The piece discusses cost pressures in the AI industry, highlighting token minimization, the rise of agents, and big capex spending, and questions whether energy use and memory constraints will translate into enduring profits.
Claude Code's Caveman skill strips non‑essential linguistic elements from model responses, making the AI reply in a parody of a coding Neanderthal.
Rupert Goodwins wrote in The Register that Caveman is the latest and most surreal consequence of token minimization as the AI industry confronts sharply rising operating and capital costs, with the sector expected to spend a trillion dollars on capex this year.
Goodwins cites the Bank for International Settlements' comparison of the AI capex carnival to historical infrastructure booms-canals, railways and electrification-to argue the spending frenzy may not translate into enduring profits.
He warns that AI is consuming energy, components, datacenter capacity and developer attention, a competition that has driven annual inflation of 300-400% into the memory supply chain and damaged OEM economics.
Goodwins says token minimization and the rise of agents—"a whole new world of tokens‑a‑gogo" exposed by OpenClaw-have produced a "tokenpocalypse", prompting enterprises to reassess model value as revenue proves less elastic than vendors assumed.
Goodwins concludes, "Douglas Adams may have died 25 years ago, but he's still writing the script."
