Sign InOpen Brain
AI EngineerVideoSource Linked

It’s Tokens All The Way Down: How RLMs are Different — Kevin Madura, AlixPartners

RLMs keep large inputs in a programmable environment, letting a model inspect them with code and delegate subsets to submodels instead of attending to every token at once.

AI Engineer · Sep 9, 2026
Open Source Open MarkdownOpen JSON
Source Summary

A recursive language model treats context as a variable inside a Python-style REPL, writes code against it, and can delegate selected work to another model. The talk contrasts this with asking a base model to find **12 numbers across 30,000 tokens** directly.

Practical Implication

Consider the pattern for large logs, data frames, repositories, or other inputs that benefit from deterministic slicing and computation. Define typed inputs and outputs, cap iterations, and return only relevant intermediate results to the main model.

Agent-Ready Context
A recursive language model treats context as a variable inside a Python-style REPL, writes code against it, and can delegate selected work to another model. The talk contrasts this with asking a base model to find **12 numbers across 30,000 tokens** directly.

Consider the pattern for large logs, data frames, repositories, or other inputs that benefit from deterministic slicing and computation. Define typed inputs and outputs, cap iterations, and return only relevant intermediate results to the main model.

The cited benchmark rose from **2.6% to 45.4% accuracy**, but the coding-agent comparisons were described as preliminary and potentially unfair. Cost, stopping behavior, sandboxing, and benefits over a carefully built conventional agent still require task-specific tests.
Connected Context · Feed7 Judgment

This introduces programmatic context manipulation as an alternative to placing a large input directly in a model or distributing it across a fixed agent chain. Deterministic slicing and computation can reserve model attention for selected evidence, with delegation kept optional and bounded. The benchmark supports testing the pattern, but preliminary comparisons leave its advantage over a carefully engineered conventional agent unresolved.

Why We Killed Our Multi-Agent Pipeline — Subbiah Sethuraman and Abhilash Asokan, ZS AssociatesThe RLM pattern aligns with replacing fixed specialist chains: deterministic processing handles selection while one main model retains end-to-end ownership and delegates only bounded investigations.Anthropic's CCA Exam as a Field-Guide for Agentic Engineering — Frank Coyle, UC BerkeleyIts iteration caps, typed boundaries, and selective intermediate returns instantiate the checklist’s guidance to narrow subagent scope, isolate outputs, inspect stopping, and control context growth.Agent swarms and the new model economicsBoth separate coordination context from delegated work, but RLMs add deterministic code over the source context; Cursor’s result suggests the benefit may depend as much on harness structure as on model choice.The New Primitives: Building AI Native Software — Kwindla Kramer, DailyThe RLM supplies a concrete mechanism for the broader primitive of compressed context and selective delegation, while retaining explicit computation and stopping controls around those operations.
Context Map
agentcodingdata#reasoning#subagents#harness-engineering
Uncertainty
The cited benchmark rose from **2.6% to 45.4% accuracy**, but the coding-agent comparisons were described as preliminary and potentially unfair. Cost, stopping behavior, sandboxing, and benefits over a carefully built conventional agent still require task-specific tests.