Agent memory with a frozen model
In a post published on August 26, 2026, Recuris describes two memory loops and a fixed meta-agent that proposes repairs based on execution traces, without changing model weights.
Source: Recuris evolui a memória de agentes com um LLM congelado (arxiv.org). Text prepared with AI from this source.
What happened and what to do
In a post dated August 26, 2026, Recuris describes long-horizon agents that use working and experiential memories to guide tasks. A fixed meta-agent analyzes execution traces and proposes targeted memory repairs. The approach aims to improve agent behavior without changing model weights.
A company could apply this idea by instrumenting its agents to record executions, organize relevant memories, and evaluate recurring failures. An analysis workflow can suggest controlled memory changes, with review and monitoring before deployment; retraining the model is not required to test this type of improvement.
How the consultancy can help
Wendelmaques can diagnose agent memory workflows and risks, define a scoped implementation for trace collection, evaluation, and controlled repairs, and support operation and monitoring on the company's infrastructure.
Next step
Send a short description of your case and the agents involved to receive a scoped diagnosis and implementation proposal.
Let's assess your project
I reply with scope and a proposal.