oneusefulthing.org
|
ksl
|
|
Ethan Mollick’s latest guide draws a sharp line between AI models and the harnesses wrapping them – a distinction most users still ignore. His take: Claude, ChatGPT, and Gemini are nearly interchangeable at the model layer, but the apps built around them diverge wildly in what they actually let you do. Google’s Gemini matches its rivals in raw capability, yet ships with an interface that strips away the agentic features making these tools useful for real work. The buried thesis here isn’t about picking a chatbot. It’s that anyone still typing prompts into a conversation window is stuck in last year’s paradigm. Mollick’s framing tracks with a broader shift across knowledge work where deploying agents, not chatting with AI, is becoming the productivity divide that counts.
