OpenAI chief economist career advice: don't copy success
OpenAI chief economist career advice rarely sounds this blunt: stop trying to copy the résumé of someone you admire. Ronnie Chatterji, who also teaches at Duke University's Fuqua School of Business, told the "Summation" podcast in an episode released Tuesday that studying a high achiever's career and retracing their steps is "relatively bad advice" (Business Insider). He was pushing back on what he called the "success leaves clues" mindset, the idea that following the same schools, jobs, and moves as someone who made it big will produce the same outcome for anyone else (Business Insider).
His reasoning holds up under scrutiny. A public career timeline almost never shows the failed attempts and lucky breaks that shaped it, and the economic conditions that made someone's path work may already be gone (Yahoo News Malaysia). That argument lines up with independent research on how people misjudge success, and it carries extra weight now that AI is actively reshaping which career strategies still make sense in a changing job market. Here's what the evidence shows, and what to do with it instead of copying a résumé.
Why copying successful careers is bad advice
Chatterji's core objection is simple: retracing someone's résumé step by step treats one visible, polished outcome as a repeatable formula (Business Insider). Nature's analysis of science-career advice describes the same pattern under a formal name, survivorship bias. The people who stayed in a field long enough to become mentors and advice-givers are, by definition, the ones who made it through. Everyone who left along the way isn't around to explain what happened to them (Nature).
That distortion shows up in controlled testing too. A study published in Judgment and Decision Making found that participants shown examples of successful college graduates predicted graduates succeed more often 87% of the time. Participants shown examples of successful dropouts predicted graduate success only 32% of the time, a 55-point swing driven entirely by which small, cherry-picked sample they happened to see (Cambridge Core). Even when participants knew the sample was biased, they still let it sway their bets (Cambridge Core). The researchers describe this kind of distortion as common practice in media and communication generally (Cambridge Core).
Line those findings up against Chatterji's comment and the picture is consistent. Whoever ends up successful enough to get profiled is a small, unrepresentative sample of everyone who tried a similar path. Following their sequence of moves means learning from survivors only, with no way to compare their outcome against the people who took a similar risk and didn't land the same way.
Should you follow successful people's career paths?
Setting bias aside, Chatterji's second argument is about timing. The choices that worked for a given founder or executive were "very context dependent," and he added that it's "very rare to find someone self-aware enough" to admit their approach might not work for anyone else today (Yahoo News Malaysia). He used starting a company as an example: the right time to found a business depends on how dynamic the broader economy is at that moment, not on an individual's ambition or personal timeline (Yahoo News Malaysia).
That timing problem is measurable right now because of AI. OpenAI's labor-market framework, published earlier this year across more than 900 occupations, estimates that 18% of jobs face relatively higher short-term automation risk, 24% may see task reorganization alongside declining employment, 12% could grow as AI lowers costs and expands demand, and 46% face comparatively less near-term change (OpenAI). The report frames these figures as a map of where labor-market pressure may emerge first, not a forecast of who loses a job (OpenAI).
OpenAI is careful about the limits of its own numbers, too. The report states that AI exposure alone is too blunt a measure to predict which jobs will actually be automated, redesigned, or expanded (OpenAI). It also disclosed that it corrected an error that had improperly swapped archetype classifications for a small subset of occupations, a fix that changed the very figures and table behind these percentages (OpenAI). Read these numbers as a directional map of where change may land across occupations broadly, not a prediction about any one person's job.
If the economic and technological backdrop has already shifted since a role model's career took off, copying their sequence of moves means copying a strategy built for a labor market that may not exist anymore by the time it reaches someone new.
How to use career role models for inspiration, not imitation
Chatterji's alternative isn't to ignore people who succeeded. It's to change what gets copied. "We should use our heroes, our career heroes, our occupational heroes for inspiration but not step-by-step advice," he said (Yahoo News Malaysia). His distinction: "Success leaves clues, but it's not in the step-by-step... it's often in the habits" (Yahoo News Malaysia). The transferable part of someone's career is how they behaved, not which job title came before which.
Instead of modeling a career on a specific founder, executive, or investor, Chatterji recommends identifying a problem worth solving and positioning yourself where that problem is actively being worked on (Yahoo News Malaysia). He names specific capabilities he expects to matter regardless of which path someone follows: coding, math, formal reasoning, and understanding how AI works, along with judgment, flexibility, resilience, and the ability to spot problems worth solving as AI takes on more routine tasks (Yahoo News Malaysia).
That translates into a four-part check before modeling any decision on someone else's career:
- Separate the behavior from the circumstance. Note how the person made decisions, chose problems, or recovered from setbacks. Set aside their specific employer, degree, and founding year, since those were shaped by conditions that may not repeat.
- Check the target occupation's current requirements. Before assuming their path applies to a role you want, look up that occupation's current education level, typical entry point, and projected demand instead of relying on how one person got in years ago.
- Compare that against current listings and personal constraints. National data doesn't account for local hiring patterns, finances, location, or existing experience. Pull current job postings in the relevant market and industry to see what employers are actually asking for now.
- Ask a mentor about the failures, not just the wins. If someone experienced is willing to talk, ask what didn't work, what they'd time differently, and what alternatives they turned down, not just the highlight reel that made it into a bio.
None of this rules out learning from someone experienced. It just changes what gets borrowed: judgment and problem selection instead of a copied timeline.
What to research before modeling a career on someone else's path
Before treating anyone's career as a template, pull current occupational data instead of their origin story. The U.S. Bureau of Labor Statistics' Occupational Outlook Handbook lists current education requirements, typical pay, and job-growth projections by occupation, while O*NET breaks down the specific skills, tasks, and entry requirements tied to that role. Compare what those sources show against the actual problem or industry worth working in, then check that against current job postings in the location where the work would happen. That comparison, not someone else's résumé, is what should decide the next move.
