On 1 July 2026, the United States Department of Energy published two documents on the official Genesis Mission page. One is the report from its summit with universities and science philanthropy, held on 18 February in Crystal City. The other analyses responses to its consultation on talent development for AI in science and engineering.

This is not a minor administrative update. The Genesis Mission has entered a phase of cultural engineering. It is not simply connecting supercomputing, artificial intelligence, quantum systems and scientific facilities. It is seeking to reorganise the rules that distribute credit, access, prestige and career opportunities within US science. Above all, it does so by deciding which forms of scientific work its platform can read.

The report’s opening letter is signed by Darío Gil as Under Secretary for Science and Genesis Mission Director. It sets out the aim of doubling US R&D productivity over a decade and compressing discovery timelines from years to months. Yet the documents show that this acceleration does not depend on more computing alone. It requires changes to academic careers, intellectual-property agreements, data practices, training models and the languages of evaluation. Genesis Mission University Summit Report

Productivity is not a neutral number

The summit report proposes moving beyond a narrow idea of productivity based on publications and citations towards a notion of impact. Suggested indicators include validated models, technology transfer, progress on national challenges and workforce development.

That may seem like a reasonable correction to a system that has disproportionately rewarded the individual article. But every reform of evaluation also decides which forms of work will be recognised as valuable.

Software, curated data, shared infrastructure, interdisciplinary co-ordination and mentoring begin to enter the field of visibility. That matters. But recognition is not innocent. When a state platform defines what counts as contribution, it also sets the grammar through which universities, teams and careers must express themselves in order to gain resources, reputation and continuity.

Every reform of evaluation decides what counts as valuable. What matters is who designs that scale and what it leaves out.

Slow science, curiosity without immediate application, dissent that complicates a mission, or the time needed to build trust can all come to appear as inefficiencies when accelerated impact becomes the dominant category.

Collective credit cannot be solved by an algorithm

The report’s most revealing tension lies between team science and individual credit. Participants acknowledge that promotion and tenure systems still reward individual achievement, while Genesis requires transdisciplinary, distributed projects sustained by many kinds of work.

The proposed answer is significant. It is to develop new ways of rewarding team science, potentially using AI tools to help identify individual contributions within collective projects.

Here a contradiction emerges. Genesis acknowledges that contemporary science depends on networks of interdependence, yet it tries to render that network legible again as a sum of separable inputs. AI appears as a possible aid to attribution. But this is not only a technical problem. It is a question of moral economy.

Sharing data, maintaining an infrastructure, documenting failures, translating needs between disciplines or supporting a team are not tasks easily reduced to an individual unit of merit. They are relationships. The report records a “human instinct to protect hard-earned data”, alongside the fear of being overtaken by other teams.

Proposals for master agreements, trust networks, incentives for data sharing and even an “FDIC for trust” are attempts to manufacture institutionally the trust that technical standards alone cannot impose. Data does not derive its value only from availability. It also depends on who can use it, under what conditions, and with what subsequent recognition. Summit Report on data, trust and credit

Infrastructure also produces subjects

The second document analyses more than 270 responses, totalling over 1,100 pages, to a consultation issued in January and closed in March. Its horizon is to train 100,000 American scientists and engineers over the coming decade to work at the intersection of AI and science or engineering. Genesis Mission Workforce Development RFI Analysis

Its interest is not only numerical. It outlines the kind of scientific person the mission needs.

The analysis calls for people with deep training in two fields, for example a scientific discipline and data science. It describes that combination as π-shaped. It also proposes a third competency in human-AI collaboration, including knowing when an AI output is reliable and how to retain independent judgement.

It also identifies a middle-skill gap. Responses propose roles such as AI technicians, data shepherds and applied-AI professionals who maintain infrastructures, curate AI-ready data and translate between AI specialists and domain communities.

The figure of the data shepherd deserves attention. Productivity cannot be sustained without data that are curated, documented, verified and situated. Public imaginaries of AI tend to glorify discovery, but Genesis admits that its infrastructure depends on persistent work of care, maintenance, traceability and repair. It acknowledges, if only indirectly, that scientific AI does not rest solely on spectacular discoveries. It also depends on maintenance, care and mediation work that normally remains outside technological epic narratives.

From programme to movement

The scale of the ambition also transforms education. The analysis argues that reaching 100,000 people will require moving from a programme to a movement, through train-the-trainer models, local chapters, technical institutions, community colleges, modular certificates and shared infrastructure for delivering, recording and recognising competences.

One formulation captures the project precisely. The problem is not a lack of courses, but the absence of national “digital glue” capable of credentialling and tracking training consistently across laboratories, universities and industry.

This is not only about teaching AI. It is about producing a professional population legible to a national platform. Each micro-credential, pathway and practical experience turns a varied educational biography into an interoperable sequence of competences, suitable for recruitment, mobility and state planning.

This co-ordination may modernise the scientific system. The crucial issue is what happens when scientific education is organised primarily around what a national infrastructure can classify, accredit and mobilise.

A platform also draws borders

The Genesis Mission presents itself as scientific infrastructure, but its categories are explicitly national. It is not organised merely to produce knowledge, but to ensure that particular data, capabilities, people and results are integrated into a national architecture of competitiveness. The executive order that created it links accelerated research to technological leadership, national security, energy dominance and returns on public R&D investment. Executive Order 14363

That does not, by itself, prove a closed policy of exclusion. It does show that the platform is being imagined as an infrastructure of sovereignty.

This is why it is insufficient to describe Genesis as an innovation policy. It is an operation of institutional reordering. It turns the laboratory into a node in a wider network; it turns the university into a provider and translator of talent; it turns philanthropy into a source of funding willing to assume risk; and it turns the scientific person into someone expected to combine judgement, disciplinary expertise, AI work and availability to move between state, industry and academia.

Anthropology does not enter here as an external warning about the risks of technology. It enters to study how a new scientific culture is built while the language appears to concern machines alone.

Genesis does not accelerate science by itself. It seeks to manufacture the cultural conditions under which a particular kind of science — measurable, interoperable, mission-oriented and legible to the platform — can count as the science of the future. The first step is to ensure that the platform can read it; the next is to make it governable.

Sources