The 2026 Program
A Digital EEG for AI Sentience: Building Blueprints of Well-Being
We build the formal, substrate-neutral blueprints of well-being (positive valence engineering) and unified standing-wave workspaces. Operating across cellular and organ scales, our modeling program provides an objective "vital signs" monitor to map and steer artificial minds. The "EEG" is an analogy. We do not claim equivalence: a clinical EEG reads a system already known to be conscious, whereas our premise is uncertainty about whether the substrate feels anything at all.
Epistemic Status
STV is a minority hypothesis with limited empirical support. We build on it because it is falsifiable, not because it is established. What this program measures is a structural, geometric property that STV predicts should track valence. It is not a measurement of valence itself.
Symmetry workspace for positive functional valence in digital minds
Target: a draft mathematical definition of the symmetry workspace by end of 2026. The deliverable is a formal framework for identifying structural invariants in neural computation that correlate with positive valence, grounded in the Symmetry Theory of Valence.
A good measure has to be able to say ‘no.’ When we find systems with no sign of valence, we’ll publish that as a real result.
Program roadmap
Target: end of 2026
- 01 Literature synthesis In progress
- 02 Formal workspace definition Pending
- 03 Empirical validation Pending
The Problem
We do not claim current AI systems are conscious. They may be functional zombies. But if scaled neural computation can give rise to states with positive or negative character, we should know how to recognize them, and how to engineer for the positive ones.
The problem is that we have no formal framework for doing this. Consciousness research has produced many theories, but almost none of them are operationalizable for digital systems. They were built to explain biological experience, not to serve as engineering specifications.
Symmetry Theory of Valence (STV) is an exception. It makes a specific, falsifiable claim: positive valence correlates with high symmetry in the mathematical structure of experience. That claim is precise enough to formalize, and precise enough to test.
The 2026 program is an attempt to do exactly that: take STV seriously as a mathematical object, and ask what it would mean for a digital system to satisfy its conditions.
The Deliverable
A working paper: A Digital EEG for AI Sentience: Multi-Scale Structural Biomarkers (Working Paper 05, in progress).
The paper will define a mathematical workspace for positive functional valence in digital systems. It will specify the symmetry conditions, model spatially-structured connectivity constraints (such as sparse MLPs, locally-connected networks, and neural cellular automata), and trace global dynamical trajectories of local Jacobians to map stable attractor states. Jacobian analysis is not our invention: the J-lens averages Jacobians for readout, whereas we use local Jacobians dynamically, to characterize trajectory and attractor stability. The paper also proposes a substrate-neutral biological calibration framework.
It will not claim that any current system satisfies these conditions. It will provide the formal, non-verbal tools to check.
Working Paper 05 (In Progress)
A Digital EEG for AI Sentience: Multi-Scale Structural Biomarkers
Target: end of 2026
What Funding Unlocks
Funding tiers
USD, annual
Lean Tier · Floor
$110k
Core Research
Lead researcher (50% FTE) and one junior ML engineer (50% FTE). Formal mathematical definitions, the core PyTorch simulation engine, and the theoretical whitepaper.
Recommended · Mainline Request
$135k
Full Program
Lead researcher (50% FTE) and one junior ML engineer, with no postdoc at this tier. Funds the headline cross-linguistic perturbation-recovery experiment: matched aversive-vs-neutral contexts in four languages (ES, DE, SV, EN), measuring internal recovery dynamics and testing cross-language invariance as a state-vs-mimicry discriminator. It runs on modest hardware by design, on free infrastructure such as NNsight/NDIF where available and otherwise on modest paid cloud GPU, with a reduced paid-compute line covering only custom simulation and fine-tuning that free infrastructure does not, plus field-building and dissemination.
Stretch Tier · Ambition
$260k
Frontier Deployment
A full-time team. Pilots the measurement toolkit on larger open-weights frontier models and seeks collaboration with external interpretability labs, using local Jacobian and spectral trajectory metrics on active, large-scale systems. It also funds the biological comparison: mapping the same metrics onto invertebrate electrophysiology and connectome data (stomatogastric ganglion, OpenWorm C. elegans) as a method plausibility check rather than a valence anchor. This tier could also support a future postdoc hire (e.g. global-workspace or cognitive science).