The Agent-Era Career
When AI commoditizes problem-solving, your career becomes the ungradeable parts: choosing what to build, judging if it's good, and finishing past where the machine stops.
Read Original Summary used for search
TLDR
• AI will automate anything with an answer key—the durable career skill is selecting which problems matter and judging solutions, not solving them
• Optimize for scarce resources (reputation, judgment) not abundant ones (capital, AI output)—reputation compounds, quick bucks from vibe-coding are worthless
• Build judgment through deliberate practice: do critical work without agents, read 1000x more code than you write, flag every wrong AI suggestion to accumulate taste
• Agents give you 70% of a feature fast; the last 30% (edge cases, architecture, polish) is the whole game and where all value concentrates
• Increase both your xG (opportunities from public work) and finishing rate (judgment to convert them)—big opportunities come from work done in public, not job applications
In Detail
Addy Osmani (ex-Google Chrome/Gemini lead) argues that AI's core competency—anything with an answer key—will commoditize problem-solving itself, making the scarce skill choosing what to build and judging quality. The career advice shift: optimize for scarce resources like reputation and judgment, not abundant ones like capital or AI-generated code. His open source work had zero direct payoff but compounded into opportunities; quick bucks from vibe-coding are worthless when shipping is trivial.
The framework for building durable judgment: engage in deliberate practice by doing critical work the hard way without agents, building deep mental models. Read a thousand times more code than you write. Keep a private log of every wrong AI suggestion you catch—that's where taste accumulates. Treat agents like human delegation: scope tasks, define done, calibrate trust (autonomy is a per-task dial, not a rank), and verify results. The 70/30 rule: agents deliver 70% of features quickly, but the last 30%—debugging edge cases, architecting correctly, polishing—is the whole game. When first drafts are free, finish is the product.
Career opportunities follow an xG model from soccer: your reputation creates chances (xG), judgment converts them (finishing rate). All of Osmani's big opportunities came from public work, never job applications. You can't script which chances arrive, only whether you're positioned to take them. The world isn't short on opportunity; it's short on people who can find the right problem, verify the machine solved it, and finish past where it stopped. Your attention doesn't scale with AI output—protect it ruthlessly. The ungradeable career is choosing what matters, judging honestly when you've got it, and answering for it in public near hard problems.