Lea as Identity Infrastructure: Why the Researcher App Should Become the Researcher Layer

1. Overview Researcher-focused social platforms face a structural paradox. The more specialized they become, the less they sustain the serendipitous exchange they set out to enable. ResearchGate has over 25 million registered users, yet fewer than 10% actively discuss research there [1]. Most treat it as an online CV [2]. Academic Twitter worked not because it was academic, but because it was Twitter. Lea, the researcher-focused Bluesky client presented by Maria Antoniac at ATmosphere 2026 [3], is the most promising attempt to rebuild what Academic Twitter offered. Its verification system, ORCID integration, and paper discussion pages address real gaps. This essay proposes that Lea's greatest potential lies not as a standalone app but as the identity layer for the emerging AT Protocol research ecosystem.
2. Definitions Specialization paradox: The structural contradiction in which a social platform optimized for a specific profession eliminates the ambient diversity that made general-purpose platforms productive for that profession. Logging into a researcher-only space implicitly demands "researcher mode," turning the environment into an obligation rather than a habitat. Identity layer: An infrastructure component on AT Protocol that structures and serves credentials and output records tied to a DID. It does not provide timeline, reply, or feed functions. It operates through labelers and lexicons, making its data accessible to any client on the protocol. Output-announcement bias: The observed pattern in which researchers' social media activity concentrates on two poles: announcing finished work (papers, books, datasets) and expressing isolation. The middle register, sharing thinking-in-progress, remains structurally absent. Intellectual property risk and competence face defense are the primary inhibitors.
3. Propositions P1: Productive intellectual exchange among researchers requires a general-purpose environment, not a researcher-specific one. Academic Twitter sustained knowledge exchange because it was embedded in a noisy, mixed-use stream. A researcher encountered papers between cat photos and political commentary. The cognitive cost of entering "research mode" was zero. Researcher-only platforms invert this condition. ResearchGate is instructive: it offers Q&A forums, discussion features, and follower networks. A Nature survey found that the most common activity on ResearchGate was simply maintaining a profile in case someone wanted to get in touch. The platform functions as a discoverability tool, not a community of social interaction [2]. A PLOS One study confirmed the majority of users had never asked or answered a question there, and most did not intend to increase their activity [4]. P2: Lea's most durable asset is its verification and credential infrastructure, not its client interface. Antoniac's team has manually verified 472 researchers, built ORCID authentication, connected to OpenAlex for co-author networks and publication records, and designed a composable verification model with progressive layers [3]. These components solve a real problem: on AT Protocol, there is currently no standardized way to confirm that a person is a working researcher. As a labeler, this verification data could be consumed by any AT Protocol client. As a set of lexicons, the paper pages and researcher profiles could be indexed and displayed by other apps. Confining these assets inside a single client limits their reach to Lea's user base. Opening them as protocol-level infrastructure extends their reach to the entire network. P3: The identity layer gap extends beyond researchers. Engineers, designers, translators, illustrators, planners: no one on AT Protocol has a structured way to present credentials and output history tied to their DID. Researchers have the advantage of existing identifiers (ORCID, OpenAlex, institutional emails). A researcher-first implementation could establish the pattern, with lexicon definitions for other output types following. Lea, having already built the verification pipeline and credential integrations, is positioned closer to this role than any other project in the AT Protocol ecosystem.
4. Corollaries C1 (from P1): If Lea operates primarily as a full client, its usage pattern will likely converge toward announcement and profile maintenance, replicating ResearchGate's trajectory rather than Academic Twitter's exchange culture. C2 (from P2): Releasing Lea's verification labeler as public AT Protocol infrastructure would allow Paper Skygest, Chive, Semble, Leaflet, and other research-adjacent tools to reference Lea's credentials. Lea becomes not an app competing for screen time but the trust anchor of a distributed research ecosystem. C3 (from P2, P3): The sustainability problem Antoniac identified [3] is alleviated by infrastructure specialization. A labeler and a set of lexicons require orders of magnitude less maintenance than a feature-complete client. C4 (from P1, P3): An identity layer on AT Protocol, tied to DIDs, produces a portable credential record that survives platform failure. This addresses the structural weakness of ResearchGate (proprietary, centralized) and researchmap (national, siloed) simultaneously.
5. Open Questions What conditions must a labeler meet for other clients to trust its verification decisions? AT Protocol's trust model for labelers remains an open design problem. Lea cannot become an identity layer without a credible answer. How should output types be defined in lexicons for professions that lack standardized identifiers? Researchers have ORCID. Designers, translators, and engineers do not. The lexicon design for a cross-professional identity layer is uncharted. Antoniac has invested significant effort in safety features: Block Party-style mass blocking, follower surge alerts, harassment triage [3]. These are client-layer functions. If Lea pivots toward infrastructure, where do these features live? The answer may determine whether the pivot preserves or abandons the community-protection mission that drives the project.

Discarded Hypotheses

• Researchers will share thinking-in-progress if given a safe enough platform. Intellectual property risk and competence face defense are structural barriers independent of platform design.
• Lea can sustain itself as a full client through grant funding. Grant cycles are mismatched with product maintenance timelines; no evidence of secured funding was presented.

References [1] Van Noorden, R. (2014). "Online collaboration: Scientists and the social network." Nature, 512, 126–129. https://www.nature.com/news/online-collaboration-scientists-and-the-social-network-1.15711 [2] "ResearchGate." Wikipedia. https://en.wikipedia.org/wiki/ResearchGate — Nature's 2014 survey reported that fewer than 10% of scientists would actively discuss research on ResearchGate, with 40% preferring Twitter. [3] Antoniac, M. (2026). "Lea: A Social App for Researchers." Presented at ATmosphere Conference 2026, Vancouver. Talk listing: https://atmosphereconf.org/talks | VOD: https://www.youtube.com/watch?v=NC1lPAv-EMU| Bluesky post: https://bsky.app/profile/mariaa.bsky.social/post/3mj3qk7k64s2z [4] Kramer, B. & Bosman, J. (2018). "Survey on opinions and usage patterns for the ResearchGate platform." PLOS ONE. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0204945 [5] Antoniac, M. — Academic profile: https://maria-antoniak.github.io/