Companies are increasingly using the Caveman plugin, which causes AI models like Claude to respond concisely and without civility in order to reduce spending on tokens. The creator of the tool, Julius Brusse, developed it in April 2026, noting that a significant part of the costs go to an "unnecessary prose" - courtesy, caveats, and transitional phrases. He estimated that this approach would reduce the consumption of tokens by about 65%.

Rising AI support costs are forcing companies to switch from fixed subscriptions to pay-for-use, making saving resources a priority. Instead of friendly, expanded answers, models now generate sharp jumping orders, sacrificing "humanity" for efficiency. This creates a paradox: to afford the use of advanced AI, companies deliberately limit its expressive capabilities.

If AI usage prices continue to rise, the economic feasibility of its implementation may be in doubt - especially if the cost exceeds the value of human labor. While Caveman and similar solutions help to contain costs, the long-term sustainability of such a model remains uncertain. The irony is that technology requires artificial simplification for accessibility.