01Research domain
Intelligence Systems
Research into efficient inference, autonomous reasoning, adaptive computation, and AI-assisted software development.
Why this domain
Language models are now general-purpose components, but the cost of running them, and the reliability of the systems built on top of them, limit where they can be used. We study both sides: how much of the computation spent on inference is actually necessary, and how software agents can reason about intent before acting on it.
Open questions
- 01 How much of the computation in a forward pass contributes to the output for a given input, and can that be determined cheaply at run time?
- 02 Can inference-time optimizations be made model-agnostic, so that they apply to existing models without retraining?
- 03 When a request is underspecified, how should an agent decide between asking, assuming, and testing a hypothesis?
- 04 What does a verifiable plan look like for a coding agent, and how do you check that the work matches it?
Working themes
Efficient inference
Identifying and avoiding redundant computation during inference, with explicit accounting for quality loss and detection overhead.
Adaptive computation
Spending computation in proportion to the difficulty of an input rather than uniformly across all inputs.
Autonomous reasoning
Explicit hypotheses, plans, and checks as first-class objects in agent behaviour rather than implicit steps in a prompt.
AI-assisted software development
Agents that change real codebases, where every action must be reviewable, attributable, and reversible.
Connections to other domains
-
Cost of computation
Inference efficiency only matters if it survives contact with real hardware, drivers, and runtimes. Measurement methodology is shared with Computing Infrastructure.
Computing Infrastructure → -
Agents in tools
AETHRA's agent bridge is a working case of an AI system editing a complex, stateful environment through reviewable transactions.
Simulation & Real-Time Systems → -
Ambiguity and planning
Reasoning techniques only prove themselves on real, messy tasks, where failures are concrete and observable. Applied systems are where that testing happens.
Applied Research Systems →