Harness as a Language: A Minimalist Agent Framework With Maximal Expressivity

An assistant chooses the wrong API or sends arguments your product cannot use.

What you would change

Apply the proposed tool-selection or argument-generation change in a sandbox.

Apply the proposed tool-selection or argument-generation change in a sandbox.

Compare against

Your current agent with the same tool schemas and task budget.

1. Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.
2. Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.
3. Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.
4. Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test.

What to measure

Correct completed actions, invalid calls, retries, latency, and cost per completed task.

When to keep it

Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation.

recommendation

test

research area

tool use

technical change

Recursive LLM-invoke primitive replacing agent harnesses

published

2026-09-22

The approach is implementable with third-party APIs and a small team and shows a measured cost and quality gain over specialized systems, but its generality is validated only on two benchmarks and requires workload-specific testing before production adoption.

1. claim 01Paperraft inference
   Read the brief interpretation

   Replaces specialized external systems such as memory frameworks (Letta/MemGPT) and self-improvement harnesses (ACE) with a single recursive code-mode LLM primitive, invoke, plus built-in constraint and monitoring hooks, eliminating manually designed tools and external memory or file systems. Costs are low in engineering complexity but shift burden to prompt design and API token spend, since all history lives as variables in a code environment; reported results run at roughly half the cost of Letta while scoring 8 points higher. Failure modes include unbounded growth of interaction history in the code environment on very long horizons, reliance on the LLM's code-generation reliability for recursive calls, and evaluation limited to two benchmarks, so gains may not transfer to other workloads.

   This is a Paperraft inference. It is not pinned to a verified PDF passage.

   Inspect the source
2. claim 02Paperraft inference

   JAZ invoke outperforms Letta (MemGPT) by 8% at half its cost on the recall-heavy portion of StuLife, and outperforms ACE by 4% at lower cost on AppWorld self-improvement.

   This is a Paperraft inference. It is not pinned to a verified PDF passage.

   Inspect the source