Motarme

The best essay written about AI this year just dropped and nobody in tech wants to hear it. Manidis names something most of the industry is allergic to discussing: the majority of AI deployment today produces the sensation of work, not work itself. He calls these “tool-shaped objects.” You can hold them, use them, watch tokens stream across your screen, monitor chain-of-thought reasoning, swap models, add tools, build agents that call other agents that generate memos nobody reads. The number goes up. The output doesn’t. He compares it to FarmVille. No matter where you click, your farm expands. The direction of your effort is irrelevant. The experience of productivity is the product. And the market for feeling productive is orders of magnitude larger than the market for being productive. The data backs this up almost perfectly. McKinsey surveyed ~2,000 enterprises in 2025 and found 6% qualify as “AI high performers” with meaningful bottom-line impact. Six percent. Meanwhile enterprise AI spending went from $1.7B to $37B since 2023. A 22x increase in two years. Larridin found 89% of enterprises use AI but only 23% even measure ROI. Most organizations literally cannot tell whether they’re running a farm or running FarmVille. But here’s where it gets interesting. That 6% who figured it out? They’re seeing 3-4x higher productivity gains than beginners, redesigning entire workflows, and pulling away at a rate that makes the spending rational for everyone else as a defensive bet. OpenAI’s enterprise data shows workers in properly deployed organizations saving 40-60 minutes daily. Wharton found AI deals convert to production at 47%, nearly double the 25% rate for traditional SaaS. So the tool-shaped object problem isn’t that the tool is fake. The problem is that 77% of organizations never built the measurement framework to know which side of the gradient they’re on. They’re spending because spending feels like strategy, not because they can connect tokens consumed to value produced. This is how every general-purpose technology diffuses. Factories installed electric motors and wired them to the same belt-drive layouts they used with steam engines. Productivity didn’t move for 30 years. It took managers who understood that electricity meant redesigning the factory floor before output followed. The AI adoption curve looks nearly identical, just compressed. The Mac Mini agent setups, the orchestration layers, the billion-dollar frameworks: these are belt-driven electric factories. They will be replaced. But by the 6% who redesigned the floor, not by people who saw the belt-drive phase and concluded the motor was decorative.