---
title: "Nicolas Finet on X: Link Sharing and Updates"
url: https://stacklist.com/card/cb462f7e-296c-4d59-887c-f18a7cec17ec
source_url: "https://x.com/nifinet/status/2078851409068654639?s=12"
stack: https://stacklist.com/stack/ebe7c82a-3c5d-43a2-a3e9-ce20328a0dfc
summary: "Nifinet shares an article about building self-improving outbound systems on Codex, referencing Andrej Karpathy's experiment where an agent autonomously ran 700 experiments on training code and improved model performance. The system demonstrates automated optimization through iterative experimentation and benchmark-driven selection."
tags: "self-improving-systems, ai-agents, machine-learning, codex, automation, neural-networks, benchmarking"
key_entities: "Nicolas Finet (person), Andrej Karpathy (person), Codex (technology), self-improving-agent (concept), automated-experimentation (concept)"
classification: "snippet"
content_hash: "sha256:a07a938096a41a01ac6ed40fe5a02eb0907636604b2d7ef0dedd8d1f0af72a07"
acp_version: "0.2"
token_counts_approximate: 96
visibility: public
agent_accessible: true
status: "final"
---

# Nicolas Finet on X: Link Sharing and Updates

Nicolas Finet @nifinet Article How to Build a Self-Improving Outbound System on Codex Earlier this year, Andrej Karpathy (@karpathy) pointed an agent at his own training code and let it run for two days. It ran 700 experiments, kept the 20 that beat the benchmark, and made the model... 2:36 PM · Jul 19, 2026 337.2K Views 16 0 1 6 33 0 3 3 519 0 5 1 9 2.5K 0 2 . 5 K Read 16 replies