Codex Founder Tibo States That With Some Minor Adjustments, Codex Can Operate Continuously for Days to Complete Complex Tasks
Codex founder Tibo stated that with some minor adjustments, Codex can now operate continuously for days to complete complex tasks. The team is about to release an update to make this capability easier for everyone to use.
Tibo also asked the community on X: What is the most difficult task you have seen GPT-5.5 successfully complete?
In market dynamics, developers and corporate resources are accelerating towards long-duration AI agent tools, with an increase in OpenAI Codex subscriptions and Agent usage. The ability to handle long-term complex tasks has become a core competitive advantage, while short-term conversational tools are under pressure, with capital concentrating on AI engineering platforms that support multi-day autonomous operation.
Source: Public Information
ABAB AI Insight
Tibo, as the head of the Codex project, has previously pushed for simplifications in the agent architecture. This statement marks another significant advancement for Codex from a short-term code generation tool to a true long-duration autonomous agent. The team is addressing core pain points such as context retention, tool invocation persistence, and error recovery through minor adjustments.
In terms of capital pathways, OpenAI is focusing computational power and engineering resources on long-term agent infrastructure. Future updates will lower the user configuration threshold, with the strategic goal of making Codex the default platform for developers handling multi-day complex engineering tasks, while enhancing paid conversion and data flywheel effects through higher intensity usage.
Similar cases include early attempts like Auto-GPT, which failed due to context and persistence issues, and recent orchestrators like Symphony supporting long tasks. Currently, Codex is in a critical transition phase from "assisted coding" to "multi-day autonomous engineering agent."
Essentially, this represents a technological replacement: short-term conversational AI is being replaced by long-duration autonomous agents. The mechanism lies in breakthroughs in model persistent memory, toolchain stability, and error self-recovery capabilities, allowing developers to fully delegate multi-day complex workflows, leading to a shift in pricing power from traditional human coding and short-term tools to AI agent platforms that support reliable long-term operation, significantly raising the productivity ceiling in software engineering.