A write up on the MiMo Code harness. This was posted the other day but didn't get much attention. I thought it deserved more.
The article covers compression, retrieval, tiered memory, deterministic workflows (converting the user's prompt to a program before executing it), and more.
> We observed that the most prominent bottlenecks vary across different time scales: the quality of single-turn decisions within a session is mainly constrained by computation; the continuity of multi-turn tasks within a session is mainly constrained by state management; and improvement across sessions is mainly constrained by the mechanism for distilling experience. These three time scales correspond exactly to computation, memory, and evolution. MiMo Code is designed around these three themes.
A write up on the MiMo Code harness. This was posted the other day but didn't get much attention. I thought it deserved more.
The article covers compression, retrieval, tiered memory, deterministic workflows (converting the user's prompt to a program before executing it), and more.
> We observed that the most prominent bottlenecks vary across different time scales: the quality of single-turn decisions within a session is mainly constrained by computation; the continuity of multi-turn tasks within a session is mainly constrained by state management; and improvement across sessions is mainly constrained by the mechanism for distilling experience. These three time scales correspond exactly to computation, memory, and evolution. MiMo Code is designed around these three themes.