I will be giving the keynote at EuroMLSys 2026 in Edinburgh on April 27.
In the talk I will be arguing that code used to be the durable artifact. You wrote it once, maintained it, reviewed it. It carried four things implicitly: what the system does, what it remembers, how it runs, and what it touches. Then generation got cheap. Developers started throwing code away and regenerating it. In some places, nobody reviews it at all.
The four responsibilities code was silently carrying do not disappear just because the code does. They become orphans. If you want agentic software to actually work, you need explicit contracts for each of them: intent, state, composition, effect. The ML-systems community already builds pieces of this. ClawVM, our paper at the same workshop, is one such piece: the state contract. The rest of the talk is about the other three, and why assembling them is the blueprint we do not yet have.
What Survives When Code Doesn't?
Large language models have significantly reduced the cost of code generation. An increasing share of code is now produced with AI assistance, and developers increasingly treat implementations as disposable rather than precious. Code is moving away from its traditional role as the primary durable artifact in software development. Yet code was never purely about implementation. The maintained codebase has historically served as the concrete foundation for four essential guarantees: the system's intended behavior (intent), what it carries forward across executions (state), how behavior is organized at runtime (composition), and what it may change in the external environment (effect). A skilled developer using AI still upholds these guarantees through review and expertise. But as human oversight diminishes, whether through autonomous agents, no-code tools, or contexts where no expert review layer exists, the guarantees become fragmented and difficult to enforce. Current agent frameworks seldom provide unified contracts for them. This talk argues that these four guarantees must be made explicit as enforceable contracts, forming the outline of a computational model for agentic software. The ML-systems community, with its roots in operating systems, databases, and distributed computing, is well positioned to lead this effort, and relevant components are already emerging across the field. Many of the pieces exist. The blueprint does not.
More from Edinburgh.