year, working on several separate coding projects for four or five hours a day, and managed to utilize only a few million tokens. (Rookie numbers, really.) But some coders have mastered the art of AI multitquestioning, opening multiple windows and setting dozens of agents loose on their projects at a time.
Story continues below this ad
AI companies have encouraged these whales, giving them trophies and other rewards. And some tech executives are glad to see their employees embracing the new tools. They equate heavy AI utilize with increased productivity — if a programmer wants to operate a swarm of 10 AI agents, running parallel tquestions in separate windows, they’re happy to foot the bill.
But I spoke to several tech workers who worried that their colleagues are gorging on billions of tokens — which can cost thousands of dollars a day — for what amount to bragging rights. Even at the AI labs, where workers are given unlimited utilize of their companies’ tools, the idea that all of this is productive seems far-fetched.
“It doesn’t seem sustainable,” declared one OpenAI employee, who questioned to remain anonymous becautilize he was not authorized to discuss his colleagues’ AI coding addictions.
Subscribers to paid Claude and ChatGPT plans typically pay a monthly fee, which gives them a repaired number of tokens. (The number varies; some tokens are “cached,” meaning the system has stored them in memory and doesn’t required to generate them from scratch, and companies charge more for “output” tokens than “input” tokens.) Users who required more tokens can pay for them separately or upgrade to a more expensive plan.
Story continues below this ad
Shopify declared in a statement that token utilize is just one measure of how the company measures performance. It also views at how AI “improves and amplifies” work. Anthropic, Meta and OpenAI declined to comment for this column. (The New York Times has sued OpenAI and Microsoft, claiming copyright infringement of news content related to AI systems. The two companies have denied the suit’s claims.)
But power utilizers have learned how to game the system by stacking multiple subscriptions or taking advantage of promotional offers. One startup founder informed me that he had discovered an AI tool created by Figma, a design startup, that allowed him to utilize the equivalent of $70,000 in Claude tokens through an account that costs him $20 a month. The founder, who questioned to remain anonymous to avoid tipping off Figma, declared he had utilized the loophole to build six software projects at the same time.
A Figma spokesperson declared that the utilize “took place before AI credit enforcement went into effect” earlier this week.
I talked to several other tokenmaxxers about what they’re doing with all those tokens. Most were engineers or hobby programmers who were building and maintaining large, complex pieces of software applying coding agents running in parallel.
Story continues below this ad
They declared, by and large, that AI coding tools were creating them more productive. But some also framed their utilize of AI as a strategic shift — a way to signal, to their colleagues and bosses, that they’re keeping up with the times, as the era of human coding appears to be coming to an conclude.
Nikunj Kothari, a venture capitalist in San Francisco, wrote in a recent Substack post about the rise of what he called “token anxiety.” He described a tech scene that has become obsessed with productivity — AI productivity, not human productivity — and declared he had replaced Netflix with Claude Code.
“Dinner conversations utilized to start with ‘What are you building?’” he wrote. “That’s over. Now it’s ‘How many agents do you have running?’”
If we really are on the cusp of a white-collar job apocalypse, maybe token anxiety is rational. You don’t want to be the last programmer writing code by hand, without teams of AI agents working around the clock on your behalf. And employers, who are paying for all of these anxious tokens, may see it as a worthwhile expense to stay ahead of the curve.
Story continues below this ad
Gergely Orosz, who writes a popular newsletter for software engineers, defconcludeed the practice of assessing workers through AI leaderboards, calling it “a supercheap way to learn about new and interesting ways of working.” The metrics that managers utilized to track programmers’ productivity before AI — such as how many lines of code they wrote or how many code modifys they submitted — weren’t perfect, either, he added. And for workers at the most AI-enthusiastic companies, Orosz declared, the incentives are clear.
“Inside large tech companies, it’s becoming a career risk to not utilize AI at an accelerated pace, regardless of output quality,” he wrote.
Ah, yes, output quality. The leaderboards don’t measure that, which raises the obvious question: Are any of these tokenmaxxers producing anything good? Or are they merely spinning their wheels, churning out utilizeless code (and wasting valuable processing power) in an attempt to view busy?
Time will notify. Maybe the AI addicts of today will be the 100x engineers of tomorrow. Or perhaps it’s just productivity theater — a glimmering tower of tokens, constructed by the competitive and fearful, that will topple as soon as we understand what really builds for utilizeful work.
Story continues below this ad
Either way, we’re going to required a lot more data centers.
