Continuous Funding: What If Scientists Were Paid After Every Useful Result?

Getting your Trinity Audio player ready...

Scientists are usually funded in large, infrequent decisions. A researcher writes a proposal, waits through peer review, and—if selected—receives enough money for a project lasting several years. Once the grant ends, the researcher must compete again.

Continuous research funding proposes a different rule: whenever a scientist produces a verifiable result that is useful to science, the funding system issues an appropriate reward.

The result need not be a famous breakthrough. It could be a reusable dataset, a corrected theorem, an improved experimental protocol, a negative result that prevents duplicated work, or software that makes other research possible.

This would not eliminate grants or salaries. Some research must be financed before results exist. Instead, continuous funding could add a missing feedback mechanism: scientific value demonstrated after the work should influence where money flows next.

What Is Continuous Research Funding?

Continuous research funding is a result-responsive model in which scientists receive repeated payments as their contributions are published, verified, reused, extended, or shown to be valuable.

A simplified cycle would be:

1. A researcher publishes an identifiable scientific output.
2. The output is checked for validity, originality, documentation, and relevance.
3. Evidence of scientific utility accumulates.
4. A funding system assigns a proportional reward.
5. Later evidence can produce additional payments.

The payment would therefore be neither a conventional salary nor a single prize. It would be a continuing economic response to useful work.

This idea is related to retroactive public-goods funding, but it operates at a finer resolution. Retroactive funding often evaluates a project after a substantial period. Continuous funding could reward individual contributions throughout a researcher’s work.

Why Traditional Grants Are Not Enough

Traditional grants solve an indispensable problem: researchers need laboratories, equipment, data access, salaries, and time before they can produce results.

The National Institutes of Health, for example, describes research-project grants as support for costs such as salaries, equipment, supplies, travel, and institutional expenses. NIH grant payments are generally made in advance rather than being calculated after each scientific output.

This prospective structure is necessary, but it has limitations.

Proposals reward predictions

Grant committees must estimate which proposed projects will succeed. Yet a persuasive proposal is not the same thing as a useful scientific contribution. Applicants may be rewarded for presenting confidence, fashionable topics, institutional credibility, and carefully optimized plans—even when real discovery will require changing those plans.

Funding decisions are discontinuous

A scientist may produce ten useful intermediate results during a five-year project, but the funding system normally does not respond ten times. The principal financial decisions occur before the project and during occasional renewals.

Small contributions can disappear

Scientific progress depends on more than headline discoveries. It also depends on:

• proofs of specialized lemmas;
• replications and failed replications;
• curated datasets;
• repaired software;
• improved measurements;
• taxonomies and standards;
• documentation;
• null results;
• corrections of published errors.

These outputs can create real value without being large enough to win a conventional prize or justify an independent grant.

Recognition often arrives too late

Citations, awards, and appointments may eventually recognize an important contribution, but the delay can be measured in years. A researcher still needs resources while the work is developing.

Continuous funding would shorten the distance between demonstrated usefulness and financial support.

What Counts as a Useful Scientific Result?

The central difficulty is not making payments. It is defining and measuring usefulness without reducing science to a crude popularity contest.

A useful result might be:

• A validated discovery: a new empirical observation, theorem, material, mechanism, or method.
• A reusable research object: software, data, samples, protocols, formalizations, or benchmark suites.
• A corrective contribution: an identified error, failed replication, revised proof, or better uncertainty estimate.
• An enabling result: infrastructure or theory that permits other researchers to work more effectively.
• A clarifying negative result: evidence that a plausible approach does not work under specified conditions.
• A synthesis: a rigorous connection between previously separate results or disciplines.

Scientific outputs can already be assigned persistent identifiers. DataCite supports identifiers and metadata for datasets, preprints, software, samples, images, and other research resources—not only journal papers. This infrastructure makes a broader definition of scientific output technically feasible.

The payment unit should therefore be the verifiable contribution, not necessarily the paper.

How a Continuous Funding System Could Work

Register each contribution

Every eligible output would need a persistent record containing:

• its authors and contributors;
• publication date and version history;
• relevant data or code;
• declared dependencies;
• funding sources;
• licenses;
• subsequent corrections;
• links to related outputs.

Existing infrastructure such as DOIs, ORCID identifiers, research-organization identifiers, repositories, and open citation graphs could provide much of this foundation.

Crossref already operates a grant-linking system intended to connect awards with research outputs through structured metadata. In January 2026, Crossref also added support for including grant identifiers across its record types, improving the technical ability to connect funding with publications and other outputs.

Evaluate several dimensions

No single metric should determine payment. A continuous funding model could combine evidence concerning:

• validity;
• originality;
• reproducibility;
• demonstrated reuse;
• explanatory power;
• practical utility;
• contribution to later work;
• quality of data and documentation;
• difficulty of the problem;
• correction of important errors;
• underserved scientific value.

Citation counts could be one signal, but not the governing signal. Citations can indicate attention without proving correctness, importance, or positive dependence.

Issue an initial reward

After minimum verification, the researcher could receive a modest payment. This would recognize the contribution without pretending that its long-term importance is already known.

The reward should generally be small when evidence is limited. That reduces the damage from mistaken evaluations and makes it possible to fund many contributors rather than placing an entire budget behind a few predicted winners.

Add later rewards as evidence develops

A result could earn further funding when:

• independent researchers reproduce it;
• another project reuses its code or data;
• later work depends on its theorem;
• a corrected version resolves an important defect;
• practitioners adopt the method;
• expert evaluation confirms its significance;
• the contribution becomes part of scientific infrastructure.

This is the defining feature of continuous funding: evaluation remains open as scientific evidence changes.

Divide rewards among dependencies

Scientific outputs rarely arise in isolation. A new result may depend on a dataset, theorem, library, instrument, and earlier experimental method.

A well-designed system could distribute part of each reward upstream. When a highly useful discovery depends on an obscure but essential earlier contribution, the earlier contributor would receive a share.

This would create a scientific contribution graph rather than a winner-takes-all leaderboard. The AI meritocracy model for research funding proposed by Science DAO describes such a system as an open network in which scientific dependencies and multiple dimensions of merit influence funding.

Continuous Funding Is Not the Same as Milestone Funding

The two models overlap, but they answer different questions.

Model	Main question	Payment timing	Evaluation target
Traditional grant	Is this proposal worth supporting?	Before and during research	Planned project
Milestone funding	Did the contractor complete the agreed step?	At predefined stages	Contractual milestone
Prize funding	Who solved the specified challenge?	After success	Final achievement
Retroactive funding	What past work created value?	After impact becomes visible	Completed project or contribution
Continuous funding	What useful result has appeared now, and how has its value changed?	Repeatedly	Evolving scientific output

Milestone payments are usually negotiated in advance. DARPA materials, for example, describe fixed payable milestones as payments based on successful completion of agreed accomplishments. DARPA programs are also commonly structured in phases.

Continuous funding need not know the result in advance. It could reward an unexpected theorem, a newly discovered limitation, or a dataset that becomes valuable for a use nobody originally predicted.

The Main Benefits

Less dependence on proposal-writing skill

Researchers would still need proposals for advance funding, but proposals would no longer be the only gateway to resources. A person who has already produced useful work could receive funding on the basis of evidence rather than institutional promises.

Better support for independent researchers

An unaffiliated scientist may be unable to satisfy the administrative requirements of major grants. A result-based system could assess an openly published contribution regardless of where it was produced—provided identity, authorship, validity, and compliance can be verified.

Faster rewards for small results

A useful correction should not need to wait for a major prize. A specialized library should not need millions of users before its maintainer receives support. Continuous funding can match small contributions with small but timely payments.

Incentives for open and reusable work

Payments could increase when outputs are documented, machine-readable, reproducible, and openly reusable. This would make research quality economically relevant rather than treating documentation as unpaid secondary work.

Funding that follows unexpected discovery

Exploratory science frequently produces something different from what the original proposal predicted. Continuous funding can respond to the actual result instead of judging it against an obsolete plan.

Serious Risks and Failure Modes

Continuous funding is not automatically fair. Poor implementation could generate new distortions.

Metric gaming

Researchers might divide work into artificially small outputs, trade citations, generate low-value datasets, exaggerate dependencies, or optimize titles and abstracts for automated evaluation.

Countermeasures would require duplication detection, contribution clustering, graph analysis, adversarial testing, expert audits, and penalties for coordinated manipulation.

Popularity bias

Widely used work is not always more scientifically important than deep foundational work. Some mathematical and theoretical results may remain underused for years before their significance becomes visible.

Funding formulas therefore need protected channels for long-horizon, foundational, minority, and negative-result research.

Unsafe automation

An AI model may misread a paper, overestimate novelty, miss a fatal methodological flaw, or reproduce biases present in its training data.

AI can assist with evidence extraction, comparison, classification, dependency discovery, and anomaly detection. It should not be treated as an infallible scientific judge. High-value or disputed decisions need human review, transparent reasoning, appeal procedures, and public audit trails.

Unstable researcher income

Scientists cannot pay rent entirely through unpredictable micro-rewards. Continuous payments should supplement stable employment, fellowships, institutional funding, and prospective grants—not replace every predictable source of income.

Long-term fellowships remain valuable precisely because they give researchers freedom and stability. The Royal Society, for example, describes its University Research Fellowships as long-term, flexible support for building an independent research career.

Neglect of unsuccessful but necessary exploration

A careful experiment may fail to discover the hoped-for effect while still being scientifically well designed. A system that pays only for successful positive findings would intensify publication bias.

The model must distinguish useful negative knowledge from careless or undocumented failure.

Why a Hybrid Funding Portfolio Is Better

No single funding mechanism can serve every kind of science.

Continuous result-based payments are strongest when outputs are visible, attributable, and independently assessable. They are weaker when research requires substantial capital before any public result can exist.

A resilient system would combine:

• grants for uncertain preliminary work;
• fellowships and salaries for stability;
• milestone funding for defined technical programs;
• prizes for specific hard targets;
• continuous payments for verified contributions;
• retroactive funding for long-term public value.

Economic analysis of scientific grant funding has similarly emphasized that grants, prizes, patents, and procurement serve different functions, with grants particularly important for exploratory early-stage research.

The correct objective is not to replace all grants with one algorithm. It is to build a portfolio in which each funding mechanism addresses a different failure mode.

AIIM as a Proposed Continuous Funding Engine

AIIM, the AI Internet Meritocracy, is a proposed system for evaluating scientific contributions and distributing funding according to demonstrated merit.

Its intended role is broader than selecting a few annual prize winners. AIIM could repeatedly evaluate open research outputs, estimate their relationships to other contributions, and allocate available donations across a scientific dependency network.

In principle, this would allow:

• one small reward after initial verification;
• further rewards after independent confirmation;
• additional payments when downstream research depends on the result;
• partial rewards to important upstream contributors;
• reassessment when an output is corrected or challenged.

Science DAO describes AIIM as an automated scientific prize, but “prize” should not imply a single competition with one winner. The more ambitious concept is a continuously updated allocation mechanism.

This remains a proposal requiring empirical validation. Before handling substantial funds, such a system would need open evaluation criteria, red-team testing, governance safeguards, financial controls, conflict-of-interest rules, and accessible appeals.

The Real Change: Paying for the Scientific Process

Modern science often rewards positions, proposals, completed papers, and rare prestigious victories. Continuous funding would add another layer: repeated compensation for the many verifiable contributions through which knowledge actually grows.

The guiding principle is simple:

A useful scientific result should not need to become famous before it becomes fundable.

A corrected proof may deserve a small payment. A reusable dataset may deserve another. A foundational tool may continue receiving rewards as new work depends on it. A major discovery may receive much more—but its supporting contributors should not be economically invisible.

Continuous funding would not remove uncertainty from science. It would make funding more responsive to evidence already created.

The strongest model is therefore not “pay only after success.” It is:

fund promising exploration before the result, reward useful contributions when they appear, and continue updating those rewards as their scientific value becomes clearer.