Palace/Library

The Cost of Staying

careeraistartupsjudgmentopportunity-cost

The Cost of Staying

Author: Anonymous Bloomberg Beta investor Type: Essay — career strategy in the AI era

Core Frame: The K-Curve

The AI era is creating a K-curve divergence in tech careers: those who repositioned early are on the upper arm (compounding); those still deliberating are on the lower arm (falling behind). The key variable: time. Every quarter spent in the wrong seat, the gap widens and compounds.

"The valuable skill in tech went from 'can you solve this problem' to 'can you tell which problems are worth solving and which solutions are actually good.'" The scarce thing flipped from execution to judgment.

Career Segment Analysis

FAANG

  • Increasingly reviewing AI-generated outputs rather than building from scratch
  • Those staying: betting stability + comp > frontier proximity
  • Those leaving: bet that frontier is where next decade of career value compounds, and every quarter of delay is compounding missed
  • Both bets are rational. Only one is time-sensitive.

Quant

  • Still works — absurd pay, hard problems, immediate P&L feedback
  • Emerging tension: the full quant toolkit (ML infrastructure, data obsession, statistical intuition) is exactly what AI labs need. One has a ceiling; the other doesn't.
  • Those leaving: hit the point where finance felt "bounded" in a way it didn't before

Academia

  • The line between university labs and funded startups is blurring in academia's disfavor
  • A 20-person research startup can now do in a weekend what takes an academic lab a semester (compute costs)
  • The most ambitious PhDs aren't choosing academia vs industry — they're choosing theorizing about experiments vs actually running them

AI Startups (Application Layer)

  • Clever features get commoditized by model updates every quarter
  • Those thriving: stopped caring about model capabilities, started caring about what models can't take away — data moats, workflow capture, integration depth
  • The sharpest moves: getting excited about plumbing, not demos

Research Startups (New Center of Gravity)

  • 10–30 people doing genuine frontier research competing with organizations 50× their size
  • Daily workflow: kick off training runs, come back and apply judgment to results — taste to distinguish signal from noise
  • Passive leverage: you set experiments in motion; compounding happens whether or not you're at your desk
  • Skills and network transfer even if the specific company fails

Big Model Labs

  • Prestige vs proximity tradeoff
  • Most interesting research concentrated among a few senior people; everyone else 3 layers removed from the frontier
  • Big-lab pedigree is depreciating as labs get more corporate; "I did frontier work at a place where my judgment shaped the direction" is appreciating

The Clock

A year ago, deliberation was cheap. Today: the tools are compounding. People who moved 6 months ago are building on last quarter's learnings. The divergence between the mover and the waiter is already compounding — not because opportunities disappear, but because the people already there are getting further ahead.

"The best people want to be close to other people who have tricks they haven't learned yet, at places with enough compute to actually run the experiments."

Related Sources