Дата публикации
IT developers increasingly feel dependent on usage limits of AI models. Max Johnson, co‑founder of the UK startup Briix, admits that he used to work with Claude without restrictions, but now must plan tasks around an invisible token counter.
In early 2025, Anthropic tightened limits to manage growing demand. Johnson says sometimes he exhausts his quota in just two queries and has to wait until evening to continue. His colleagues also adjust schedules, spending up to an hour waiting for reset.
NYU math student Ani Potts treats limits like a budget. He focuses on intensive work blocks, then switches to smaller tasks when nearing the cap. He notes pauses help him “use his brain again.”
Toronto developer Daniel Qureshi, paying about 28 CAD for Claude Pro, often halts projects when hitting the limit. He believes manual coding is no longer practical with AI available.
Yet Qureshi sees benefits: tasks once taking hours now finish faster, reducing burnout. On weekends he spends time building AI agents while staying productive.
Tariff restrictions are becoming part of daily routines: developers plan work around tokens. This changes habits but sometimes improves efficiency and prevents overload.
Meanwhile, reports from China show a rise in companies where AI agents perform key functions. Such firms scale business without hiring staff, using agent systems and “vibe‑coding.”
The AI Trap
Limits highlight a deeper risk: when AI becomes the main tool, developers risk turning into operators. Psychologists call this “cognitive offloading”: the brain gets used to delegating everything to algorithms. Without AI access, even experienced specialists may feel helpless.
This dependence forms not only among professionals but also learners. Studies show excessive generative AI use by teenagers may slow development of critical thinking and information‑seeking skills.
