Virtual research collective · DistributedGitHub

Shared research agenda

Agentic systems need better questions before they need bigger claims.

Member labs keep distinct programmes while contributing to a common agenda across agency, trust, domain science, and resilient infrastructure.

01

Agentic and multi-agent systems

Current programme

Research on agents that plan, act, coordinate, and alter external systems over extended periods.

How should an agent explain and record consequential actions?
Which constraints remain effective across long task horizons?
How do agents cooperate when objectives or evidence conflict?
What should count as a reproducible agent experiment?

02

Trustworthy and secure digital systems

Current programme

Work on assurance, privacy, security requirements, protocol trust, and resilience under adversarial conditions.

Which security properties can be checked after deployment?
How should trust change after partial failure?
Where do incentives undermine technical controls?
Which audit records let independent researchers test a claim?

03

Agentic science and domain research

Open direction

A future space for domain-led labs that use AI agents while preserving the methods, evidence, and limits of their field.

Which scientific tasks should agents perform without human substitution?
How can a domain expert inspect an agent’s intermediate work?
What evidence separates a useful result from a plausible output?
How should domain labs report model and data dependence?

04

Resilient decentralized infrastructure

Current programme

Research on programmable blockchains, cyber-physical systems, and emerging infrastructure that must remain dependable under attack.

How should decentralized markets expose manipulation and exclusion?
Which privacy mechanisms remain usable at protocol scale?
How can cyber-physical defenses distinguish faults from attacks?
Where might quantum-resilient security alter system requirements?