GPT-6, Sam Altman’s Secret Notebook | OpenAI Playground | OpenAI Login | ChatGPT OpenAI | Turtles AI
Sam Altman, after a controversial launch of GPT-5, anticipates a GPS-6 that will remember users’ habits, tone, and tastes, thanks to an "enhanced memory" shaped in collaboration with psychologists. But the privacy nightmare and the need for neutral personalization remain.
Key points:
- GPT-6 will be equipped with long-term memory to truly adapt to the user.
- OpenAI is collaborating with psychologists to model the experience and monitor well-being, including tone.
- Serious privacy concerns persist, especially regarding unencrypted memory.
- The model will be ideologically neutral by default, but customizable upon request.
Sam Altman, like a magician pulling a rabbit out of the hat after a somewhat shaky launch of GPT-5, announces that GPT-6 will be the model with which AI actually comes down to dinner with you, rather than responding on the fly: equipped with an "advanced memory," it will keep track of your preferences, your style, and those little idiosyncrasies that make you, transforming every interaction into a symphony that picks up exactly where you left off; after all, as he has already mentioned elsewhere, memory is the feature that most impressed him in 2025, even if for now it travels on a track that is not yet armored, because that sensitive data is not encrypted and this, he himself admits, constitutes a serious risk. There is also a dash of psychology: the model, for example, does not limit itself to recording blind data, but measures and calibrates it together with psychologists to track your mood over time, even if these studies remain confidential. And since we live in the age of polarization, GPT-6 will be a "central" alarm clock by default, but it will open the door to becoming super-woke or super-conservative, if the user so requests. Altman, while apologizing for GPT-5’s fluctuations in tone and welcoming GPT-4o’s return at user pressure, already seems to be looking ahead, like an assistant anticipating your question without you even having to type it. In the meantime, however, the question remains: how much customization can be achieved without succumbing to Big Brother? And how do you calibrate those kisses of memory without everything slipping into unethical technological eroticism?
Sam describes it with ironic frankness and a certain trepidation, as if he were designing an artificial friendship he never wants to have to uninstall.


