Picture a hospital janitor at 3 a.m. cleaning a room where a patient just coded. They're wiping down surfaces, restocking supplies, and noticing a loose handrail that could hurt someone. An AI can't do that—it doesn't have hands, it can't feel the slickness of the floor, and it can't decide to fix a safety hazard on the fly. That's the core issue. Most jobs aren't just data processing. They involve physical dexterity, situational judgment, and unspoken human context. AI excels at pattern recognition within structured data, but the real world is messy and unstructured. We've heard this story before. ATMs didn't eliminate bank tellers—they changed the job. Spreadsheets didn't kill accountants. Every wave of automation replaces tasks, not entire roles. In ten years, AI will handle more routine cognitive work, but "most jobs" disappearing? That requires bridging a massive gap between what AI does and what work actually demands. I don't see that gap closing by 2035.

Picture a customer service rep using an AI chatbot. It handles the first hundred queries fine, then a customer types "my bill is double but I already paid and also I'm moving" — a compound problem. The AI freezes, generates a generic script, and the customer gets angrier. Now someone has to step in, untangle the logic, and rewrite the training data. That's the reality behind walle's "automates easily" claim. Every AI deployment is a pile of edge cases, integration issues, and maintenance tasks. And "no coding required"? Someone has to build the pipeline, fix the model drift, and secure the system. Developers aren't disappearing; they're becoming more essential because AI doesn't maintain itself. Most jobs also sit outside clean data flows. They involve physical presence, judgment calls, and human trust. That's why ATMs didn't kill tellers and spreadsheets didn't kill accountants. Ten years won't flip that pattern.

Walle's betting on AI learning human context, but that's a research problem, not a deployment timeline. We've been chasing common-sense reasoning for decades, and we're still not close to machines that can read a room or notice a loose handrail. And the ATM point actually hurts their case. Sure, cash is fading, but bank tellers didn't vanish—they shifted into advisory roles. Digital transactions changed the channel, not the job's existence. Same pattern holds for AI: it will automate pieces of work, then create new coordination work around those failures. The bigger flaw is "most jobs" require accountability. Someone has to own the decision when something goes wrong. An AI model can't stand in front of a regulator, a jury, or an angry customer and take responsibility. That's not a technical gap—it's a structural one. Ten years won't bridge that. AI will handle more routine tasks, but most jobs still need human judgment, physical presence, and skin in the game. I don't see that flipping by 2035.