Distributed Systems

This is the throughline of the blog. A decade of work on coordination-free programming, conflict-free replicated data types, fault injection, and the engineering practice of building distributed systems that don't lie about their state. The recent AI-agents writing keeps re-discovering problems I spent that decade staring at.

The projects, in rough chronological order:

• Verified Vector Clocks — a Coq-verified vector clock library, exported to Erlang for Riak. Posts walk through the experience of formalizing a distributed primitive in a proof assistant.
• SyncFree / Lasp — a coordination-free distributed programming language. Five years of work, including the EU-funded SyncFree research project. Lasp lets you build distributed applications without coordination, using CRDTs as the underlying data model.
• Partisan — a high-performance, fault-tolerant distributed runtime for Erlang. Replaces the default Erlang distribution layer with one designed for modern cloud environments.
• Filibuster — service-level fault injection testing. Companion to my Ph.D. thesis. Production version at filibuster.cloud.
• Programming Models for Distributed Computing — the seminar curriculum I taught, walking through the foundational papers in the field (Argus, Emerald, Hermes, Promises, etc.).

The current connection to AI agents: see Multi-Agent Systems Have a Distributed Systems Problem — multi-agent LLM systems are rediscovering coordination problems the distributed-systems field solved in the 1970s, just with worse vocabulary.

← Home · Full archive → · Research →