Research interest: Pre-representational computation—the operations that exist before and enable representation (whether human language, embeddings, or any other representational layer).
Specifically: which compositions of linear algebra operations (projection, attention, regularization, weight sharing) preserve structure, and how do systems learn priors over which operations to apply when.
Verification-centered empirical research. For alignment researchers: verifiable behavior and scalable oversight. For quantitative researchers: robust coordination mechanisms.
Corpus
Research artifacts, deployed systems, and essays in one sequence, ordered by contribution rather than category.
14 entries, ordered by how much each contributes to the case.
Grounded Commitment Learning
ResearchExperimentMulti-agent coordination through verifiable behavioral contracts. Agents commit to observable behaviors rather than inferred mental states—enabling external verification without access to internal representations. Grounded in Hart-Moore incomplete contracts theory (Nobel Prize in Economics, 2016).
40.4% hold-up reduction (95% CI: [37.2%, 43.5%]) · r = -0.972 punishment paradox, p < 0.001
derive_real_headline_stats.py · Re-derived by CI · Repository
Persistent AI advisors, the people they model, and the limits of “more context”—why a user model should preserve the history of its claims rather than a polished conclusion.
Interactive mirror · Event-sourced state, applied · References
Chat with the Research
SystemEvaluationProductionGrounded RAG chatbot over this site, red-team hardened. Published eval-gate numbers including the bars still unmet, and the real defects the harness found and root-caused—rate limiter, citation injection, retrieval crowding.
Collaborative Nested Learning
ResearchExperimentExtension of Google Research’s nested optimization: 5 timescales with 9 bidirectional knowledge bridges. Addresses catastrophic interference where fast learning degrades slow-learned representations, via normalization constraints that preserve component distinctiveness during optimization.
+89% accuracy at high regularization, where baseline collapses
Bridge ablation, re-derived by CI · Reproduce in Colab · Repository
Aegis
SystemSystems-architecture layer beneath agent frameworks: durability, verification, and policy—not orchestration. Event-sourced state for resume/replay from any checkpoint, a tool gateway enforcing policy at invocation time, and GCL commitments as first-class objects with explicit failure modes.
303 tests passing
Single-node only. No performance benchmarks yet, and no production deployment.
What the 303 tests cover · Repository · Event-sourced state applied to models of people
HALA: Human-AI Layered Architecture
EssayResearchPattern language for human-AI role allocation: 28 patterns across 5 layers. Emerged from deployment experience where organizational dynamics—not technical capabilities—were the binding constraint.
Agentic Delivery: Specs, Supersession, and Eval Gates
EssayEvaluationHow this site’s content was written by an AI coding agent under a written spec with supersession rules, a fact inventory, and eval gates—including the defect the process found in its own verification record.
Document Understanding Pipeline
ProductionProduction extraction with multi-provider LLM routing and human-in-the-loop validation, shipped at HCA Healthcare. HIPAA-compliant architecture prioritizing audit trails over model optimization; corrections captured as reward-shaped labels for offline fine-tuning.
Strategy Tournament
SystemExperimentA backtesting and evaluation framework for comparing systematic strategies under realistic execution constraints—slippage, latency, partial fills—and robustness stress tests, so a strategy that only looks good against a frictionless backtest gets caught before capital does. Addresses evaluating adaptive systems when the evaluation itself can be gamed.
Intelligence Control Layer
SystemProductionThree-layer governance architecture for autonomous agents in high-stakes domains: a constitutional layer of frozen constraints that cannot be overridden, a human-supervised strategic layer, and an autonomous tactical layer. Implements HITL review queues, Platt-calibrated confidence scoring, and bandit-based exploration/exploitation.
Built under employer confidentiality: no public repository, quantified result, or stated limitation. Ranked on what can be shown, not on what it is worth.
Adversarial Intelligence Platform
ProductionReal-time voice-AI agentic system for high-stakes adversarial contexts. Multimodal pipeline combining streaming ASR, paralinguistic signal extraction, and RAG over domain corpora, with a multi-agent dialectical architecture, predictive outcome modeling, and nested temporal learning. Built as First Staff Engineer at the AI division of a multi-billion-dollar vertical-SaaS private equity firm.
Built under employer confidentiality: no public repository, quantified result, or stated limitation. Ranked on what can be shown, not on what it is worth.
Multi-Agent Orchestration Platform
ProductionProduction agentic workflow coordination.
Reproducibility
EvaluationThree flagship claims—the punishment paradox, Hart-Moore hold-up reduction, and the CNL bridge ablation—re-derived by CI on every push to the source repos. Not a promise that a number cannot change; a standing invitation to check whether it still holds.
Limitations
Applied research with empirical validation in specific contexts. Results demonstrate effectiveness within tested conditions; generalization bounds require further investigation.
Scaling boundaries (~100 agents for GCL, Split-MNIST scale for CNL) may reflect experimental design constraints rather than fundamental limitations.
Open Questions
Questions I find myself returning to—some tractable, some speculative, all shaping how I think about adaptive systems.
Structure preservation under composition
Which compositions of projection, attention, and regularization preserve structure vs. destroy it? Neural networks compose these operations, but we lack a theory of which compositions maintain useful invariants. This connects to interpretability: if we can't enumerate the computational primitives a system uses, we can't fully interpret its behavior.
Minimal generating set
What's the minimal set of operations that generates the others? Current architectures use attention, convolution, normalization, nonlinearity—but these may not be primitive. Identifying the minimal basis is a prerequisite for mechanistic understanding.
Meta-learned priors over operations
How do meta-learned priors over operations develop during training? The space of operations available for adaptation is itself learned. Understanding this development may reveal why certain capabilities emerge at scale and others don't.
Missing computational primitives
Are current architectures missing computational primitives present in biological neural systems—continuous-time dynamics, local learning rules, dense recurrence, routing as control, structured memory—and would adding them change the space of learnable behaviors?
Alignment for systems with learned operations
What does alignment mean for systems whose operations are themselves learned? If the computational primitives change during training, alignment targets a moving substrate.
This suggests alignment may need to constrain the space of learnable operations, not just the outputs. Current approaches assume fixed computational primitives—an assumption that may not hold for sufficiently capable systems.
Commitment as lossy compression
A commitment is a lossy compression of intent into verifiable behavior. What's the rate-distortion tradeoff? More specific commitments are easier to verify but lose flexibility; more abstract commitments preserve optionality but resist verification. There may be an optimal commitment "resolution" that depends on trust level and coordination complexity.
If any of these resonate, I'd welcome the conversation: jason@jasonstiltner.com