At a London anti‑AI rally, activists from Pause AI handed out flyers that riffed on the South Park “underpants gnomes” meme: “Step 1: Grow a digital super mind. Step 2: ?. Step 3: ?.” The joke captures the current state of generative AI—companies have built powerful models and promise sweeping change, but the regulatory and integration pathway remains undefined.

Researchers at Anthropic released a study flagging managers, architects and media professionals as the occupations most likely to feel AI disruption, yet the analysis rests on task‑level capabilities rather than real‑world performance. A separate February experiment by Mercor tested agents from OpenAI, Anthropic and DeepMind on 480 banking, consulting and legal tasks; every model failed to complete the majority of assignments, underscoring the gap between benchmark success and workplace utility.

The discord stems from divergent incentives and an information vacuum that lets bold claims sway markets without solid evidence. Sustainable transformation will require transparent model reporting, coordinated evaluation frameworks, and a clear “Step 2” that aligns technology with existing workflows. Until those pieces fall into place, AI’s promised profit remains speculative.