The new GPT-5.5 is hailed as a game-changer in the AI space, and, as a result, OpenAI’s CEO, Sam Altman, is eager to advertise how great it is for general users and coders. In one of his recent posts, Altman now says that the updated Codex is so good that he fears missing out on working with the advanced models.
In a post on X (formerly Twitter), Altman says that he wants to switch to a polyphasic sleep schedule – involving multiple short sleep cycles instead of one long stretch. The reason behind Altman’s declaration is down to this – he fears missing out on the transformative capabilities of the company’s latest AI coding tool.
“I am switching to polyphasic sleep because GPT-5.5 in Codex is so good that I can’t afford to be sleeping for such long stretches and miss out on working,” stated Altman, dropping hints of praise for the new AI model.
Altman’s statement, however, was presented as a context for another analogy detailed in the post, where he mentioned the notion of rapid AI advancement taking away jobs and affecting the economy. The post takes a stand that cutting-edge tools like Codex appear to be accelerating the pace of innovation dramatically, rather than limiting opportunities.
But what is polyphasic sleep?
For those unaware, polyphasic sleep is a pattern of sleeping that includes multiple sessions of sleep within a 24-hour day, instead of once (monophasic) or twice (biphasic). This sleeping pattern involves short and scheduled naps that often total far less than the recommended 7-9 hours of sleep for humans. It is said that polyphasic sleep aims to increase productive waking hours and minimise the total sleep time.
Will advanced AI affect work standards?
Altman’s lighthearted take on the topic of utilising advanced AI tools comes amid the ongoing debate of AI’s impact on the future of work. While a major chunk of the industry believes that AI could reduce the need for human labour, there are others who assert that AI will act like a tool to augment human capabilities, especially in the post-AGI era.
Altman’s statement expands the idea that modern AI tools allow for higher levels of capability augmentation, allowing people to raise the bar of output and get into creative problem-solving, thus freeing up the human mind from basic jobs. With GPT-5.5, OpenAI is promising a major upgrade in the overall performance capability of the AI tool, especially for coders and other professional aspects.
Why is there excitement around GPT-5.5 Codex?
Codex, for those unaware, is OpenAI’s AI-powered coding assistant and is now based on the newly released GPT-5.5 model. With Codex, developers can generate, debug, and optimise code at unprecedented speeds.
So far, developers using the tool have reported significantly faster iteration cycles, thus turning what once took hours or days into minutes. Some of the major upgrades that the latest Codex brings to the table include:
Agentic task execution: The GPT-5.5-based model doesn’t just write code snippets. Now, it can plan, navigate codebases, and execute multi-step tasks like debugging, refactoring, and adding tests.
In-app browser integration: Codex can now operate a bundled browser plugin to view local development servers. This allows it to reproduce visual bugs, click through UIs, and verify fixes in real-time.
Scoped modifications: A standout feature with this variant is its “outcome-first” logic. Instead of massive rewrites, it prioritises precise, minimal changes that preserve existing interfaces and system stability.
Self-correction: The GPT-5.5 Codex is significantly more effective at identifying its own errors and iterating on them without needing constant user intervention.
OpenAI experimented with rolling out Codex across an entire company on NVIDIA infrastructure. “We tried a new thing with NVIDIA to roll out Codex across a whole company, and it was awesome to see it work,” he shared in a related post. NVIDIA engineers reportedly gained notable productivity gains while running the GPT-5.5-based Codex on advanced GB200 NVL72 rack-scale systems.