SMF Labs.

Where We Sit in the Ecosystem

The voice biomarker space has organized itself into three camps over the last eighteen months. We do not sit inside any one of them. We sit at the intersection of all three, which is both our advantage and the thing that makes our funding story unusual.

Bernard's memo and my own context on Modality

Bernard sent me a memo on May 7 outlining how he sees Modality.AI fitting into our work. His framing is that they are a methodological predecessor whose findings validate our approach, not a competitor we need to beat. I want to add a piece of context that is mine to add. I have known about Modality for some time. David Suendermann-Oeft has presented at BioFuture, which I attend with Dr. Walter Greenleaf, with whom I co-authored the VR in Healthcare publication. So when Bernard's memo arrived, the ecosystem context was already mine. Bernard, this is not me catching up to your analysis. It is me confirming it from a different angle. We are aligned on where Modality fits.

The most important practical piece of Bernard's memo is the UMCU connection. Modality's cross-lingual ALS work was conducted with Professor Leonard van den Berg and Dr. Ruben van Eijk at UMC Utrecht. Bernard knows both personally and is in active discussion with Ruben about their 85-pALS longitudinal cohort. That puts us in a different position than a platform vendor seeking a clinical partner. We are pursuing scientific co-investigation grounded in a long-standing personal connection.

On Canary specifically

Henry O'Connell, Canary's CEO, signed our LOI in principle on April 11, 2026. He explicitly accepted moving to a consortium agreement covering IP, data governance, commercialization rights, and revenue sharing. The conversation paused for personal reasons in April. It is dormant, not dead. Henry has been waiting. The decision is whether to re-engage this week or to hold the conversation until after Mexico when we have more leverage from the Baja announcement. My current view is that we re-engage this week with a phased structure that earns trust through delivery rather than promising it through paper. I would like Bernard's read before I send the email.

Where we actually sit

We are the only group operating credibly across all three camps. Bernard's phonological subspace methodology is genuinely cross-cutting. For camp one, Paper 4 (drafted) and Paper 1 (under review at PLOS Digital Health) establish the scientific basis for severity assessment as a clinical trial endpoint. For camp two, our cross-lingual capability is what Canary lacks. They have features tuned for English. We have validated severity assessment across sixteen languages without retraining. For camp three, the work with Kyutai on dysarthric speech-to-speech, plus the Mexico ALS pipeline through TecSalud, plus the ElevenLabs voice clone delivery, gives us the only end-to-end story from measurement through restoration.

The honest read: Modality is two years ahead of us on regulatory paper trail. We are likely two years ahead of them on underlying architecture and language coverage. Canary has a commercial machine we do not have. We have linguistic reach and an ethical narrative they cannot manufacture. The right strategic posture is to hold our position at the intersection and let the structural advantages compound.

What I Have Learned From Recent Research

A few things genuinely changed my thinking over the last several days. I want them in the file.

On Modality and the post-Euphonia landscape

Modality has been very productive in 2026. They published in Nature Scientific Reports in April. They presented at Target ALS 2026 and AD/PD 2026. Their 137 peer-reviewed publications and 78,000 participants are real and matter. What our Track 4 work does, that they cannot do, is provide training-free severity assessment that works across sixteen languages without ever needing pathological data from those languages.

Google Project Euphonia announces expanded multilingual results in June 2026. Whatever the announcement looks like, it will reframe the post-Euphonia ecosystem. Our positioning has to account for this. The framing should assert our differentiation in the post-Euphonia landscape: training-free, cross-lingual, patient-led, ethically governed. None of those four properties are likely to be Google's. That is the wedge.

On Canary, structurally

Canary is in the population screening camp, technically a different market from Modality and from the clinical trial endpoint focus. Their feature library would benefit from access to our cross-lingual data. Their commercial infrastructure (Epic, Teams, payer relationships) is exactly what we lack.

My current view, which I want Bernard to validate before I send the re-engagement email: we re-engage Canary this week with a phased structure. Phase one is a non-binding consortium MOU naming three workstreams (TecSalud Mexico pilot, NIH R01 collaboration, Eurostars September application). No transfer of corpus access, methodology IP, or commercial rights at this stage. Phase two, contingent on Mexico pilot delivery, is a sponsored research agreement at $75K to $150K with defined corpus subset access and joint publication structure. Phase three, only after phase two delivers, opens the broader commercial conversation.

On regulatory pathways

The realistic timeline for either FDA or CE under MDR is twenty-four to thirty-six months. Modality has spent the last two years paying that regulatory tax. We have not. That gap is real, and we should not pretend otherwise. What we should claim is the architectural advantage: training-free methods are inherently more stable and less prone to overfitting, which the FDA's new AI/ML Lifecycle Management guidelines favor.

The EU AI Act enforcement begins August 2, 2026. That is also our advantage. If we pass EU bias and representativeness standards in 2026 with a sixteen-language corpus, we have a story no US competitor can tell. The Brussels Effect is real: EU AI Act compliance becomes a global signal.