The Context Window of a Career
Your AI has a context window. So does your career. The question is who's holding it.
A context window is the accumulated information a model draws on to produce a response, and the bigger it is, the more coherent the output. Every major AI company is pitching continuity right now, because a system that knows your history serves you better than one that starts from scratch every time. That's also a decent description of what a good mentor does.
What Mentors Actually Hold
Mentorship isn't advice-giving or scheduled one-on-ones. The thing a mentor holds is a running model of you, accumulated over real interactions, under pressure, across time, that lets them say "you're ready for this" or "you're not there yet" in a way that carries actual weight, because it's grounded in observation rather than a guess.
The most important professional knowledge isn't articulable — it lives in judgment, in pattern recognition, in how someone handles a situation going sideways. Research suggests that the vast majority of what an organization knows is tacit in exactly this way. All of it transfers person to person, through proximity and time. None of it transfers any other way.
The Gap Is Already Real
Most employees don't have a mentor, and for Gen Z, the gap is sharper. A survey across 44 countries found that the overwhelming majority of Gen Zs said mentorship mattered to them, roughly half wanted their managers to actively teach them, and only about a third said that was actually happening. That's a lot of people waiting for something that isn't coming.
A long-running study tracking employees over more than a decade found that mentored employees were promoted five times more often than those without, and most stayed versus fewer than half who weren't mentored, the difference between a career that compounds and one that stalls. Entry-level job postings have fallen sharply in recent years, and young developer employment has dropped too, which means the apprenticeship relationship is becoming less available precisely when the tools that seem to substitute for it are everywhere.
For previous generations, professional context accumulated through relationships: a manager who watched you present badly and helped you understand why, a senior colleague who caught you solving the wrong problem before you'd sunk three weeks into it. That's a context window. It persists, it's observational, and it includes things you never explicitly said.
AI memory doesn't do this, and not because the technology is immature. There's a gap researchers have identified between how AI memory feels and what it actually is. It stores logs of what was said, but it can't reconstruct what you were like under pressure, how your judgment has shifted, or whether you're actually ready for the next thing. Conversation logs are a transcript, not a story.
What You Become Without the Window
In one survey of more than a thousand interns, nearly all of them used generative AI weekly. When asked what was most valuable about their internship, they ranked mentoring relationships second only to hands-on experience, with AI-powered learning near the bottom. The very thing they were using constantly was the thing they valued least as a developmental experience. They knew the difference, even if they couldn't always articulate it.
A lot of Gen Z report interacting less with colleagues because of AI, and I think that's where the real cost builds quietly, not in any single interaction but in the aggregate, in the professional context that never forms because there's nobody holding it. The AI gives you answers, reduces the friction of not knowing something, but it doesn't watch you grow. It doesn't notice you're ready, and it doesn't stake its reputation on your promotion.
The question probably isn't whether AI tools make early-career professionals more productive. They probably do. The harder question is whether that productivity is being purchased at the cost of something that compounds differently: the accumulated context that tells you which problems are worth solving, how to reason about things you've never seen before, when to trust your own instincts. That knowledge doesn't transfer through prompts. We seem to be running a reasonably large experiment on what professionals become when the context window of a career never really opens, and perhaps nobody is tracking the results.