After six months, ChatGPT will take over your life.

  It's the first time in the history of the world.

OpenAI CEO Ottman once again shocked the tech world.

In a recent conversation with Silicon Valley interns, he bluntly said "the ultimate AI assistant" is only "one model mutation / 6 months".

Over the next six months, we're about to embrace a world where the next generation of ChatGPT can watch your screen, record every meeting and call, and perfectly master the whole context of your life.

We've become extremely practical, and we've only evolved a generation of models.

Note that he's not talking about "stronger memories" or "larger contexts", but continuing to watch your screen, record every meeting and call, and master the perfect contexts of your entire life.

This will be a change in the nature of human-machine relationships: from dialogue to coexistence. But this time, does Ottoman predict reliable?

Over the past two years, Ottman has repeatedly dreamed of a vision: hoping for a very small reasoning model with a trillion tokens in the background that can remember all your life and then continue to add to it.

At the Sequoia AI Ascent conference last May, Ottman made a very compelling observation.

He made a rough but clear distinction:

Elderly people use ChatGPT as an alternative to Google;

People in their 20s and 30s use it as a life coach.

Students are directly involved in the operating system.

A further trend is that young people "have to ask ChatGPT first".

These are all signals of a new form of AI in the future. But this time, he's not just saying "memory", he's directly pointing to screen perception and meeting volume recordings.

In the next six months, the next generation of AI will fully understand you.

This will be a real leap from "occasional use" to "continuous presence", from "single task" to "accumulation of context at a lifetime level". The more life you put into this AI, the more it costs to leave it.

In fact, it has already begun to emerge: ChatGPT's desktop end is now able to access the context on the screen in voice mode, and the enterprise version already has session recording and summary functions.

On April 21 this year, OpenAI launched a codex in its programming product.ChronicleResearch preview functionality.

It generates memories of "what you've been doing lately" through intermittent screenshots + OCR in the background, and saves Markdown files encrypted locally to help the codex understand "that" or two-week-old project without requiring users to repeat it repeatedly.

Following the launch, Greg Brockman, President of OpenAI, said two words: "Surprisingly magical".

Its internal code is:"Telepathy" is an inspiration.As Ottman himself put it: "The word as the name implies, that's what it feels like".

But the Chronicle is all about living life to the fullest.It primarily serves the codex programming workflow, capturing recent screen content in full context rather than continuously recording meetings/calls or building a complete personal file. Users can pause at any time.

Official documents clearly warn against injecting, unencrypted storage locally and requiring user authorization.

Another move by OpenAI is the Dream V3 launch of ChatGPT in June this year.

Dreaming V3 is the latest architecture of the ChatGPT memory system: the backdrop automatically synthesizes memory states from conversational history, with time perception, reduced obsolete information, and enhanced fact memory and preference to follow.

Users can view/edit through memory abstract pages.

The most interesting thing is that it hasAutomatically evolvingIf you say "I'm going to Singapore in July" and the time is up, the system automatically changes the memory to "You went to Singapore in July".

In the internal evaluation of OpenAI, the performance of this update was significantly improved.：

The actual recall rate increased from 67.9% in 2025 to 82.8%;

The following preference is raised from 55.3% to 71.3%;

Time-sensitive accuracy increased from 52.2% to 75.1%.

And the computing efficiency has been increased by about five times, which is why it can be lowered to the free user level.

Dreaming V3 and Chronicle are parallel but different product lines with a focused on-screen context (Codex programming scene), a focused dialogue synthesis (Chat GPT general).

Together, they are closer to the "always online, deeply personalized" direction described by Ottman, but are still in the preview/step-by-step push phase of research.

Putting the lens a little further away, you'll find that over the past few months, almost the entire AI industry has been heading in the same direction at the same time.

Google is on the corporate path.

Memory Bank is available on Gemini's enterprise-level Agent platform. Developers can connect their own AI agents to a pool of session memory capabilities to automatically refine user preferences, key nodes, and explicit instructions from conversations.

This line is currently more of a developer infrastructure and has not yet entered the Gemini dialog box of ordinary consumers, but the direction is consistent with OpenAI.

And Anthropic is not staring.

In March, lasting memory was given directly to all Claude users, with a free paid version available, and the backend summarized your data based on conversations.

In April, Managed Agents on the developer side added a long-term memory disclosure, with enterprise customers such as Netflix taking over the production environment and a 97% reduction in enterprise user errors.

The third party was also down, but tried the opposite.

The goal of Mem0 is to become a cross-platform memory middleman that combines OpenAI, Anthropic, and open-source models so developers don't have to be locked in a building.

This is precisely how the industry has realised that if every giant builds its own memory island, users and developers who are trapped in it will ultimately suffer.

Almost all the giants are now pushing in this direction.

But nobody really wants to come on stage and say that your AI memory is now locked in the platform you use most often. From ChatGPT to Claude, memory is empty, everything starts from scratch.

This is the most concealed of the models.

The model's strength may be topping the rankings for three months, but the thicker the memory, the longer it takes, the higher the cost of leaving.

When you look at this layer, you can see that Ottman's words are not just about the product vision, but also about the business premise.