I’m back from a bit of a hiatus after being in the UK! I’m here for a few more days if you’d like to reach out - very exciting upcoming projects are coming up.
Logbook | Dossier | Institutions
I’m calling it: we’re on the brink of a researcher fertility crisis.
A few weeks ago, I came across a viral reel by surgeon Dr. Steven Cyr, who, as he scrubs up, comments on the brink of a population crisis: ironically, that the US doesn’t have enough medical students to replace practicing doctors.
He’s right. We talk endlessly about demographic declines, but there’s a replacement crisis in careers that require professional degrees – including research-based jobs – that few are naming.
Sure, we have more PhD students than ever before, but that’s the same stat people cited about apparent overpopulation 10 years ago. The real story is the exodus that follows a peak.
Applications to US biomedical PhDs fell 13% between 2018 and 2024. 60% of PhDs either prolong or drop out of their PhDs. More than 40% of postdoctoral researchers (out of a survey of 45,500) will leave academia altogether. Meanwhile, “get that coin” has become Gen Z’s career motto (not “get that Nobel.”)
The numbers signal that a mass defection from research is just around the corner – maybe already happening.
The Declining “Researcher Fertility Rate” (RFR)
I remember being ~10 years old when adults around me were freaking out about overpopulation.
Less than a decade later, many developed countries were lamenting that the Total Fertility Rate (TFR) was well below an idealized replacement rate of 2.1 children per woman.
But for many countries, fertility has been below replacement for decades. What this illustrates is how we (by we, I mean broader society) are usually reactive to things. We stay “comfortable” until we are in the midst of them happening, and then scramble to head back. I suspect the same narrative will happen in research, where we tout the horn far after we’ve passed a decline because we haven’t felt it yet.
Effectively, the TFR drops when the number of women of childbearing age who choose to have children decreases (or the number of children they have decreases).
In research, we can build a similar metric. What happens when the number of PhDs who choose to stay in research and become PIs decreases? I call it the dropping researcher fertility rate (RFR).
In the table above, the replacement threshold is calculated to show that there are 7-11 eligible researchers available for each PI job that opens up.1
Note: the RFR cannot be calculated by dividing the total number of new PhDs by retiring PIs, as this doesn’t reflect the number in research. We have the biggest population of PhDs ever, but these readily-educated people are actively choosing not to continue with research. This parallels the large population of childbearing women who are choosing not to have children. That’s where the problem lies.
At first, a rate of 7-11 looks fine. Academia is one of the most bottlenecked and competitive industries to secure a PI position in, which means we should have plenty of demand, right?
Nope.
Fewer PhDs want to pursue academia or research at all, which means not only is there a group being forced out of academia (with the current shortage of jobs), there is also a group choosing out at a rapidly accelerating rate.
Keep in mind that our current figure of RFR 7-11 is based on pre-2025 data, looking at a group from the classes of ~2010-2020. We can reasonably expect that, for a variety of cultural, political, and economic factors (I expand on this below), Gen-Z PhD holders (class of ~2020-onwards) will turn away at exponentially higher rates. This number also does not include the number of mid-career scientists who leave academia, only factoring in a new cohort.
But Jolie - research careers span more than just academia! Industry researchers matter too! I scoped RFR to academia first because it’s the most measurable and we have the most readily available data for it. But even then, academia is still the largest individual share of basic research; in the US, 44% of research is done in universities, and 55% of R&D funding is allocated by the federal government. In the UK, 47% of public R&D funding goes to higher ed.
Academia is also the feeder of industry research. If RFR in academia drops in a few years, you can’t just “make it up” in industry later, because industry mostly hires from that academic basin and benefits from the newly-minted specialists that trained in academia, rather than having to train said specialists themselves. That’s part of what allows them to move quicker.
So why haven’t we felt anything yet? Age-structure lag. The cohorts that thinned out after ~2012 have just begun to hit PI-eligibility (generally speaking), so the crunch arrives late and then all at once. Unlike births, which you can count and forecast when that cohort hits reproductive age, research has no clean census: we don’t usually systematically track the total denominator (PhDs, postdocs, staff scientists, lab techs, unpaid undergrads, MS students), and outputs are lumpy and delayed. It took years to name the fertility crisis; it’ll take longer to see RFR cleanly because a research career spans decades and what people do (or don’t) discover is hard to measure in real time.
But more importantly, I’d argue the pre-symptoms are already here: swollen grad cohorts funneled into few labs, stacked postdocs, and labs stuffed with overqualified researchers maintaining cell lines or cranking out maintenance data to look productive (this is not intellectually stimulating, and can actually factor into why the drop will continue to increase - more on this below). The pain shows up first as underused talent, until those smaller cohorts age into the PI window and the gap becomes impossible to ignore. This will be interesting to track, especially as lab automation becomes mainstream.
Why a Low RFR Matters
When our RFR for researchers sits low for long enough, the pyramid flips. You end up with lots of seniors, a hollow middle, and not enough people being educated and trained to fill those spots. Even if interest spikes later, you can’t backfill fast: labs close and empty out before replacements finish training, and the window for certain discoveries closes.
What breaks first?
Idea diversity & quality. Fewer independent labs → fewer bets being made → a narrower literature.
Teaching & mentorship. With fewer mid-career PIs (the best being poached by industry at any given stage of their careers), students get less hands-on craft, thinner networks, reduced mentorship.
Governance capacity. Fewer available reviewers means even longer grant and journal queues, potentially longer panels and study sections as the number of investigators decrease per study.
Completed cohort output
In demography terms, you pick a birth cohort (say, women born in 1990) and ask:
“How many kids did this exact group have by age ~40?”
That’s ‘completed-cohort fertility.’
The research parallel is: pick a PhD class and follow that same group for ~20 years. Add up what they produce: papers, patents, grants won, students mentored - this is tricky, because many are intangible.
That total is the cohort’s ‘completed output.’
This matter because period stats can look fine even while individual cohorts are under-productive or shrinking. Cohort tracking tells you if each ‘incoming’ generation of researchers is really replacing the last.
Mentor Ratio
In demography, the dependency ratio is the number of older adults 65+ relative to working-age adults; when it rises, fewer workers support more retirees and kids.
In research, swap in mentor ratio = active PIs ÷ trainees (or its inverse, trainees per PI). When this worsens, fewer PIs per trainee, each student/postdoc gets less hands-on time, fewer first-author chances, slower letters/networking, and longer queues for reviews and grants. Throughput drops because PI bandwidth is the bottleneck.
The Cultural and Economic Drag
How did we get here? This didn’t happen overnight, and certainly not by accident. As with fertility, there are a mixture of technological and socioeconomic factors, but you can really boil it down to Gen Z’s nihilism, heavily influenced by late-stage capitalism.
When you don’t believe there’s a future worth building for, why pursue research - or anything mission-driven - at all?
Kyla Scanlon’s piece, Gen Z and the End of Predictable Progress, is a great frame for some of the things I’m about to break down. She splits cohorts into Millennials, Gen-Z-1 (1997–2002, more optimizer/striver) and Gen-Z-2 (2003+, more pragmatic/defensive), and how the approaches and philosophies of all three are different.
a) The Cultural, Social, Philosophical Collapse
1. Why Wait Until 50?
The average age for a first NIH grant is near 50 years old. In the 1980s, it was 35. This changes your worldview dramatically.
Gen Z has watched climate change accelerate, democracies wobble, and entire industries evaporate overnight. They’re the most nihilistic generation about the future we’ve ever measured. Now you’re asking them to grind for 25 years to maybe run their own lab?
They’re not optimizing for the slow burn, high-impact route because they genuinely don’t believe the institutions (or jobs, or specializations) will exist by the time they’d inherit them. It’s an adaptive pattern recognition and active avoidance. When everything feels like it's collapsing, you optimize for quick wins and faster security. Get in, get paid, get out.
2. Things No Longer Compound
Compounding as a life strategy is fading in practice for Gen Z.
Previous generations could visualize their career arc: entry level → senior → manager → director → VP. Each year built on the last.
That mental model is gone. Gen Z has internalized that careers are non-linear, companies are temporary, and entire fields can be automated away between your freshman and senior year. The idea of spending 7 years on a PhD, then 5 on a postdoc, then fighting for tenure until you’re 45 assumes a stability that feels fictional.
If jumping from place to place is to be expected, why would you lock yourself into one spot - and take the slowest path possible?
3. Commitment Issues
Gen Z avoids anything irreversible, but it’s rational as a single actor.
Law school? AI’s coming for that. Medical school? Telemedicine and robotics are restructuring the whole field (or it’s years and years of overpaying the school, then years and years of being underpaid).
A research career feels pretty irreversible, especially if time is currency. Seven years for a PhD that qualifies you for... more research. Your exits are frankly limited, especially compared to peers at tech companies or consulting firms. Two years at Google is a passport to essentially anywhere in tech. Two years building an audience is an asset you own forever. Two years in a PhD program? You’re not even done with coursework, the work isn’t “yours” yet, and you don’t have the finances to invest in something else.
This is the same commitment phobia that kills birth rates. Kids are irreversible. PhDs are slightly less, but still somewhat, irreversible. Gen Z has learned that in a world of constant disruption, optionality is the only real asset, a place where you can pivot fast. Research offers none.
4. The Great Meaning Inversion
We’ve completely inverted what “meaningful work” means.2
For previous generations, the formula was clear: sacrifice now for meaning later. Go through grad school to cure cancer. Accept temporarily low pay to push the boundaries of human knowledge. The meaning justified the sacrifice, especially if respect, leadership, and reputation compounded linearly and predictably.
But meaning means nothing compared to optionality in 2025, meaning is cheap if/when your niche doesn’t exist in 20 years, meaning can’t cover your expenses. When you can’t afford rent, when healthcare is tied to employment, Maslow's hierarchy reasserts itself.
The cultural programming has flipped. On Dialectic, Eugene Wei makes an astute observation that language tells us much about the values we care about. “Get that coin” seems harmless, but it’s a survival strategy. The kids who choose meaning over money aren’t seen as noble anymore.
5. Status Hemorrhage
Being a researcher is seen as low-status now.
If you tell a non-academic you’re doing your postdoc and watch their face, there’s a tiny microexpression of pity mixed with confusion that lasts for just a beat. “Oh... that’s... interesting. So, like, you're still in school?”
There’s also something to be said about the collapse of scientific role models. Kids want to be MrBeast because MrBeast is winning by every metric their culture values. No modern biologist is.
The Economic Reality
1. Stipend Erosion
The typical US PhD stipend generally ranges from US$10,000-45,000. A postdoc can expect ~US$61,000. Early-career researchers have only seen incremental nominal increases since the early 2010s; adjusted for inflation, real pay has eroded. Compare that to direct workforce entrants (bachelors) at $69,000, and managing costs of living in Boston, San Francisco, New York.
All the while peers’ salaries are starting at $180,000 plus equity out of undergrad (tech) or $95,000 plus bonuses (consulting).
The opportunity cost isn’t abstract, and it hurts to watch everyone else build wealth.
2. The Risk-Reward Calculation Is Broken
Here’s the scatter plot that should be in every university admission office:
PhD-to-PI track: ~10-20% chance of tenure-track position, median lifetime earnings $3M
Failed startup founder: 80-90% failure rate, but repeat founders have a 20% success rate on next venture, with successes leading to significantly higher lifetime earnings
FAANG engineer: Near-zero failure rate once hired, $4-6M+ lifetime earnings depending on seniority and equity, can exit essentially anytime
The PhD feels like VC risk with bond returns. You’re taking 7+ years more to complete your schooling, for the privilege of maybe earning less than a dental hygienist (for reference: median $94,000/year). The variance seems to be all downside.
3. The Job Market Is Statistical Violence
Let’s be clear about what “job prospects” in academia means:
~1,367 PI positions open per year (based on a ~3% PI retirement rate in the US, out of ~45,500 active PIs)3
10,000-15,000 candidates competing for them
7-11 candidates per position before you factor in postdocs from previous years (our original RFR)
4. The Hidden Costs
The economic hit isn’t just your immediate salary. It’s:
Delayed retirement savings: Starting your 401k at 35 instead of 22 costs you ~$800K-1M in compound interest
Healthcare: Some grad student insurance doesn’t cover vision, dental, or mental health costs
Geographic lock-in: You go where the postdoc is, not where houses are affordable
Time tax: Try explaining to a partner why you need to move to Wisconsin to sit in a lab for 12-16 hours a day, making your $56,000 salary. (Sorry, Wisconsin).
The Numbers Don’t Lie (But Universities Might)
Universities publish median salaries for PhD graduates that include the one person who founded a unicorn startup. They don’t publish the postdoc grinding for $50K at age 35. They celebrate the Nobel laureate, not the thousand adjuncts teaching intro bio for shockingly low wages.
Cultural nihilism makes people unwilling to accept economic sacrifice. The economic reality reinforces that the culture was right to abandon ship. Each confirmed exit becomes another data point that choosing research is irrational. The story becomes self-fulfilling.
And just like with fertility, once the dominant cultural narrative sets in, no amount of funding can fix it. Very rarely can you pay people enough to believe in a future they’ve already written off.
Two tempting “fixes” that make λ worse
Fix #1: “Just admit more!”
On paper, widening the funnel sounds equitable. In practice, if you don’t also expand PI capacity, stipends, and time-to-independence, you get credential inflation. More trainees chase the same number of mentors, first-author slots, and faculty jobs. The bar to independence drops (because committees must move people through), but the work doesn’t get better: PI bandwidth is the scarce input, and you’ve diluted it. Selection also turns perverse: the people willing (or forced) to endure 8–10 years on low pay are disproportionately those without strong outside options. The ones you most want to retain have alternative offers and leave sooner.
TLDR: if you mushroom the headcount without deepening freedom and resources, you manufacture mediocrity at scale.
Fix #2: “Label grants ‘interdisciplinary’ and ‘risky,’ to encourage constant novel research.
Nice words don’t change payoffs. In a low-trust climate (constant programme churn, priorities that flip with leadership) winners of “risky” calls will most likely still do the rational thing: spend the money safely. This is especially true for young researchers who hedge into ‘safe’ projects because they want to survive if/when the scheme gets rebranded next cycle, and they still have careers to build. Panels generally stay conservative (career risk is asymmetric), so the grants drift back to familiar names and methods. You tend to get performative novelty that appears to support risks or moonshots, but is incremental in practice.
So What Do We Do…?
The ‘right’ solution highly depends on whether you think there’s more merit in resuscitating the current academia model - in which you can feed tenure tracks, pump funding into academia, professorships - or evolving research into something different.
I lean toward adaptation, but have included possible solutions in both avenues (they’ll probably happen simultaneously). Some of these are borrowed from the pro-natal playbook.
1. Cash-front incentives – additional “starter stipends” indexed to cost-of-living. In essence, increasing PhD/postdoc stipends. This could help at the margins, but still pales in comparison to the opportunity cost at non-academic options.
2. Short-cycle entry points – Nine-month “micro-thesis” fellowships that give students the optionality to ‘renew.’ Gen Z loves optionality. The challenge: academia tends to reject anything this flexible, and this can be high lift when short-term undergraduate or masters-level research terms are meant to do this.
3. Loan insurance – Government backstops your debt if you exit before year 3. De-risking exploration is smart. The trust issue exists, though, because the discussion of student loan forgiveness has ping-ponged.
4. Immigration fast-tracks – Expedited visas for under-35 researchers. This one works. America’s research dominance was built on importing talent, and there are pools of readily available talent in foreign countries (see China, India).
5. Narrative campaigns – Pro-natalists highlight women who have experienced both motherhood and career success to encourage young women to have children. Status and stories are memetic. Highlight PhD success stories, with positive examples.
The Structural Shifts
5. Distribute research to where young people are flocking to work.
Google, OpenAI, and Arc Institute4 are just a few industry teams that have already absorbed significant talent that might have gone to academia. Many more already have R&D arms set up.
This is beginning in biology and software, but I see a future where we have mini-labs embedded everywhere (non-profits, studios, etc). If we normalize flow between them - like default joint roles (0.5 FTE university <> 0.5 FTE industry lab), or portable grants that fund the person, letting early-career researchers carry grants across academia/industry without penalty, this preserves and increases the optics of optionality. This could also appear as co-supervised PhDs that include industry PIs; universities guarantee thesis standards, and projects benefit from shared data/compute via MOUs.
6. Rethink credentialism.
A seven-year doctorate is a fine route to deep craft and the license to supervise labs, but the skills of good research (problem selection, experimental design, data analysis, critical writing) can be taught and demonstrated far faster than that. If both academia and industry are only willing to hire PhDs, we turn away people who could do excellent, rigorous work but didn’t pursue the PhD (for whatever reason - earning potential, time, topical uncertainty being large factors).
The PhD has lost value over time. I’ve spoken to ~50+ recent PhD graduates who defended in the last 3 years: almost none would recommend the traditional PhD path to a newcomer in 2025. They see it as a stamp, not a learning loop.
We would benefit from multiple on-ramps being seen as signals of research competence: two-year research fellowships, apprenticeship models, direct-to-research programs that culminate in a portfolio rather than a thesis. These need to be seen as valuable indicators of ability, even if they aren’t as long as a PhD.
This could be debunked or be proven wrong. But the only real way to test if this works is to either hire or fund individuals who are credentialed by evidence and ability (think open methods exams, replication sprints, trials), not just time.
What if we just don’t need researchers?
There’s an argument in the fertility debate that this is nature’s way of optimizing. Maybe the world is telling us that we need less humans; maybe we don't need the volume of researchers we thought - maybe this is natural selection that leads to fewer, better-supported, more focused ones.
This is fair. With lab automation and robotics, foundation models that can design assays, and synthetic data that can all push the marginal cost of an experiment lower, maybe we do need fewer researchers per unit output. This is exciting! Leverage goes up. But it’s also scary, because automation mostly scales exploitation (running more of what we already know how to do) rather than exploration (selecting the right novel, messy questions).
There’s also the moral problem: what do we tell the generation of PhDs who did everything ‘right’ and now face a collapsed job market? These are people who are well into their adulthood, and for many, have already invested a decade, often more, of irreplaceable time. We owe them more than “thanks and good luck,” right?
And finally, the long-term practical problem: consolidating research into fewer hands means those hands choose what gets studied. When you limit who can research, you limit what questions get asked, which creates cyclical effects downstream with future grants, peer review, evaluation committees, the list goes on.
The sane target is resilient capacity: keep RFR above replacement in foundational areas, use automation to raise researcher leverage, and build flexible on-ramps so curious people can try science without waiting seven years first.
The Realistic Path Forward
Gen Z won’t buy into the current system. As it stands, they’re not really interested in saving it. If we’re concerned about the way research moves forward, we need to build something to capture their talent before they write off research entirely.
The fertility parallel holds: once a generation decides something isn’t worth it, no amount of policy can fully reverse that cultural shift.
We have maybe a few years to figure this out before the cultural narrative sets permanently.
Clock’s ticking.
The Control Group
Resources for researchers who refuse to be part of a failed experiment, or things I found particularly interesting.
Dean Lee’s “De-Risk Your PhD” Series (LinkedIn) - Practical advice from someone who completed the PhD, tips that nobody tells you about to win the game of academic survival
Renaissance Philanthropy + ARIA’s FRCL (announcement) is creating a “third way” between academia and industry, called a FRC that mixes contracts and grants. They’re early days and just announced their launch, and accepting applications!
To calculate the “replacement threshold” — that is, the ratio of researchers in a position to become new principal investigators (PIs) to the number of PI slots actually opening up per year, we need to focus on the real academic bottleneck: PI (faculty) job openings, not the number of fresh PhDs.
1. Number of PI Slots Opening Up Annually
> Active U.S. PI workforce: ~45,555 (NIH/clinical research estimate).
> Annual PI turnover/retirement rate: 3% is a typical estimate for faculty retirement or departure.
> PI jobs opening per year:
45,555 × 0.03 = 1,367
So, about 1,370 PI jobs open up annually, mainly via retirements and some additional institutional growth.
2. Researchers Eligible for PI Roles
New research PhDs awarded (2023): ~57,862.
However, not all new PhDs are immediately eligible for PI roles. The true “pool” includes:
> Current postdocs reaching the end of their term
> Adjunct/contingent faculty applying for TT roles
> Some mid-career researchers switching institutions or disciplines
NSF and higher-ed hiring data suggest roughly 10,000–15,000 competitive applicants are on the market for PI roles each year, given U.S. research PhD/postdoc flows. However, most job market data show that the number of new hires for tenure-track/PI roles is about 3,000 annually, significantly lower than the pool of applicants.
Nathan Lee has an interesting piece that touches on the meaning crisis and the cultural decay in venture capital, from the perspective of a young person.
See Footnote #1.
Clearly ARC Institute and FROs exist as an alternative to research, and is very, very close to what academic research is in its best case. If this is true, doesn’t that mean we’re fine? As Patrick Collison noted in his 2023 chat with Dwarkesh, new vehicles like focused research organizations (FROs) and the Arc Institute give PhDs high-agency, well-resourced roles that fix some pain points (long time-to-independence, weak tooling, brittle funding). Encouraging, but still early: headcounts are modest; we don’t yet have longitudinal rates (placement into PI-like leads, output/impact, spinouts); coverage is narrow (biomed > other fields; Bay Area > everywhere else); and there’s the upstream hump of getting the “phase-2” Gen-Z cohort (born after 2003) to start PhDs at all after 2025’s morale shock in US research funding.
This is such a good read. Sharing with my son, who is a postdoc, ASAP.
Certainly it's the economics that's the main problem here, not the culture? The interesting bit about declining birth rates is how you can't just pay people to have kids, but surely the success of high-paying AI labs (and before them, FAANG engineer jobs) is evidence that technical people are incentivized to do research/technical work by good pay.