Encounter notices structural problems in immuno-oncology trial designs that the sponsors often cannot see from inside their own assumptions — and works with them to address those problems before the trials read out.
We read ongoing immuno-oncology trials through a structural lens and look for patterns in the published phase 1/2 data that imply the phase 3 population has not been stratified on the axis the mechanism actually operates on. When we find one, the sponsor gets a dated private letter describing the observation, its implication for the pivotal, and a verification task they can run on their own data to confirm or rule out the question.
Alongside the private path, we maintain a living scorecard on Zenodo of structural predictions for trials we have read. Every prediction is timestamped before readout, falsifiable, and scored honestly when the trial reports. Predictions are never removed. Misses stay on the record.
The bug bounty analogy is the closest institutional analog. Independent researchers notice vulnerabilities that the system owners cannot see from inside their own assumptions; they disclose privately first; the dated disclosure itself is the artifact, regardless of what happens next. Encounter applies the same model to clinical trial design. Private disclosure is the default path. The public scorecard exists because the value of a structural read compounds only if the prediction is dated and public before the trial reads out.
We never name sponsors publicly in terms of how they responded to a private letter. Engagement is a private matter between Encounter and the sponsor. The scorecard publishes the call, not the conversation.
Every cancer drug does its work at a specific step in the immune response. Chemotherapy destroys tumour substrate. Checkpoint inhibitors release the brake on T cells. Bispecifics drag effector cells to the target. Different steps, different drugs.
And every patient population is stuck at a specific step. Some tumours are invisible to the immune system and the response never starts. Some are visible but the T cells can't get in. Some get infiltrated but the effector response is exhausted before it finishes the job. Different populations, different bottlenecks.
Trials fail, surprisingly often, because the combination addresses a step the population isn't held at — or adds a drug at a step the existing regimen already covers. The pattern is almost always visible in the earlier-phase data before the phase 3 begins. Almost nobody looks for it, because doing so requires an instrument the field does not have.
We have built that instrument. The Encounter Principle is the product of several years of structural theory, formal derivation, and retrospective validation against dozens of published immuno-oncology pivotals. It is not a heuristic. It is not a scoring rubric. It is a formally derived decomposition of the cancer-immunity cycle at a resolution the clinical-immunology literature does not operate at.
Every Encounter read is an application of the Encounter Principle to a specific trial. The directional call — whether the combination addresses where the population is actually held, or stacks on what it doesn't — follows from the decomposition. There is no judgement call in the final step. The judgement is in the instrument, and the instrument took years to build.
The Encounter Principle is a closed system. It is not taught in any clinical-immunology curriculum, not described in any standard textbook, and not derivable from the existing immunotherapy literature by reading the literature more carefully. It is its own thing.
Two paths exist. Sponsors who have received a private letter can engage directly on the question it raises; that conversation starts with a reply. Sponsors who want a structural read on a trial they are designing can request one through the service page. Either way, the first conversation is about whether the framework can answer the question at hand. If it cannot, we say so.
Encounter is led by Raimo van der Klein, an independent researcher based in the Netherlands and the author of the Generative Geometry framework.
