the idea
the fro as institutional technology — the lineage it inherited, what the model proved, what stayed unfinished, and who carries it next.
lineage
- 1940s–osrd and the program managervannevar bush's osrd — 'science needs structure, not just money' — and the empowered program-manager model that became darpa (1958). convergent cribbed the milestones and the manager, then bound them to a nonprofit chassis.
- 1925–2000sthe corporate labsbell labs and xerox parc as the lost systems-research homes; the arora-style decline-of-corporate-science evidence and clancy's corporate-lab analogy supply the 'what was lost' half of the argument.
- 1950s–the philanthropic instituteshhmi's janelia (the people-not-projects contrast), broad, the lmb, max planck, the mrc — the institutions that bet on outsiders and gave the field its existence proofs.
- 1990–the megaprojectsthe human genome project, the lhc, hubble — the aspiration anchors. fros are pitched as reproducible mini-versions: 'mini manhattan projects' instead of once-a-generation megaprojects.
- 2018–2021the science-reform wavethe replication-crisis and diminishing-returns literature (collison and nielsen in the atlantic, collison and cowen), covid's fast grants, the day one project, and the parallel org births — new science, arcadia, arc, impetus, parpa — that ricón's genealogy documents.
- 2019–2021the boyden-lab linethe direct causal chain: boyden-lab brain-mapping failures → rodriques' thesis chapter → marblestone's fas fellowship interviews (with kalil) → the day one memo → the schmidt futures network spinout. the idea's genealogy is unusually traceable.
what convergent proved
- fros can be built at allroughly ten us fros plus the uk cohort in about four years, 'almost $400m' raised from '30+ individuals and institutions' (self-reported), and elite talent recruitable — 'nobel laureates approach us.'
- the public goods shipprism (>750x label diversity, 8x tracing accuracy), plexdia's ~30x parallelization, lean improvements underneath deepmind's imo-silver system, [c]worthy's oae efficiency map, cultivarium's organism portal and growth prediction, eve bio's pharmome dataset, forest 1, and the mothra array.
- the vocabulary spread'fro-shaped problem' entered the vocabulary — policy adoption by aria, the uk's research ventures catalyst, nsf tech labs, the white house ai action plan, and ifp's x-labs proposal.
- the incubator can be productizedthe gap map, essential technology, and the aria founder residency — roughly a hundred applications distilled to five teams and two launched fros — amount to a cohort model of institutional manufacturing.
- the structure generates ideasmarblestone's claim, via reinhardt's bottling lightning: 'the mere existence of focused research organizations as a new funding structure has led to people thinking of ideas that wouldn't even have become conceivable.'
what stayed unfinished
- one fro, one benefactora $20–50m philanthropic ticket per fro caps the count against a claimed 100–200-problem backlog. buck's bbn critique names fundraising itself as the bottleneck.
- midstream money is unsolvedbudgets must be front-loaded because there is no next-funding-round kpi — the field notes concede it directly.
- the founders needed managingmilestone design was reworked from darpa-rigid to 'north star' plus adaptive goals, scientist-founders needed executive coaches, and dissemination turned out to require 'a kind of product development.'
- the endgame is untestedchi-fro's move to the probabilistic computing foundation is the first non-death transition; the sunset doctrine was publicly walked back; who owns the tool at year seven remains unpriced.
- the flagship output is partialprism labeled roughly 49% of cells and about one in five labeled neurons shared barcodes — 'suited to partial reconstruction,' per mckenzie's technical read, not yet dense connectomics.
- the missing counterfactualno randomized or counterfactual comparison against equivalent grant or arpa spend exists; buck's 'create variation' call is unanswered.
the honest ledger
- forsome science needs an organization, not a grant — coordinated engineering beyond academia's incentives and below industry's profit threshold. fros are the missing middle between grants, startups, megaprojects, and arpa portfolios (the day one memo; the nature comment).
- fora third career track: professional researchers who are not trainees. 'nobel laureates approach us' — talent acquisition, the field notes say, was easier than feared.
- againstnever measured against the counterfactual — no randomized or matched comparison of an fro against the same money spent as grants or an arpa program. the critiques so far are structural, not empirical (buck's advice for funders; mastroianni's strong-link frame).
- againstwho gets to found an fro? a $20–50m fundraising threshold filters founders, and the alternative-funder wave risks reproducing the credentialing networks it was built to escape (buck's bbn and slow-cancellation critiques).
- againstthe focus is the gap — time-bound, narrow-scope fros leave 'a lot of work and many types of people that they won't cover' (reinhardt's bbn post).
- nuancebbns and parpas are complements, not competitors — buck's contract-research shops and reinhardt's program-manager model cover the work fros structurally can't.
- nuancefederal adoption may drift — nsf's tech labs language ('transition... to commercially viable platforms') hints the fro could lose its nonprofit, public-goods, and sunset features on the way to scale.
- againstscience is a strong-link problem — progress depends on the best work, which funding-structure reform may not touch. the frame's applicability to fros is inferred, not asserted (mastroianni).
successors
- speculative technologiesthe parpa made independent — ben reinhardt's program-manager model as its own nonprofit, the complementary template the discourse uses to define what a fro is not.
- astera institutefunder and incubator — ran rejuvenome, incubated forest and [c]worthy, hosted the parpa, and employed marblestone as consulting research director for longevity.
- bind researchthe independent uk fro — disordered proteins, competitively funded through research ventures catalyst, outside the convergent line.
- futurehouserodriques' own shop — a private ai-for-science lab that sits outside the canonical fro line; the co-inventor's second opinion.
- renaissance philanthropykalil's post-schmidt vehicle — the 'big if true science' accelerator and the xtx ai for math fund; convergent's sibling in the same orbit.
- nsf tech labsthe federal adoption — nsf tip's tech labs pilot, up to $1b for fro-style independent labs; convergent answered the rfi and offered to be implementation partner.
- aria frost ukaria's frost-uk program plus the founder residency — the state-partnered cohort model that produced meridial and echo labs.
- ifp's x-labsthe institute for progress's x-labs proposal — x02 execution awards as fros, x03 block grants as incubators; the mission-control pattern generalized into policy language.
- the parallel cohortnew science, arcadia, arc institute, impetus — the new science organizations ricón's genealogy files convergent beside.
- the comparative brain connectome initiativemarblestone and payne's ifp proposal for a federally funded fro to map whole mammalian brains — the connectomics fro pitched at government scale.
open questions
- does an fro beat the same money spent as grants or an arpa program — and will anyone ever measure it?
- who owns the public good at year seven — post-sunset maintenance is unpriced, and only chi-fro has tested a transition.
- does federal adoption through nsf tech labs preserve the fro features — nonprofit, public goods, sunset — or drift toward commercialization?
- how large is the fro-shaped problem space — convergent estimates 100–200; the gap map is the live count.
- can the incubator model itself replicate beyond convergent's schmidt-anchored balance sheet?
- where do the openness norms land — cc-by-nc licenses versus true public domain?
AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.