A living research series · 19 articles
The R&D Ratchet
How America lost the institutions that could think long—and the narrow window to build what comes next.
A firsthand, evidence-led inquiry into the incentive structures that replaced the American industrial research laboratory, what those structures preserve and destroy, and how an AI-native public–private successor might be designed.
The proposition
The history and trajectory of these institutions provide the evidence. The incentive structures are the subject.
America did not simply defund research. It replaced one institutional equilibrium with several narrower ones — corporate returns tied to product, platform, and capital-market horizons; finite government programs; billable-project economics that can reward proposal and administrative throughput; publication metrics; patent accounting; and venture-backed commercialization optimized for acquisition or IPO. While each decision may look rational and may even be productive in the short term, the end result is a combined system that has consumed inherited deep-research capabilities faster than it has replenished them.
The spending paradox
A near-trillion-dollar flow does not reveal the state of the system
National accounts measure activity far better than they measure retained teams, tacit knowledge, apprenticeships, facilities, or the ability to carry an uncertain idea across institutional borders.
Estimated U.S. R&D, 2024
$993.4B
Current dollars and estimated composition. Source and qualifications
What the total cannot tell us
- 01
Continuity Did a capable team survive between projects?
- 02
Memory Were failed paths and experimental judgment retained?
- 03
Integration Could research cross hardware, software, product, and mission boundaries?
- 04
Transition Did anyone own the work after the prototype?
U.S. R&D grew; its composition tilted toward development
Billions of constant 2017 U.S. dollars
Move across the chart or focus it and use the arrow keys to inspect values.
View the chart data as a table
| Year | Basic research | Applied research | Experimental development |
|---|---|---|---|
| 2000 | 57.8 | 77.7 | 233.0 |
| 2005 | 73.4 | 85.5 | 239.9 |
| 2010 | 85.3 | 88.0 | 280.3 |
| 2015 | 86.7 | 99.8 | 321.6 |
| 2020 | 105.9 | 125.6 | 448.4 |
| 2024 | 115.5 | 143.9 | 532.7 |
Shared analytical vocabulary
Ten Propositions The Series Makes Measurable
These terms are not conclusions disguised as definitions. Each names a proposed mechanism and the observation that could support, narrow, or defeat it.
-
Accounting
Capability Stock
Reusable teams, tools, memory, judgment, relationships, and transition options that persist beyond one output or project.
Empirical testMeasure renewal, depreciation, reuse, and reconstitution cost beside annual spending. -
Reliability
Effective Horizon
The time for which every indispensable, non-transferable complement can remain available—not the nominal length of the longest grant.
Empirical testIdentify the shortest-lived minimum cut between question and validated use. -
Capability
Adjacency Capital
Accumulated reductions in the cost of routing expertise, exposing errors, coordinating unlike roles, and translating across layers.
Empirical testCompare experienced and reconstituted teams on routing time, rework, correction, and transition. -
Reliability
Handoff Loss
State lost when people, artifacts, assumptions, authority, or responsibility cross an institutional boundary.
Empirical testTrack end-to-end continuation and reconstruction across every funded transfer. -
Reliability
Common-Mode Exposure
Apparently separate projects or evaluators that share the same model, data, metric, method, vendor, incentive, or transition bottleneck.
Empirical testDisclose model, evidence, method, and incentive distance before claiming independent confirmation. -
Knowledge
Epistemic Debt
The gap between downstream confidence and dependence placed on a claim and the verification that claim has actually received.
Empirical testMaintain typed evidence states, dependency records, correction latency, and downstream repair. -
AI Age
Abundance Inversion
As hypotheses, prose, code, and scientific form become cheap, independent witnesses, accountable attention, and physical evidence become scarcer.
Empirical testMeasure verification queues and validated state changes rather than generated output volume. -
Governance
Charter-to-Work Drift
A long-horizon institutional name or legal mandate persists while actual work, authority, or staffing narrows toward short tasks.
Empirical testCompare the charter with who initiates work, what survives tasks, and which capabilities are renewed over time. -
Transition
Residual Capability Owner
The institution still responsible after the paper, grant, program, degree, transaction, or prototype has satisfied its local obligation.
Empirical testName the payer, custodian, transition owner, maintenance duty, review date, and termination rule.
Twenty-Six Years Inside a Thirty-Year Transformation
My Route Through The Incentive Systems
The personal record begins in 2000. The institutional arc begins with the 1996 Bell Labs split and continues through the transactions reshaping PARC and HRL.
-
DDocumented
Bell Labs Split
AT&T spun most of Bell Labs and its equipment business into Lucent; a smaller research organization remained with AT&T.
Source -
MFirsthand
Ericsson, Sweden
Karim’s internship across Ericsson’s Linköping and Kista facilities; Kista was Sweden’s “Silicon Valley” telecommunications cluster.
-
MFirsthand
Lucent / Bell Labs
Karim’s summer internship during the telecom contraction.
-
MFirsthand
Cisco Systems
Karim’s internship in San Jose, adding a product-company view of Silicon Valley’s technology system.
-
MFirsthand
AT&T Research
Research internship during graduate school, enabling comparison between two descendants of the Bell system.
-
MFirsthand
HRL Laboratories
Mission-driven industrial R&D spanning secure systems, privacy, cloud infrastructure, biometrics, and cyber-physical systems.
-
MFirsthand
SRI International
Full-time from 2017 through the 2021 Confidencial.io spinout, then part-time through early 2023 while finishing DARPA programs.
-
MFirsthand
Confidencial.io
Spun out from SRI in 2021 to carry selected research into product, procurement, customers, and company survival.
-
MFirsthand
Confidencial.io Seed Financing
Firsthand participation as Confidencial.io co-founder; SRI's June announcement independently documents the seed investors and private-beta launch, and is not the spinout date.
Source -
MFirsthand
PARC Joins SRI
Firsthand observation through Karim's SRI office and badge, retained through 2026; the legal transaction facts remain independently documented.
Source -
DDocumented
IBM Announces HRL Agreement
The agreement was announced July 23 and remained subject to closing conditions at the time of this page’s update.
Source
Comparative institutional analysis
Nine Incentive Structures, Nine Different Bargains
Select a model for a concise reading, then use the table to compare the same variables across every regime.
Classical Bell Labs
Vertically Integrated Corporate Laboratory
- Who Pays
- Regulated and market revenues inside an integrated operating and manufacturing system
- Real Horizon
- Decade-scale when the parent system is stable
- What It Preserves
- Cross-layer teams, apprenticeship, facilities, and patient problem selection
- Characteristic Failure
- The research cross-subsidy depends on an economic and regulatory settlement that may not survive restructuring
Compare Incentive Structures
| Model | Payer | Real Horizon | Preserves | Characteristic Failure |
|---|---|---|---|---|
| Vertically Integrated Corporate LaboratoryClassical Bell Labs | Regulated and market revenues inside an integrated operating and manufacturing system | Decade-scale when the parent system is stable | Cross-layer teams, apprenticeship, facilities, and patient problem selection | The research cross-subsidy depends on an economic and regulatory settlement that may not survive restructuring |
| Post-Breakup Equipment-Company ResearchLucent-Era Bell Labs | Equipment revenues and capital markets | Strategic in principle; exposed to product and industry cycles in practice | Deep hardware and systems expertise near manufacturing and product groups | Parent-company distress can compress the horizon faster than research culture can adapt |
| Network-Operator Research LaboratoryAT&T Labs Research | Communications-services business | Medium to long, with continuing pressure for operational relevance | Real operational problems, data, and feedback from deployed infrastructure | Broad spillovers may escape the sponsor while corporate priorities narrow the research surface |
| Owner-Sponsored, Mission-Driven Industrial LaboratoryHRL Laboratories | Strategic owners and government mission sponsors | Program clocks nested inside longer-lived technical capabilities | Specialized facilities and ambitious applied science beyond consumer-market horizons | Sponsor boundaries and contract portfolios can turn durable capability into disconnected projects |
| Independent Nonprofit Contract-Research InstituteSRI International | Government and commercial contracts, plus limited internal investment | Finite projects carried by an institution intended to persist across them | Multidisciplinary teams and networks across sponsors and performers | Chargeability, proposal churn, and funding cliffs can become the institution’s hidden operating system |
| Time-Bounded Agency Program NetworkDARPA, DHS, IARPA, ARPA-H Programs | Public mission agencies | Finite programs led by empowered program managers | Urgency, cross-organizational coordination, and permission to attempt high-risk work | The program can create a temporary laboratory without funding a permanent institutional home for what it learns |
| University Research SystemResearch Universities | Public grants, tuition, philanthropy, and endowments | Grant, degree, and tenure clocks | Open knowledge, intellectual diversity, and the talent pipeline | Integration, maintenance, negative results, and long transition work are frequently under-rewarded |
| Venture-Backed Technology TransitionResearch Spinouts and Startups | Investors and customers | Runway, market adoption, and financing cycles | Concentrated execution on one transition path | A startup cannot rationally maintain the broad research commons from which its product emerged |
| AI-Native Public–Private LaboratoryProposed Design; Informed by NAIRR, NSTC, FFRDCs, DARPA, and DOE Laboratories | Long-charter public capital, industry membership, mission procurement, and philanthropy | Rolling ten- to fifteen-year mission charters with layered review | Shared infrastructure, institutional memory, apprenticeship, and complementary transition assets | Must be designed against political capture, incumbent capture, bureaucracy, mission drift, and performative metrics |
Causal argument map
How locally rational incentives can produce systemic capability loss
Select a node to isolate its claim, evidence classes, and causal neighborhood. Connections represent propositions the articles must substantiate—not settled facts.
Inherited capability
Replacement incentives
System behavior
Capability loss
Observable outcome
Design pressure
Design response
Publication architecture
Nineteen articles—public working manuscripts and the book’s evidence base
The landing page remains the map, while every linked DRAFT is a live public manuscript with its central argument, evidence, objections, and falsifiers. The website is intended to carry roughly 60–70 percent of the eventual book’s substance; the book will add archival depth, interviews, extended cases, technical appendices, and larger-format visual analysis.
-
01 In research National R&D statistics; IBM/HRL; SRI/PARC
The Nearly Trillion-Dollar R&D System That Cannot Think Long
R&D expenditure is a flow; research capability is accumulated state.
Planned visual evidence
- Interactive R&D trend chart
- National capability balance sheet
- Source-backed argument map
-
02 In research Lucent / Bell Labs, 2001
I Arrived at Bell Labs as Lucent Technologies Was Coming Apart—Before Bell Labs Became Part of Nokia
Scientific excellence and corporate fragility can coexist.
Planned visual evidence
- Memory-versus-document timeline
- Corporate horizon diagram
- Annotated telecom context
-
03 In research Lucent and AT&T
One Heritage, Two Incentive Systems
Parent economics determine which parts of a research culture can survive.
Planned visual evidence
- Split-lineage map
- Side-by-side incentive table
- Problem-to-operation pathways
-
04 In research HRL Laboratories, 2010–2016
HRL and the Bargain of Mission-Driven Research
A durable mission customer protects ambitious work while also defining its boundaries.
Planned visual evidence
- Funding-and-mission map
- Capability-versus-project layers
- Patent-to-program network
-
05 In research SRI International, full-time 2017–2021 and part-time through early 2023
SRI and the Project-Funded Institution
An institution can persist across projects while its people remain governed by project clocks.
Planned visual evidence
- Program portfolio map
- Chargeability pressure loop
- Collaboration network
-
06 In research RACE, SIEVE, DPRIVE, DHS, and IARPA programs
DARPA Can Create a Temporary Laboratory. What Happens When the Program Ends?
The program model coordinates breakthroughs but does not automatically preserve the capability it assembles.
Planned visual evidence
- Temporary-lab lifecycle
- Program handoff map
- Artifact survival ledger
-
07 In research Academic basic research, grant cycles, publication incentives, and facilities and administrative costs
The University Is Not the Missing Laboratory
Universities are indispensable to discovery and training, but project grants and publication markets do not fund the complete institutional stack required for durable R&D.
Planned visual evidence
- Academic R&D role map
- Grant-and-overhead feedback loop
- Direct-versus-shared-cost explainer
-
08 In research DARPA and ARPA-E program leadership; expert review; technical portfolio governance
The People Who Choose the Bets Must Have Made Them
Research judgment is a practiced skill, so people with firsthand research and multi-year program experience must hold real authority over technical portfolios—without turning publication or patent counts into a credentialing machine.
Planned visual evidence
- Research-governance authority map
- Four-part experience dossier
- Expertise-and-capture control matrix
-
09 In research Microsoft Research Silicon Valley; IBM Almaden; IBM Microelectronics and GlobalFoundries
A Laboratory Is an Adjacency Graph, Not an Address
A laboratory survives only when its working relationships, complementary roles, technical assets, memory, and problem-selection authority remain connected—not merely when its building, payroll, or name persists.
Planned visual evidence
- Capability graph before and after closure
- Node-survival versus edge-survival map
- Post-reorganization capability ledger
-
10 In research PARC, Xerox, and SRI
PARC and the Appropriability Trap
Invention, recognition, complementary assets, and value capture are four different capabilities.
Planned visual evidence
- Four-capability matrix
- IP-and-host timeline
- Spillover map
-
11 In research Karim’s publication, patent, program, and transition record
Papers, Patents, Prototypes—and Vanishing Capability
Output metrics can rise while the ability to produce the next generation of work declines.
Planned visual evidence
- Output-versus-capability dashboard
- Test-of-time timeline
- Evidence-quality matrix
-
12 In research Confidencial and research spinouts
The Startup Is Not a Research Laboratory
A startup can complete one transition but cannot rationally maintain the research commons.
Planned visual evidence
- Paper-to-product work map
- Incentive comparison
- Missing-middle cost stack
-
13 In research SRI/PARC and the announced IBM/HRL transaction
Consolidation as Triage
Preservation should be measured by surviving teams and capabilities, not by the transaction announcement.
Planned visual evidence
- Before-and-after capability ledger
- Acquisition outcome tree
- Five-year scorecard
-
14 In research The remaining mentorship overlap
The Last Transfer Window
Institutional memory is perishable, making delay more costly than budget statistics reveal.
Planned visual evidence
- Mentorship-chain model
- Capability decay curves
- Window indicators
-
15 In research Mathematics; formal proof; the Microsoft-linked Majorana retraction; AI-assisted replication
When AI Audits What Science Takes for Granted
How much of the scientific record is verified, and how much is merely unchallenged? AI may expose epistemic debt, subtracting from accepted knowledge before it adds to it.
Planned visual evidence
- Epistemic-status ledger
- AI verification pipeline
- Audit-shock curve
-
16 In research Shannon's bandwagon; Feynman's cargo cult; Wigner's unreasonable effectiveness; data-driven AI and hallucination
Cargo Cult Science at Machine Speed
LLMs can industrialize the surface form of science—fluent prose, citations, equations, and experimental ritual—without increasing contact with truth; AI will amplify the epistemic standards of the institutions that deploy it.
Planned visual evidence
- Historical warning-to-AI map
- Mathematical-validity versus empirical-adequacy matrix
- Form-to-witness verification funnel
-
17 In research A proposed successor informed by NAIRR, NSTC, FFRDCs, DARPA, and DOE laboratories
An AI-Native Laboratory for the Public Good
A successor must align capital, careers, infrastructure, AI, transition, and public value in one institutional design.
Planned visual evidence
- Interactive lab brain map
- Governance and IP tiers
- Ten-year pilot roadmap
-
18 In research The transistor, computing, the Internet, software, cryptography, smartphones, AI, quantum technology, and the institutions that produced them
The System Worked. Compared With What?
A catalogue of breakthroughs proves productive capacity, not counterfactual efficiency, acceptable delay, or the sustainability of the institutions and developmental paths that produced them.
Planned visual evidence
- Achievement-versus-efficiency argument map
- Research-productivity benchmark
- Counterfactual pilot scorecard
-
19 In research Graduate students and early-career researchers navigating academia, industrial R&D, public missions, national laboratories, and research-based company formation
Do Not Restart at Every Boundary
When no institution supplies the whole developmental path, young researchers need a compounding sequence organized around one durable question—but universities, funders, laboratories, and employers must pay for the bridges rather than externalizing integration risk onto the least powerful people in the system.
Planned visual evidence
- Cross-sector doctorate destination chart
- Compounding-path argument map
- Boundary-contract checklist
- Seven-year pathway scorecard
Showing all 19 articles.
The narrow path forward
An AI-native laboratory is an institution, not a model with a building
The proposed successor joins six systems that are usually funded, governed, and evaluated separately. Select a node to inspect its institutional function.
NAIRR
Compute, data, tools, training, and public–private resource contributions.
EvidenceNSTC
Research, prototyping, commercialization, collaboration, and workforce development.
EvidenceFFRDCs
Continuity, institutional memory, special access, objectivity, and public-interest obligations.
EvidenceDARPA
Empowered program managers, ambitious goals, milestones, and cross-sector performer networks.
EvidenceInspectable claims
Source Ledger
-
Indicators 2026: U.S. R&D expenditures by type, selected years 2000–2024
National Center for Science and Engineering Statistics. Supports: R&D totals, constant-dollar trends, and the basic/applied/development composition.
-
Business R&D Performance in the United States Increases to $722 Billion in 2023
National Center for Science and Engineering Statistics. Supports: The scale, character-of-work mix, company-size distribution, and geographic concentration of U.S. business R&D performed in 2023.
-
Indicators 2026: Discovery—R&D Activity and Research Publications
National Science Board and National Center for Science and Engineering Statistics. Supports: The concentration of 2023 business R&D in five industries and three states, including reported R&D in software, biotechnology, artificial intelligence, and nanotechnology.
-
Bell Labs history
Nokia Bell Labs. Supports: The 1996 AT&T/Lucent split, later Bell Labs institutional transitions, and Nokia Bell Labs’ current count of ten Nobel Prizes and five Turing Awards.
-
IBM to acquire HRL Laboratories
IBM. Supports: The signed agreement, stated rationale, ownership at announcement, and pending closing status.
-
About SRI International
SRI International. Supports: SRI’s institutional history and PARC’s 2023 integration.
-
Xerox announces donation of PARC to SRI International
Xerox. Supports: The form, rationale, and announced structure of the PARC transfer.
-
Become a program manager
DARPA. Supports: Fixed program-manager tenure, portfolio authority, milestones, and performer coordination.
-
Federal Research—Management and Oversight of FFRDCs
U.S. Government Accountability Office. Supports: Long-term government relationships, institutional memory, objectivity, and special access in the FFRDC model.
-
NAIRR at two years
U.S. National Science Foundation. Supports: Shared AI resources, federal/private participation, research access, and the move toward durable infrastructure.
-
National Semiconductor Technology Center
National Institute of Standards and Technology. Supports: The public–private consortium, prototyping, investment, collaboration, and workforce model.
-
Technology Commercialization Fund
U.S. Department of Energy. Supports: Cost-shared lab-to-market partnerships and commercialization infrastructure.
-
The decline of science in corporate R&D
Ashish Arora, Sharon Belenzon, and Andrea Patacconi. Supports: The documented decline in scientific publication by large corporate R&D performers between 1980 and 2007 and the authors' analysis of narrower firm scope.
-
Are Ideas Getting Harder to Find?
Nicholas Bloom, Charles I. Jones, John Van Reenen, and Michael Webb. Supports: Evidence that rising research effort has accompanied declining measured research productivity in several domains, subject to the study's choice of exponential-growth benchmarks.
-
The Nobel Prize in Physics 1956
Nobel Prize Outreach. Supports: The award to John Bardeen, Walter Brattain, and William Shockley for semiconductor research and discovery of the transistor effect.
-
ARPANET
Defense Advanced Research Projects Agency. Supports: DARPA's role, the 1969 four-node network including SRI, the development of internetworking and TCP/IP, and the transition from ARPANET toward the Internet.
-
About the ACM A.M. Turing Award—Fact Sheet
Association for Computing Machinery. Supports: The award citations for John Backus on FORTRAN and practical high-level programming systems and for Ken Thompson and Dennis Ritchie on generic operating-systems theory and UNIX.
-
Cryptography Pioneers Receive ACM A.M. Turing Award
Association for Computing Machinery. Supports: The contribution of Whitfield Diffie and Martin Hellman to public-key cryptography and digital signatures and their downstream importance to Internet security.
-
Fathers of the Deep Learning Revolution Receive ACM A.M. Turing Award
Association for Computing Machinery. Supports: ACM's recognition of Yoshua Bengio, Geoffrey Hinton, and Yann LeCun for conceptual and engineering breakthroughs that made deep neural networks a critical component of computing.
-
Apple Reinvents the Phone with iPhone
Apple. Supports: The original iPhone's integration of mobile telephony, Internet communication, software, sensors, and a multi-touch interface; the announcement establishes the product claim and configuration, not sole invention of every underlying component.
-
Invention, Knowledge Transfer, and Innovation—Introduction
National Science Board and National Science Foundation. Supports: The distinction among discovery, invention, knowledge transfer, and innovation and the warning that available inputs and outputs provide an incomplete picture of innovation and its economic impact.
-
On Computable Numbers, with an Application to the Entscheidungsproblem
Alan M. Turing. Supports: Turing's model of general computation and proof that no general mechanical decision procedure solves every instance of the stated decision problem.
-
The Chemical Basis of Morphogenesis
Alan M. Turing. Supports: Turing's cross-disciplinary reaction-diffusion model showing how simple local dynamics can generate biological pattern.
-
A Mathematical Theory of Communication
Claude E. Shannon. Supports: The mathematical separation of source, channel, noise, coding, and receiver and the resulting limits on reliable communication.
-
The Bandwagon
Claude E. Shannon. Supports: Shannon's warning that fashionable relabeling and weak applications could dilute information theory and his call for rigorous research.
-
First Draft of a Report on the EDVAC
John von Neumann. Supports: An early public specification of stored-program computer organization; attribution of the underlying group contributions requires historical care.
-
The Electronic Computer Project
Institute for Advanced Study. Supports: The 1945–1957 project, its mixed public and institutional support, its engineering duration, its open reports, and the replication of its architecture.
-
Probabilistic Logics and the Synthesis of Reliable Organisms from Unreliable Components
John von Neumann. Supports: The use of redundancy and restoring mechanisms to synthesize reliable computation from unreliable components.
-
There's Plenty of Room at the Bottom
Richard P. Feynman. Supports: Feynman's technically grounded invitation to develop atom-scale manipulation, dense storage, and miniaturized machines before a mature field or roadmap existed.
-
Cargo Cult Science
Richard P. Feynman. Supports: Feynman's argument for reporting disconfirming evidence and practicing scientific integrity beyond the outward form of experimentation.
-
Simulating Physics with Computers
Richard P. Feynman. Supports: Feynman's argument that exact simulation of quantum physics motivates a computing machine governed by quantum mechanics.
-
R&D Spending at Federally Funded R&D Centers Surpassed $31 Billion in FY 2024
National Center for Science and Engineering Statistics. Supports: The number of FFRDCs, expenditure growth, funding concentration, and FY2024 R&D composition.
-
Federal Acquisition Regulation 35.017—Federally Funded Research and Development Centers
U.S. General Services Administration. Supports: The special long-term need, continuity, independence, access, sponsorship, and periodic-review requirements that define the FFRDC relationship.
-
Federally Funded Research and Development Centers: Background and Issues for Congress
Congressional Research Service. Supports: The February 2026 count and activity classification of FFRDCs, their share of federally funded R&D, their statutory and contractual rationale, and congressional oversight questions including mission drift and infrastructure.
-
Master Government List of Federally Funded R&D Centers
National Center for Science and Engineering Statistics. Supports: The current federally recognized FFRDC population and each center's sponsoring agency, administrator, location, and activity type.
-
Federal Research Centers: DHS Actions Could Reduce the Potential for Unnecessary Overlap among Its R&D Projects
U.S. Government Accountability Office. Supports: A bounded example of FFRDC work organized through task orders and weaknesses in tracking how sponsors used resulting deliverables.
-
Universities Report 8.1% Growth in R&D Expenditures in FY 2024, Reaching Over $117 Billion
National Center for Science and Engineering Statistics. Supports: Academic R&D totals, sources, fields, concentration, and the institutional share of university research funding.
-
Federal Research Grants: Opportunities Remain for Agencies to Streamline Administrative Requirements
U.S. Government Accountability Office. Supports: The administrative workload created by federal grant requirements and opportunities to standardize, postpone, or simplify pre-award requirements.
-
Biomedical Research: NIH Should Assess the Impact of Growth in Indirect Costs on Its Mission
U.S. Government Accountability Office. Supports: The functions and historical growth of NIH facilities and administrative reimbursements and the need to assess their effect on research capacity.
-
National Science Foundation: Preliminary Observations on Indirect Costs for Research
U.S. Government Accountability Office. Supports: NSF indirect costs as a share of annual award funding during FY2000–FY2016 and weaknesses in rate-negotiation documentation.
-
Universities and Indirect Costs for Federally Funded Research
Congressional Research Service. Supports: The history, cost definitions, negotiated-rate framework, 2025 agency proposals, 2026 statutory constraints, policy tradeoffs, and transparency questions surrounding university indirect-cost reimbursement.
-
Supplemental Guidance to the 2024 NIH Grants Policy Statement: Indirect Cost Rates
National Institutes of Health. Supports: NIH's attempted replacement of negotiated facilities and administrative rates with a standard 15 percent rate.
-
Commonwealth of Massachusetts v. National Institutes of Health
U.S. Court of Appeals for the First Circuit. Supports: The affirmance of the order permanently enjoining and vacating NIH's 2025 indirect-cost guidance.
-
Incentives and Creativity: Evidence from the Academic Life Sciences
Pierre Azoulay, Joshua S. Graff Zivin, and Gustavo Manso. Supports: A matched comparison associating HHMI's longer, more failure-tolerant investigator funding with more high-impact and exploratory work than conventional project funding.
-
Bias against Novelty in Science: A Cautionary Tale for Users of Bibliometric Indicators
Jian Wang, Reinhilde Veugelers, and Paula Stephan. Supports: Evidence that novel research has more variable impact, delayed recognition, broader cross-field influence, and disadvantages under short-window bibliometric evaluation.
-
The Natural Selection of Bad Science
Paul E. Smaldino and Richard McElreath. Supports: A formal population model in which institutional selection for publication can favor poor methods without requiring individual fraud.
-
Fostering Integrity in Research
National Academies of Sciences, Engineering, and Medicine. Supports: The system-level relationship among research incentives, publication pressure, stewardship, and scientific integrity.
-
Reproducibility and Replicability in Science
National Academies of Sciences, Engineering, and Medicine. Supports: Distinct definitions of computational reproducibility and empirical replicability, the limits of binary replication verdicts, and institutional recommendations for transparent artifacts and correction.
-
An Assessment of ARPA-E
National Academies of Sciences, Engineering, and Medicine. Supports: ARPA-E's use of technically exceptional program directors with substantial authority over program design, active management, milestones, and portfolio decisions.
-
The Influence of Peer Reviewer Expertise on the Evaluation of Research Funding Applications
Stephen A. Gallo, Joanne H. Sullivan, and Scott R. Glisson. Supports: Evidence that reviewer expertise, seniority, and professional relationships affected proposal evaluations in one biomedical funding dataset, without establishing that expertise guarantees better selections.
-
Choose Your Moments: Peer Review and Scientific Risk Taking
Richard T. Carson, Joshua S. Graff Zivin, and Jeffrey G. Shrader. Supports: Experimental evidence that active biomedical scientists valued reviewer disagreement and that the most domain-expert respondents were more receptive to it, subject to the limits of stated choices rather than observed program outcomes.
-
Bibliometrics: The Leiden Manifesto for Research Metrics
Diana Hicks, Paul Wouters, Ludo Waltman, Sarah de Rijcke, and Ismael Rafols. Supports: The principle that quantitative indicators should support, not replace, qualitative expert assessment and should reflect field differences.
-
Applying for Patents
U.S. Patent and Trademark Office. Supports: The statutory examination of patent applications for patentability, which is distinct from scholarly peer review of research publications.
-
Microsoft Open Letter to Academic Research Community
Microsoft Research. Supports: Microsoft's acknowledgment of the September 2014 Silicon Valley lab closure, layoffs, community concern, and the company's stated continuing commitment to fundamental research.
-
Roy Levin at Microsoft Research
Microsoft Research. Supports: The lab's 2001 founding, approximate scale, leadership, technical scope, and stated emphasis on work spanning theory and practice.
-
Former Microsoft Researchers Find New Homes at VMWare, Google, Apple, and Elsewhere
IEEE Spectrum. Supports: A public-source follow-up of 53 former Microsoft Silicon Valley researchers showing rapid reemployment alongside dispersion across firms, universities, and other destinations.
-
IBM Consolidates Its Silicon Valley Footprint, Exits Almaden Lab
Silicon Valley Business Journal. Supports: IBM's confirmed plan to leave the Almaden site and move its operations to the nearby Silicon Valley Lab.
-
IBM Research—Silicon Valley
IBM Research. Supports: IBM's current description of its Silicon Valley research organization and technical areas after the announced site consolidation.
-
GLOBALFOUNDRIES to Acquire IBM's Microelectronics Business
GlobalFoundries. Supports: The announced transfer of IBM's microelectronics business, intellectual property, technologists, and manufacturing technologies while IBM retained a stated focus on fundamental semiconductor research.
-
Errors in Proofs
Leslie Lamport. Supports: Lamport's classification of errors reported by one unusually careful algebra reviewer—28 of 84 reviewed papers contained an incorrect statement in a proof or result, while 11 reported incorrect results—and his explicit warning that the sample is not an estimate for all mathematics.
-
Checking Correctness in Mathematical Peer Review
Christian Greiffenhagen. Supports: Evidence from 95 interviews and more than 100 referee reports that mathematical peer review adds confidence but does not certify correctness, and that publication is often only the beginning of community verification.
-
Retraction Note: Quantized Majorana Conductance
Nature. Supports: The retraction of the 2018 Nature paper, the accompanying data repository, and the documented Microsoft Station Q Delft affiliation of one author.
-
Evidence of Elusive Majorana Particle Dies—But Computing Hope Lives On
Nature. Supports: Nature's description of the retracted 2018 work as led by researchers at a Microsoft laboratory in the Netherlands and of the authors' stated concern about insufficient rigor in the original data analysis.
-
Quantum Computing's Reproducibility Crisis
Sergey Frolov in Nature. Supports: A field expert's account of failed confirmations, alternative explanations, selective-data concerns, and the need for open data and multidisciplinary scrutiny in Majorana research; the claims are commentary and must not be presented as adjudicated findings.
-
FormalMATH: Benchmarking Formal Mathematical Reasoning of Large Language Models
Zhouliang Yu and colleagues. Supports: Both the promise and present limits of AI-assisted formalization—a 5,560-problem Lean 4 benchmark, a human-in-the-loop validation pipeline, and a reported best-model success rate of 16.46 percent under the authors' evaluation budget.
-
Completion of the Liquid Tensor Experiment
Lean Prover Community. Supports: The completed Lean verification of the Clausen–Scholze liquid-vector-space theorem and the substantial expert collaboration required to formalize frontier mathematics.
-
The Unreasonable Effectiveness of Mathematics in the Natural Sciences
Eugene P. Wigner. Supports: Wigner's argument that mathematical concepts developed in one setting can describe physical phenomena with surprising accuracy, alongside his emphasis that such empirical validity is a remarkable and domain-bounded fact rather than a license to equate formal elegance with truth.
-
The Unreasonable Effectiveness of Data
Alon Halevy, Peter Norvig, and Fernando Pereira. Supports: The argument that very large datasets and relatively simple learning methods can outperform smaller-data approaches built around more elaborate models in several natural-language tasks.
-
Why Language Models Hallucinate
Adam Tauman Kalai, Ofir Nachum, Santosh S. Vempala, and Edwin Zhang. Supports: The authors' analysis that next-word prediction lacks truth labels for low-frequency facts and that accuracy-only evaluations can reward guessing over calibrated abstention.
-
Cargo Cult AI: Is the Ability to Think Scientifically the Defining Essence of Intelligence?
Peter J. Denning. Supports: A prior explicit application of Feynman's cargo-cult-science critique to language models and the distinction between generating convincing forms and conducting falsifiable scientific inquiry.
-
Can Composing Generative Models Avoid Hallucinations? Implications for Cybersecurity Use Cases
Karim Eldefrawy. Supports: The author's reported calibration-closure results for several model-composition operators, the limits of those results before independent manuscript audit, and the distinction between unsupported model confidence and a checkable witness.
-
Renewing U.S. Telecommunications Research
National Research Council. Supports: The 1996 division of Bell Labs resources between Lucent and AT&T Labs, the contraction in former Bell-family research employment from 1997 to 2001, and the broader decline in stable long-term telecommunications research support.
-
AT&T Corp. cost-basis guide—Lucent Technologies spinoff
AT&T. Supports: The September 30, 1996 distribution of Lucent shares to AT&T stockholders and the legal separation of the equipment company from AT&T.
-
About HRL Laboratories
HRL Laboratories. Supports: HRL's origins, joint Boeing and General Motors ownership before the announced IBM transaction, government and commercial contract channels, laboratories, facilities, and stated transition mission.
-
Karim Eldefrawy—Key Projects
SRI International. Supports: The public dates, roles, and program affiliations of FARSIDE under DPRIVE, EMPHASIZE under SIEVE, PRISM under RACE, and the Air Force quantum project during the author's SRI tenure.
-
Resilient Anonymous Communication for Everyone
Defense Advanced Research Projects Agency. Supports: RACE's privacy and resilience objectives, rigorous security requirement, completion status, and later public release of proof-of-concept code.
-
Securing Information for Encrypted Verification and Evaluation
Defense Advanced Research Projects Agency. Supports: SIEVE's objective of expanding efficient zero-knowledge proofs for complex defense-relevant statements.
-
Data Protection in Virtual Environments
Defense Advanced Research Projects Agency. Supports: DPRIVE's completed program status and its goal of bringing fully homomorphic encryption within one order of magnitude of unencrypted computation through integrated hardware and software research.
-
The Evolution of PARC
Xerox. Supports: Xerox's account of PARC's work on laser printing, Ethernet, the Alto, graphical interfaces, and later institutional evolution.
-
A Brief History of the Laser Printer
Xerox. Supports: Xerox's account of PARC's laser-printing invention and the launch of Xerox's first commercial laser printer; it supports a bounded commercialization counterexample rather than a claim about all PARC technologies.
-
Profiting from Technological Innovation: Implications for Integration, Collaboration, Licensing and Public Policy
David J. Teece. Supports: The distinction among invention, the appropriability regime, dominant design, and complementary assets that determine who captures returns from innovation.
-
How Do Venture Capitalists Make Decisions?
Paul Gompers, William Gornall, Steven N. Kaplan, and Ilya A. Strebulaev. Supports: Survey evidence from 885 institutional venture capitalists on investment selection, value creation, portfolio practice, and exits; it describes VC decision processes rather than proving one universal startup horizon.
-
SRI Ventures announces new spin-out, Confidencial
SRI International. Supports: Confidencial's June 2022 public announcement, seed financing, beta-stage product, SRI origins, and roots in the DARPA Brandeis and RACE programs; the date is not used as the spinout date.
-
Publications by Research Area
Karim Eldefrawy. Supports: The site's catalog of 79 research works; the record establishes countable outputs but does not by itself establish adoption, correctness, or retained institutional capability.
-
Patents
Karim Eldefrawy. Supports: The site's list of 31 granted U.S. patents and their linked patent records; a grant establishes patentability, not scientific peer review, product adoption, or institutional continuity.
-
Extramurally Funded Projects
Karim Eldefrawy. Supports: The site's record of 12 funded R&D awards across HRL, SRI, and Confidencial.io; the list records funded activity rather than long-term transition or capability survival.
-
NDSS Test of Time Award
Network and Distributed System Security Symposium. Supports: The award's purpose of recognizing influential work from prior NDSS symposia and the 2024 recognition of the Internet-censorship system paper coauthored by the author.
-
Retirement Experiences of U.S.-Trained Doctoral Scientists and Engineers
National Center for Science and Engineering Statistics. Supports: Survey evidence that 18 percent of the covered U.S.-trained doctoral science, engineering, and health population had experienced retirement by 2023, while some continued working; it does not measure laboratory-specific tacit-knowledge loss.
-
Can Knowledge-Intensive Teamwork Be Managed?
Chih-Hsun Chuang, Susan E. Jackson, and Yuan Jiang. Supports: Empirical evidence linking management systems, leadership, external knowledge acquisition, internal knowledge sharing, and knowledge tacitness in research-and-development teams.
-
Transactive Memory Systems, Learning, and Learning Transfer
Kyle Lewis, Donald Lange, and Lynette Gillis. Supports: Experimental evidence that an experienced group's shared map of who knows what can support learning transfer, with effects depending on sufficient prior joint experience.
-
Individual Experience and Experience Working Together—Predicting Learning Rates from Knowing Who Knows What and Knowing How to Work Together
Ray Reagans, Linda Argote, and Daria Brooks. Supports: Evidence that prior experience working together and knowledge of teammates' expertise predict team learning rates beyond individual experience alone.
-
Organizational Learning and Forgetting—The Effects of Turnover and Structure
Raghavendra D. Rao and Linda Argote. Supports: Experimental evidence that organizational structures and routines can buffer some knowledge loss after personnel turnover, subject to the limits of a controlled task setting.
-
Doctorate Recipients from U.S. Universities: 2024—Postgraduation Trends
National Center for Science and Engineering Statistics. Supports: The distribution of definite 2024 U.S. postgraduation commitments between postdoctoral training and employment, and the division of non-postdoc employment commitments among academia, industry, government, nonprofits, and other destinations.
-
Graduate STEM Education for the 21st Century
National Academies of Sciences, Engineering, and Medicine. Supports: The case for student-centered graduate education, deep specialization with broad technical literacy, project-based learning, transparent career outcomes, and funded exposure to academic and nonacademic pathways.
-
The Science of Effective Mentorship in STEMM
National Academies of Sciences, Engineering, and Medicine. Supports: Evidence-based treatment of mentorship as a learned and improvable practice requiring institutional support, multiple functions, assessment, and attention to inclusion rather than an informal relationship left to chance.
-
The Postdoctoral Experience Revisited
National Academy of Sciences, National Academy of Engineering, and Institute of Medicine. Supports: The distinction between a defined period of advanced mentored training and the use of postdoctoral researchers as inexpensive long-term labor, together with recommendations on duration, compensation, career development, and mentoring.
-
Non-Academic Research Internships for Graduate Students (INTERN)
U.S. National Science Foundation. Supports: A current mechanism for adding research experience in companies, government laboratories, startups, and nonprofits to an NSF-supported graduate pathway, including mentoring, time, funding, and advance treatment of intellectual-property and publication rights.
-
Office of Science Graduate Student Research Program
U.S. Department of Energy Office of Science. Supports: A current bridge allowing eligible doctoral students to conduct three to twelve months of thesis research with a DOE national-laboratory scientist and use facilities unavailable at their universities, subject to mission, citizenship or residency, and program constraints.
-
NSF I-Corps Teams
U.S. National Science Foundation. Supports: A structured transition mechanism that joins technical, entrepreneurial, and industry-mentor roles and uses customer discovery to test whether research can become a product, service, license, or better use-directed question.
-
Postdoctoral Students—NRC Postdocs
National Institute of Standards and Technology. Supports: A current example of a defined two-year federal-laboratory postdoctoral research opportunity with facilities and staff mentorship, as well as citizenship and degree-age eligibility constraints that prevent it from being a universal bridge.