Alex Finn recommends starting with recurring frictionProblem framing with AI is the human-led process of defining the real goal, constraints, evidence and success criteria before using a model to explore solutions. rather than a narrowly specified app idea. Fable 5 can inspect the broader workflow, identify repeated decisions and help design a personal operating layer that stores preferences and feedbackAI agent memory is stored information that an agent can retrieve and use across steps, sessions, or changing contexts. once instead of rebuilding contextContext engineering designs how relevant instructions, evidence, state and tool information are assembled for a model's task. for every task.
Alex Finn suggests asking the model to propose several directions before choosing an implementation. This creates room for the agent to surface overlooked opportunities instead of simply producing the first interface the user imagined.
Alex Finn also highlights browser delegation and model allocationModel selection chooses the AI model whose capability, quality, cost, speed, safety, and operating constraints best fit a task.. External research and multi-step web tasksComputer use lets an AI agent interact with software through interface controls such as screens, buttons, keyboards and pointers. benefit from an agent that can act across tools, while scarce premium-model usage should be saved for planning and decisions that require the most judgment.
Watch on YouTube




