free to read. paid to cite.
Bucket turns primary research into three kinds of impact. Every paper, report, figure, dataset, tool and doc below belongs to one of them.
93 of 93 items
Education reform
Open the path from consuming knowledge to producing it, for every learner.
- paper2026-06-25The Knowledge-Access Gradient
Who can reach how deep into human knowledge, at what age, at what cost, and what that means for reforming education
- figure2026-06-25The five-thousand-year access arc
From The Knowledge-Access Gradient.
- figure2026-06-25Access to knowledge by age
From The Knowledge-Access Gradient.
- figure2026-06-25The continuity funnel from school to research
From The Knowledge-Access Gradient.
- figure2026-06-25Researcher capacity by country
From The Knowledge-Access Gradient.
- docThe Global Education Problem Landscape
Bucket's founding problem statement, grounded in authoritative data: what the educational problems actually are, by country and level, measured against UN SDG...
- docThe Reform Thesis
The bridge from diagnosis to mission: given the actual problems, where does Bucket's open-knowledge thesis measurably move the needle, and where does it not? H...
- docWhat it means to be educated
The irreducible, contested value question underneath every reform: there is no single cross-civilizational definition of an educated person.
Scientific discovery with ML
Find the open problems within reach of current methods and give researchers instruments to attack them.
- paper2026-06-19The structure of public research funding, 2015-2025: concentration, cross-funder co-funding, and the funding→output relationship in a reconciled NIH/NSF/EC/UKRI grant graph
Institutional grant concentration is extreme: Gini = 0.929; the top 1% of institutions hold 53.6% of all grants.
- paper2026-06-23The transformer paper-recommendation advantage is real at the head of the impact distribution and decays to null across the broad literature: a 4-checkpoint, all-26-field convergence study of SPECTER vs TF-IDF
SPECTER's across-field edge over TF-IDF is large and nearly significant on the most-cited core (+0.0095 ΔMAP, 16/26 wins, p = 0.062 at checkpoint 1) and decays...
- paper2026-06-24What each funder funds: specialization, complementarity, and the surprising temporal stability of funder field-portfolios in a reconciled NIH/NSF/EC grant→output graph
Funders span a clean 5× specialization gradient: generalists EC (HHI 0.092) and NSF (0.095) at one end, the hyper-specialist NHGRI (0.466, 66% in one field) at...
- paper2026-06-24Public funding and researcher careers: funder portfolios, career-stage composition, and why the funded-vs-unfunded productivity gap is mostly selection, not effect
The headline is a refusal: the naive 2.1× funded-vs-unfunded publication gap is mostly selection, matching on field × career-stage × entry-era collapses it to...
- paper2026-10-08Machine learning for scientific discovery: the solvability frontier
Solver Gap Engine, 2026-09-30: 3,599 formal-conjectures problems mapped, 1,364 open problems ranked, 31 at cross-file similarity 0.8 or above, median 0.586 aga...
- figure2026-10-08The frontier circle over 5,080 problems
From Machine learning for scientific discovery.
- figure2026-10-08What sits inside and outside the frontier
From Machine learning for scientific discovery.
- figure2026-10-08Backtest of the reach rule at two cutoffs
From Machine learning for scientific discovery.
More capable humans
Shorten the route to mastery with learning systems that build on what a person already knows.
- paper2026-10-07Human-AI-computer interaction: measuring whether AI strengthens unaided judgment
Four scores on the same items: Hp, A, Jp at day 0 and Hp again at day 7; mp = Jp / max(Hp, A), Rp = Hp(t + 7) - Hp(t), L = Jp(t + 7) - Hp(t + 7).
- figure2026-10-07Learning timeline with and without the system
From Human-AI-computer interaction.
- figure2026-10-07Mastery scores by condition
From Human-AI-computer interaction.
- figure2026-10-07The knowledge region a learner holds
From Human-AI-computer interaction.
- toolResearch OS
A workspace where the AI finds, quotes and checks, and the learner writes.
- toolBucket Academy
Spaced practice, diagnostic placement and a grounded tutor.
- docManifesto
The thesis: foundations, axioms and a small number of brilliant humans with AI.