Methodology
How every number is computed and which literature justifies each choice. The signalling-theory model is the same YAML the AI agents classify against — single source of truth.
Project goal
Specify as little as possible. Get out as much as possible.
(German original: “Spezifiziere möglichst wenig, krieg möglichst viel heraus.”)
The Ehrenthal four-signal scheme is given as the classification target; the actor list is given as the input. Everything else — which sources to consult, which extraction strategies to apply, which signals to drop as noise — is left to the two multi-agent systems. The thesis evaluates the gap between the two architectures along that wide degree of freedom. The signals an Atlas.ti researcher would manually code are the ones the systems should ideally produce automatically; where they disagree is where the evaluation focuses.
The core proposition
In markets with noncommensurable performance (no shared price / share / benchmark), actors and observers rely on signals. A signal's information value rises with its cost (hard-to-fake) and its observability (publicly verifiable).
Costly, hard-to-fake signals (patents, funding, peer-reviewed research) separate genuine capability from cheap talk (Rieger et al. 2025; Suchman 1995). Low-cost signals (positioning statements) are informative only when corroborated by costly ones.
The 19 signal dimensions (three axes)
Each signal is one dimension, on three literature-grounded axes: channel (capability vs legitimacy), cost (hard-to-fake-ness), and observability (public verifiability).
| Dimension | Channel | Cost | Weight | Observ. | Grounding |
|---|---|---|---|---|---|
| Leadership / board expertise Announcements of leadership and board expertise — named senior hires, scientific advisors, board appointments. | Legitimacy | Medium | 0.9 | high | Ehrenthal et al. 2026 (14% of legitimacy signals in their dataset); Hilkamo & Granqvist 2022 on named-personality sense-making in quantum. |
| Patents Patent filings (applications + grants), design rights, trademark filings — Swissreg, WIPO, EPO publications. | Capability | High-cost | 1.4 | high | Ehrenthal et al. 2026 (12% of legitimacy signals — "scientific credibility through patents"); Rieger, Dreller & Engelen 2025 (costly trademark/patent filings predict VC funding). |
| Publications Peer-reviewed papers, pre-prints, datasets, benchmarks released by the actor or with the actor as a major affiliation. | Capability | High-cost | 1 | high | Ehrenthal et al. 2026 (7% — "publications" under legitimacy); Knight & Cavusgil 2004 on research output as costly capability evidence. |
| Awards Awards, prizes, formal recognitions (industry awards, government commendations, ranking inclusions). | Legitimacy | Medium | 0.8 | high | Ehrenthal et al. 2026 (19% — "awards"; the largest single legitimacy sub-marker in their dataset). |
| Testimonials Customer / partner testimonials, third-party endorsements, named reference quotes. | Legitimacy | Low-cost | 0.5 | high | Ehrenthal et al. 2026 (18% — "testimonials"). Low-cost in our scheme because actors mediate which testimonials surface in their channels. |
| Educational outreach Educational programmes, workshops, student outreach, MOOCs, hackathons, academic-curriculum contributions. | Legitimacy | Medium | 0.6 | high | Ehrenthal et al. 2026 (16% — "the quantum computing industry earns legitimacy by sharing knowledge through educational outreach"). |
| Funding event Funding rounds (seed → growth), SNF / Innosuisse / Horizon Europe grants, public RFPs awarded, government contracts. | Legitimacy | High-cost | 1.5 | high | Extension to Ehrenthal et al. 2026 — funding sits in their "Communication Categories" (Investor / Funding) rather than the four-signal scheme directly, but it is unambiguously a legitimacy signal in the wider Suchman 1995 (pragmatic legitimacy) and Rieger et al. 2025 frame, where capital commitment by a third party is the canonical costly signal. |
| Regulatory recognition National strategy publications, export-control inclusion, standards work, participation in policy consultations, certification milestones. | Legitimacy | Medium | 0.7 | high | Extension to Ehrenthal et al. 2026 — not in their corporate-press dataset (they study vendor communications, not state recognition), but is a Suchman 1995 cognitive-legitimacy signal critical to the Swiss ecosystem (SNF / Innosuisse / federal quantum strategy). |
| Collaborations for applications National and international collaborations for specific applications (chemistry, finance, optimisation, materials). | Legitimacy | Medium | 1.1 | high | Ehrenthal et al. 2026 (85% — by far the dominant customer co-creation sub-marker in their 2025 dataset). |
| Pilots & POCs Simulations, pilots, proofs of concept conducted with named customers. | Capability | Medium | 1 | high | Ehrenthal et al. 2026 (14% — "simulations, pilots and proofs of concept, framed to demonstrate global relevance and early adopter engagement rather than market penetration"). |
| Customer training Customer support programmes, training, enablement, dedicated developer-relations content. | Legitimacy | Low-cost | 0.7 | high | Ehrenthal et al. 2026 (3% — "low, but is expected to increase as customer collaborations grow"). |
| Cloud-platform listings Availability via cloud-quantum platforms — AWS Braket, Azure Quantum, IBM Quantum, Google Quantum AI Cloud, OQC Cloud. | Legitimacy | Medium | 0.9 | high | Ehrenthal et al. 2026 (43% — "cloud-platform listings — the primary channel through which vendors depict global reach"). |
| HPC collaborations HPC-centre integrations, supercomputer collaborations, quantum-HPC hybrid demonstrations. | Capability | High-cost | 1 | medium | Ehrenthal et al. 2026 (32% — "HPC collaborations" as a community- ecosystem sub-marker); Adner 2017 ecosystem partner topology. |
| Industry partnerships Industry partnerships, MoUs with companies, distribution agreements, ecosystem-programme memberships. | Legitimacy | Medium | 0.9 | high | Ehrenthal et al. 2026 (14% — "industry partnerships"); Song et al. 2025 coattail legitimacy. |
| Academic partnerships Academic partnerships, university joint labs, named visiting professorships, multi-institution research consortia. | Legitimacy | Medium | 0.9 | high | Ehrenthal et al. 2026 (10% — "academic partnerships"). |
| Roadmaps Public product / technology roadmaps with named milestones and dates. | Legitimacy | Low-cost | 0.5 | high | Ehrenthal et al. 2026 (76% — "roadmaps … anchor long-horizon claims"); this is the canonical "cheap talk" signal Ehrenthal et al. study: highly observable, low cost, informative only when corroborated. |
| Milestones Specific milestone announcements — "X-qubit machine by year Y", named delivery dates, beta-availability windows. | Capability | Medium | 0.7 | high | Ehrenthal et al. 2026 (20% — "milestone sequences"). Medium-cost because missed milestones are visibly tracked by analysts. |
| Technological advances Technical-capability claims — qubit counts, gate fidelity, coherence times, error-correction demonstrations, new architectures, toolchain releases. | Capability | Medium | 1 | medium | Ehrenthal et al. 2026 (11% — "technological advances"); Tomesh et al. 2022 (no unified benchmark, so capability *claims* are medium-cost until corroborated by research or IP). |
| Long-horizon claims Long-horizon visions — fault-tolerant projections beyond ~5 years, industry-transformation narratives, broad future-state claims. | Legitimacy | Low-cost | 0.3 | high | Ehrenthal et al. 2026 (4% — "long-horizon claims"). Lowest-cost future-trajectory marker; informative only as part of a coherent pattern across cheaper and costlier signals. |
The scores
| Score | Formula | Answers |
|---|---|---|
| Signal Activity Score | Σ (weight × confidence) | How much an observer should update their view. |
| Cost-Weighted Signal Score | Σ (weight × confidence × cost_mult) | Signal activity after discounting low-cost signals. |
| Low-Cost Signal Share | low_cost / total | Substituting positioning for evidence? |
| Capability–Legitimacy Ratio | (cap+1)/(cap+leg+2) | Capability- vs legitimacy-led. |
| Signal Trend | signals_7d − prev_7d | Accelerating or cooling. |
| Signal Breadth | distinct dimensions | Broad vs narrow signalling. |
Cost multipliers: high 1 · medium 0.7 · low 0.4.
Limitations
This is a Bachelor’s-thesis prototype, not a production intelligence tool, and the numbers should be read accordingly. Coverage is bounded by what is publicly visible — arXiv, actor websites, and Google News, plus optional patent records. Activity behind paywalls, in private channels, or not indexed by these sources stays invisible, which systematically undercounts smaller or stealth actors. The cohort is a fixed list of roughly 40 Swiss quantum-computing organisations observed over a single evaluation window, so the results describe the Swiss ecosystem in 2026 and do not generalise to other deep-tech fields, other countries, or other time periods.
Every signal is classified automatically by large language models against the signalling-theory scheme above — not hand-coded by a human — so the scores are best-effort estimates rather than ground truth, and the line between a costly signal and a low-cost one is ultimately a model judgement. Both AI systems run on free-tier models and free web-search backends that are rate-limited and non-deterministic, so a given day’s run can miss signals or vary from the next. The two systems also differ in more than architecture (their prompts differ too), which makes the System A vs System B comparison illustrative rather than a controlled benchmark. Finally, one person designed the systems, the schema, and this evaluation — an independent evaluator was outside the thesis scope — so the dashboard reports observable public signalling and makes no claim about any actor’s actual research capability, commercial success, or scientific quality.
References
- Ehrenthal, Gonzalez-Padron & Gruen (2026) — four-signal scheme: legitimacy / customer co-creation / community-ecosystem / future-trajectory
- Suchman (1995) — legitimacy as the receiver-side evaluation of signals
- Connelly et al. (2011) — signalling theory review (receiver condition)
- Spence (1973) — signal cost as the credibility mechanism
- Rieger, Dreller & Engelen (2025) — costly trademark signals predict VC funding
- Knight & Cavusgil (2004) — capability-based competitive advantage
- Hilkamo & Granqvist (2022) — de novo quantum market sense-making
- Tomesh et al. (2022) — no unified quantum benchmark, hence signals as proxies
- Adner (2017) — ecosystem as structure
- Mohr & Sarin (2009) — high-technology marketing strategy
- Kolbe & Burnett (1991) — content analysis of observable evidence
- Blomqvist et al. (2008) — collaborative networks as legitimacy signal
- Song, Zhao & Wei (2025) — coattail legitimacy
Schema version 0.4.2 · revised 2026-06-02. Full code, prompts and disposition on GitHub.