Temper：用成熟原语为 Agent 构建可验证运行时

Strongly agree. Agents amplify whatever primitives they build on, good or bad. Our systems research team at @datadog is building Temper: a higher level of abstraction for software that agents build and operate, made from boring, well-understood primitives. (1) OData for self-describing APIs with typed entities and relationships (2) I/O automata for behavior (3) Actor model for execution (4) Cedar for policy (5) Wasm for safe code sandboxing (6) rust as the language None of these are new, and that’s the point. Composing them lets you describe/specify an application succinctly as entities, state machines, and policies, instead of as the glue code between them. That description is small and precise enough to formally verify (SMT solvers, model checkers, theorem provers), and the runtime is tested under deterministic simulation. An agent can propose any change. A state change only lands if it passes the spec and the policy. https://t.co/cXWpzEMuPL

“Programming language syntax or languages themselves doesn’t matter anymore because the agents just do it.” This drives me nuts because it’s just not true. You need to have solidly designed primitives for agents to work with, otherwise it becomes sloppy soup.