Gergely Orosz on OpenAI's agentic software factory

OpenAI is an unusually advanced preview of what happens when agentic coding moves from IDE assistance to company-wide production infrastructure.

Logged at IST: 2026-09-17 16:03 IST

What it is: Gergely Orosz's deep dive into how OpenAI uses Codex and internal agent infrastructure across engineering and non-engineering work.

Gist: The public portion says Codex and ChatGPT Work have become central inside OpenAI, with adoption extending beyond engineers into finance, recruiting, legal, and other functions. The core shift is not just better code completion. OpenAI has wired Codex into repositories, docs, Slack, Notion, Databricks, Datadog, logs, and internal skills, so employees can delegate longer-running work with much richer context.

The most useful artifact is the "agentic software factory" pipeline: humans define outcomes, Codex gathers context, implements changes, babysits tests and CI, routes code through specialist agentic reviews, classifies change risk, handholds deployment, builds monitoring dashboards, watches production, and feeds signals back through systems such as Perf Factory and Sevbot. The result is more code, more PRs, heavier CI load, and pressure to rethink reviews, deployment, observability, and native mobile release bottlenecks.

Newsletter angle: A concrete case study in agentic engineering becoming infra, not just a devtool: the new bottlenecks move to harness quality, context access, CI capacity, risk classification, and production feedback loops.