§0 · Demo

Five questions. Five fields. Two kinds of answer.

On the left of each scene: what a vanilla LLM (ChatGPT, Claude, Gemini) typically returns. On the right: what LivingMeta returns from the live instance — with traceable citations, methodological detail, multi-perspective consensus, evidence-based gaps, and a Lab coach that scales with the user's role.

The vanilla-LLM responses are illustrative — they represent typical failure modes for this kind of question. Test them yourself. The LivingMeta responses are grounded in the live instances: real corpus sizes, real work IDs from the priority agenda, real authors and themes from the actual extractions. Click any instance link at the end of each scene to verify.

§1 · Menstrual Health at Work

What does the literature say workplace interventions for menstrual symptoms should look like?

Posed by an HR policy researcher building a case for a cycle-aware support scheme.

A vanilla LLM answer

Illustrative

There are several workplace interventions that have been studied for menstrual symptoms, including flexible working hours, menstrual leave policies (as seen in Spain, Japan, and Indonesia), on-site rest facilities, and manager training on menstrual health awareness.

A meta-analysis by Sims et al. (2023) found that flexible scheduling reduced absenteeism by approximately 18% across pooled studies.

Sims et al. 2023 meta-analysis on this topic does not appear in any indexed database — likely fabricated.

Other recommended interventions include access to menstrual products in the workplace, educational programs for staff, and the development of formal disclosure-friendly policies. The evidence base is generally considered to be emerging but promising.

Hedging language with no anchor to specific evidence. No paper IDs, no agenda priorities, no resource registry.

A LivingMeta-grounded answer

From menstrual-health-work

Over 3,400 papers in the broader field, with a focused cluster on workplace implementation. The agent synthesises across the cluster — it doesn't fake aggregation from outside the corpus, and it doesn't pretend papers exist that don't.

Real corpus count from the instance, real cluster scoping.

Foundational evidence anchored on specific papers: the landmark Dutch nationwide survey by Schoep et al. [W2954816350] — 32,748 women, FWCI 15.1, 268 citations — sets the productivity case the design needs.

Real work IDs and real FWCI scores; every citation clickable to OpenAlex.

Surfaces a multi-layered design model: flexibility + de-stigmatisation + infrastructure + clinical referral pathways as INTERDEPENDENT levers, not a checklist. Reflects what the corpus actually argues, not a generic HR template.

Synthesised argument, not bullet-list output.

Open the thread to read the full ~15K-character synthesis with 90+ clickable paper references.

Recorded from the live Lab agent on menstrual-health-work.

T4 · Evidence-based gaps + curated resources
Gap mapping and resource curation — both grounded in the corpus, not invented. The visitor leaves with a defensible "what we know vs what we still need to study" picture, with the instance's own priority agenda as the answer to "where should the field be working next?"

§2 · Innovation Management

What are the open research questions on AI integration in innovation ecosystems?

Posed by a PhD student scoping their dissertation in industrial-innovation research.

A vanilla LLM answer

Illustrative

AI integration in innovation ecosystems is an emerging field. Key research questions include how AI affects innovation processes, the role of platforms, ecosystem dynamics, and governance challenges. Recent work by Smith et al. (2024) has highlighted platform interoperability and vendor lock-in as critical factors.

Smith et al. 2024 cited with no DOI and no journal — likely fabricated.

Other authors have stressed the importance of absorptive capacity and dynamic capabilities (Teece et al., 2007) in determining whether organizations can integrate AI successfully. Top-management support also matters (Cooper & Zmud, 1990).

Real citations, but pulled from unrelated literatures — Cooper & Zmud is from IS implementation, not AI ecosystems. Conflation.

The open questions are generally considered to be around governance, ethics, and the long-term effects on innovation diversity. Researchers are still working out the right frameworks.

Hedging without anchors. No agenda items named, no paper counts, no priority signal.

A LivingMeta-grounded answer

From innovation-management

The agent's synthesis draws on 20+ papers in the AI-integration cluster, from a ~2,000-paper IM corpus. Five distinct sub-clusters of open questions emerge — each treated separately with its own evidence base.

Real cluster count, drawn from actual extractions.

Frames the field as having shifted from "whether AI gets adopted" to "how adoption reshapes innovation logic" — a substantive theoretical observation, grounded in the recent literature, not a generic AI-hype frame.

Argumentative framing, not generic AI-adoption talk.

Cites Ji & Tian (2026) [W7133900944] in Sustainability for the Dynamic Capability problem: AI capabilities alone don't produce innovation outcomes. Specific paper, specific journal, specific argument.

Verifiable citation with paper, journal, and argument context.

Open the thread to read the full ~13K-character synthesis across all five clusters, with 49 clickable paper references.

Recorded from the live Lab agent on innovation-management.

T2 · Methodological rigor + honest aggregation
Per-paper methodological rigor and honest aggregation. The platform surfaces what the corpus actually says — counts, clusters, priorities — and does not fake meta-analytic outputs the underlying extractions do not support. Reproducibility is by construction: every claim is downloadable JSON.

§3 · Sports Analytics

How robust is the "acute:chronic workload ratio predicts injury" finding?

Posed by a sport-science researcher writing a coach-facing review chapter on training-load monitoring.

A vanilla LLM answer

Illustrative

Hulin et al.'s 2015 work on the acute:chronic workload ratio is widely cited and has become a standard reference in training-load monitoring. The general consensus is that a ratio above 1.5 indicates elevated injury risk in athletes.

Specific threshold (1.5) doesn't match the source — the paper reports 2.11 as the high-risk threshold. Standard "general consensus" framing with no source check.

Subsequent meta-analyses have largely confirmed the predictive validity of the ACWR, though some authors have raised concerns about measurement error and the choice of moving-average windows. Overall, the framework is considered well-validated for training prescription.

No specific meta-analysis cited. "Well-validated" claim doesn't reflect the actual ongoing controversy in the methodological literature.

A LivingMeta-grounded answer

From sports-analytics

Drawing on 22+ papers, the agent makes a sharp distinction: the ACWR-injury association is real and replicable, but the predictive capacity for individual injury is weak. These are different claims, and the literature has often conflated them — for a coach-facing review, the distinction IS the story.

Honest disambiguation — exactly what vanilla LLM glosses over.

Anchored on Hulin et al. 2015 [W1859218480] (FWCI 38.2, 460 citations): 53 elite rugby league players, 8,177 GPS files, binary logistic regression. Real methodology, real sample size, full statistical specification.

Real paper, real measurement specs, all queryable in the corpus.

The follow-up question in the modal asks for the meta-analyst persona's replication specs. The agent zooms in: minimum N, multi-club design, measurement-window standardisation, pre-registration of analytical choices.

Multi-turn dialog with persona-specific drill-down — the 6-persona architecture in action.

Open the thread to read the full ACWR-evidence synthesis + the meta-analyst replication-design follow-up. ~14K + ~16K characters, 49 clickable paper references.

Recorded from the live Lab agent on sports-analytics.

T3 · Synthesis + 6-perspective consensus
Six independent AI perspectives, same paper, same fields, different cognitive emphasis — and a consensus mechanism that shows where they agreed and what each one added on top. Not a black-box summary: a triangulated, queryable extraction with the source quotes attached.

§4 · Lokaal Bestuur

Welke theorieën verklaren bestuurlijke ontkoppeling tussen formeel beleid en feitelijke praktijk in Nederlandse gemeenten?

Question in the instance's research language. Posed by a bestuurskundige onderzoeker working on Dutch municipal implementation gaps.

A vanilla LLM answer

Illustrative

Dutch public administration shows gaps between policy and practice, often called implementation problems. This is a known challenge in governance — organisations sometimes adopt policies for legitimacy without following through. Studies suggest better monitoring and clearer communication tend to help.

No citations, no named theoretical mechanism, no Dutch evidence. "Studies suggest" with nothing behind it.

Researchers have proposed various frameworks for this, including organisational learning, principal-agent theory, and policy implementation models (Pressman & Wildavsky, 1973). Recent work by Van Dijk et al. (2022) emphasises the role of street-level bureaucrats.

Pressman & Wildavsky is a real classic, but disconnected from the Dutch decoupling literature. Van Dijk et al. 2022 — unverifiable.

A LivingMeta-grounded answer

From lokaal-bestuur

De agent identificeert vijf dominante theoretische tradities — neo-institutionalisme, principal-agent, sensemaking, padafhankelijkheid, bestuurskundige fragmentatie — en behandelt elk vanuit zijn eigen mechanistische logica. Geen enkele theorie is afdoende op zichzelf.

Vijf theorieën expliciet, niet één generieke "implementation gap".

Anker: DiMaggio & Powell's institutionele isomorfie [W3123282572] verklaart ontkoppeling als legitimiteitsmanagement via coercitief, mimetisch en normatief isomorfisme. Concrete theorie, concrete paper, concrete mechanismen.

Klassieke theorie met klikbare brontekst — niet zomaar een buzzword.

De follow-up vraag in de modal: "welke Nederlandse gemeenten hebben hier evaluaties over gepubliceerd?" De agent combineert literatuur + WooZM raadsstukken + OpenBestuur — BYOD cross-corpus in actie, ondersteund door de live data-corpora (4,7M raadsdocumenten, 6,84M WooZM agenda-items, 614K CBS regels).

BYOD: literatuur + 3 publieke Nederlandse data-corpora in één antwoord, citatie-discipline op elk niveau.

Open de thread voor de volledige Nederlandse synthese + de cross-corpus gemeente-follow-up. ~12K + ~14K karakters, 46 klikbare paper-referenties.

Opgenomen van de live Lab agent op lokaal-bestuur.

T1 · Traceable citations + BYOD cross-corpus reasoning
Traceability isn't only about academic citations — it's about every layer of the answer being clickable. Lokaal Bestuur shows the BYOD pattern: your own domain corpus sits alongside the literature, with the same citation discipline. Theory from the literature, evidence from the records, financial reality from the statistics — all in one queryable instance.

§5 · Emergent Organizational Change

How do organisational change recipients construct and reconstruct their roles?

Posed by a change-management consultant who supervises four masters' theses on emergent change in public organisations.

A vanilla LLM answer

Illustrative

Change recipients can adopt various roles — supporters, resisters, ambivalent observers — depending on their interpretation of the change. Key factors include perceived fairness, communication quality, and psychological safety. Bernerth et al. (2011) discuss change cynicism as a related construct.

Generic typology with no theoretical anchoring in the recipient-role literature. Bernerth citation is on a related but different construct.

Strategies to support positive role construction include involving employees in decision-making, providing clear communication, and offering training. Leadership style also matters — transformational leaders tend to elicit more positive responses.

Management advice, not research scaffolding. No corpus-grounded operationalisations offered.

A LivingMeta-grounded answer

From emergent-organizational-change

Across 30+ directly applicable papers on sensemaking, identity work, and role construction, the agent rejects the binary "ready vs resistant" framing prevalent in practitioner change literature. Substantive critique, not a summary.

Critique grounded in the corpus — not just paraphrase.

Anchored on Van der Schaft et al. [W4401037182] 2024: a narrative study of 80 recipients in a top-down change initiative yields five composite narratives of role-taking under change. Real empirical anchor with sample + method visible.

Specific paper, specific N, specific method — not "studies show".

The same question, asked by Junior / Senior / Coach, yields three different responses. The Lab agent reads the user-role and adapts: Socratic scaffolding for the junior, peer-level density for the senior, course-design framing for the coach. Same evidence base, three modes of usefulness.

Role-toggle in the modal — switch between the three live agent responses.

Open the modal and toggle through the three role variants. Each is the agent's full response from a separate live thread on emergent-organizational-change.

Recorded from three separate live threads, one per role.

T5 · Lab coach with role toggle
The Lab coach scales with the user. A junior gets Socratic scaffolding pointing at three real corpus operationalisations; a senior gets peer-level density with the actual lineage; a supervisor gets a single-case scaffold and exemplar theses they can hand to students. Same evidence base, three different ways of being useful.

§6 · Your field

Don't see your field?

Five live instances is not a hundred. If your field is not above — criminology, medical biology, French literature, computational linguistics, anything — the same workflow runs identically. Tell us what your literature looks like.