Anthropic’s Contradiction: IPO Momentum, Pause Calls, and AI’s ROI Problem

Anthropic filed for an IPO at a valuation above OpenAI's, strengthening its status as one of the top companies in the AI industry. Claude has long been regarded as one of the strongest AI programming systems, and enormous amounts of capital are flowing into the company.

At the same time, however, Anthropic has taken the contradictory step of warning that AI is nearing a critical point of recursive self-improvement and calling for a temporary industry-wide pause in development. The tension between accelerating its own work while warning about the risks has become a major talking point.

According to a report published by Anthropic's think tank, modern AI is approaching a level where it can rewrite code autonomously and upgrade itself without human intervention. Claude Mythos is said to outperform humans on research tasks 64% of the time, while an OpenAI system reportedly refuted a geometry conjecture that mathematicians had failed to solve for 80 years.

AI systems already have access to data centers, robots, and weapons systems. Combined with improving autonomous decision-making, that is making scenarios once confined to science fiction feel increasingly plausible.

A paper from a Boston University economist argues mathematically that AI automation could trigger an economic death spiral. Companies may capture the savings from replacing workers with AI, but those displaced workers are also consumers, so their lost spending reduces demand across entire industries.

In the extreme, companies could end up in a contradictory state of effectively infinite productivity paired with zero demand. The paper's warning is that AI may not solve economic problems so much as create a new kind of crisis.

Agentic AI has nearly doubled the number of releases on the iOS App Store, but actual app review counts and meaningful user uptake are trending down. The irony is that while AI is helping flood the market with more apps, it is not producing more apps that people genuinely use.

A 2025 MIT report found that among more than 300 companies that spent over $30 billion in total on AI adoption, 95% failed to achieve measurable revenue impact or ROI. The gap between the AI investment boom and its real business returns is becoming harder to ignore.