AI gets good at anything with an answer key. Your career is everything that doesn't have one.

Addy's platform is a resource for web developers and performance enthusiasts, offering insights into web performance optimization, browser rendering, and front-end development techniques. Through articles, tools, and case studies, Addy offers insights into measuring and improving web page speed and user experience. Developers can learn about performance budgets, lazy loading, and optimizing web assets to create fast and responsive websites.

Addy Osmani

Addy Osmani, former Google engineering lead, shares career advice for the agentic AI era. The core thesis: AI excels at anything with an answer key, so durable careers are built on judgment — choosing what to build and evaluating whether it's good. Key principles include optimizing for scarce resources (reputation over compensation), learning to find problems rather than just solve them, practicing deliberate skill-building without agents to avoid judgment atrophy, shifting from doing to directing agents with clear specs and verification, owning what you ship regardless of who wrote it, and always pushing past the median output agents produce. Public work, finishing strong, and staying near hard problems are the compounding advantages that matter most.

The Agent-Era Career

Nice post. Reached similar conclusion from a research standpoint: <a href="https://valentinguigon.substack.com/p/the-better-judgment-project-how-ai" target="_blank" rel="noopener nofollow">https://valentinguigon.substack.com/p/the-better-judgment-project-how-ai</a>
If AI systems increasingly do the first-order work, valuable human contributions will come from interrogating the system’s assumptions, recognizing missing variables, and identifying when a confident forecast is miscalibrated or misinformed. […] Additionally, detecting AI blind spots and determining what an event is actually worth to a given individual, given its probability, should become essential new skills. This is its own form of tool-oriented metacognition: confidence about the adequacy of the system that is telling you about the world.
[…]
In a world of abundant answers, machines will likely handle information retrieval, detection of falsehood, and calibration of confidence. With a sufficiently good database, you can link any fact to any other fact and ask any question you want. For that reason, choosing what to attend to, defining the right problem, and knowing which question to ask at time t for the problem at hand will become one of the most important skills. We will need to figure out what it is that we don’t know, and what is the proper way to understand the problem at hand.