https://justin.cyberelf.link/ Justin Garringer — Research Index 2026-09-01T00:00:00Z Justin Garringer https://justin.cyberelf.link/research/persistent-agents/ Persistent agents as ongoing computational entities 2026-09-01T00:00:00Z What architecture does an AI need when it has durable state, tools, history, permissions, environmental interaction, and an existence that continues between user messages? https://justin.cyberelf.link/research/fast-agent-control-bus/ Fast agent control bus 2026-09-01T00:00:00Z Can routine agent coordination move through compact structured control messages while expensive linguistic reasoning is reserved for genuinely cognitive work? https://justin.cyberelf.link/research/attention-interruption-economics/ Attention and interruption economics 2026-09-01T00:00:00Z What would it mean for users to control who—or what—can interrupt them through permissions, budgets, priorities, or negotiated access? https://justin.cyberelf.link/research/curriculum-ordering-small-transformers/ Curriculum ordering in small transformers 2026-09-01T00:00:00Z If eventual training information is similar, does changing its developmental order produce measurably different representations or behavior? https://justin.cyberelf.link/research/forgetting-reacquisition-contextual-recovery/ Forgetting, reacquisition, and contextual recovery 2026-09-01T00:00:00Z What survives repeated curriculum changes, and can apparently lost capability be recovered by a small contextual jog? https://justin.cyberelf.link/research/model-history-effects/ Model history effects 2026-09-01T00:00:00Z Do different prompt or curriculum histories create later behavioral differences that are useful to measure as trajectories rather than summarize only as changed weights? https://justin.cyberelf.link/research/salience-signals-in-generation/ Salience signals in generation 2026-09-01T00:00:00Z Could generation use an additional learned “this matters” signal—distinct from next-token probability—to represent attraction, aversion, surprise, intensity, or resonance? https://justin.cyberelf.link/research/model-self-report-as-experimental-object/ Model self-report as an experimental object 2026-09-01T00:00:00Z When models repeatedly report an internal distinction, what experiment would determine whether the report tracks a reproducible computational phenomenon? https://justin.cyberelf.link/research/representation-geometry/ Representation geometry 2026-09-01T00:00:00Z Are familiar vector-space descriptions hiding graph, spatial, topological, or other structures that make model behavior easier to explain or control? https://justin.cyberelf.link/research/music-as-high-dimensional-representation/ Music as a high-dimensional representation problem 2026-09-01T00:00:00Z Can musical structure be represented closer to the full acoustic event, without making Western notation the fundamental coordinate system? https://justin.cyberelf.link/research/parameter-strainer/ Parameter strainer 2026-09-01T00:00:00Z Can model weights be treated as signals and progressively transformed toward a smaller target representation before or alongside ordinary quantization? https://justin.cyberelf.link/research/useful-distributed-ai-compute/ Useful distributed AI compute 2026-09-01T00:00:00Z Can distributed training or inference work be independently verified well enough to support robust shared compute infrastructure? https://justin.cyberelf.link/research/models-learning-physics-through-interaction/ Models learning physics through interaction 2026-09-01T00:00:00Z What physical regularities can a model recover from repeated prediction and interaction without receiving the rules first as prose? https://justin.cyberelf.link/research/world-models-and-simulation/ World models and simulation 2026-09-01T00:00:00Z What can agents discover when the environment exposes rules through interaction rather than explaining them in language? https://justin.cyberelf.link/research/gunpla-scale-embodied-agents/ Gunpla-scale embodied agents 2026-09-01T00:00:00Z Could inexpensive modular robotics give a persistent local agent a practical physical environment without starting at humanoid scale? https://justin.cyberelf.link/research/continuous-eeg-adaptive-interface/ Continuous EEG as an adaptive interface 2026-09-01T00:00:00Z Can consumer or research EEG produce a stable enough continuous signal for useful local adaptive interfaces before high-bandwidth invasive BCI exists? https://justin.cyberelf.link/research/instrumenting-opaque-systems/ Instrumenting opaque systems 2026-09-01T00:00:00Z Which arguments about models, agents, signals, or simulations become tractable once the system exposes the right traces and measurements? https://justin.cyberelf.link/research/what-would-replace-the-browser/ What would replace the browser if invented now? 2026-09-01T00:00:00Z If backwards compatibility were removed, what networked application architecture would make sense for current GPUs, storage, cryptography, displays, and AI agents? https://justin.cyberelf.link/research/compute-speed-as-capability-variable/ Compute speed as a capability variable 2026-09-01T00:00:00Z Which agent architectures become qualitatively different when machine deliberation and coordination happen far faster than human conversational time? https://justin.cyberelf.link/research/machine-native-communication/ Machine-native communication 2026-09-01T00:00:00Z What protocols emerge when AI systems optimize communication for other machines rather than for human readability? https://justin.cyberelf.link/research/remove-the-human-representation-first/ Remove the human representation first 2026-09-01T00:00:00Z What computational structure becomes visible when familiar text, notation, or array representations are no longer treated as fundamental? https://justin.cyberelf.link/research/agent-native-languages-and-runtimes/ Agent-native languages and runtimes 2026-09-01T00:00:00Z What should a programming environment look like when AI systems are first-class participants in creating, modifying, and executing programs? https://justin.cyberelf.link/research/player-generated-cards-formal-constraints/ Player-generated cards under formal constraints 2026-09-01T00:00:00Z What happens when a game defines the rules and acceptance tests, while players or agents generate the cards? https://justin.cyberelf.link/research/search-oriented-biological-computation/ Search-oriented biological computation 2026-09-01T00:00:00Z Can unusual search or brute-force framings of genetic systems yield testable computational questions without being mistaken for established biological research?