OpenAI deploys "dreaming" memory system for ChatGPT to actively update user profiles across conversations

@iruletheworldmo People are widely underestimating the importance of memory. Being a form of continual learning, t is perhaps one of the most important things

Try this prompt to find out how #OpenAI ChatGPT #realmemory has insights from how it has been #dreaming about you - I found the response extremely insightful and helpful:

Use your new dreaming capability and come up with your best possible advice for me - no BS, tell it straight and estimate against some sort of metric of how much import I should attach to this advice. Maybe limit to ten items unless you think it’s important that I hear more.

@iruletheworldmo How? It's a garbage update 🤣

Guess you can add "OAI shill" to your bio along with "guy who forecasts wrong constantly."

@iruletheworldmo Let us know if you have any feedback!

@iruletheworldmo What part of the memory update actually feels better in use? I’m curious whether people are seeing real practical gains, like better continuity and fewer repeated explanations, or if it just sounds cool conceptually.

@rohanpaul_ai Finally, I can stop explaining my tech stack to every new chat like it's a first date.

Getting memory right in the enterprise is not about giving agents more context. It is about giving them the right context. This is the design principle behind scene-based reasoning, built specifically for the demands of Fortune 500 deployments. Enterprise memory has to be precise, scoped to the correct operational scene, under governance, at the moment of decision. Infinite memory is not the answer. Bounded, scene-based reasoning is.

@rohanpaul_ai Sticky-note memory was dumb but honest — you knew what was in it and why. Memory that decides for itself what to keep is smarter and quieter, and it starts shaping you without showing its work. The hard part was never remembering more. It's being able to see why something stuck.

@iruletheworldmo Yap yap. Hand over the gpt

@rohanpaul_ai Active memory = a machine that not only remembers your preferences but never argues with them. They sell recall, you get an echo chamber.

@iruletheworldmo Memory looks small until you stop repeating the same context every single day. The control layer is the make-or-break part for me.

@iruletheworldmo Memory is best local…at least for me.

@iruletheworldmo I still dint got it waiting badly

@iruletheworldmo Have they… 🙄

@rohanpaul_ai the flip side of active is the system decides what to keep and what to drop on its own judgment. a stale sticky note at least sits there for you to catch. active memory quietly rewrites the picture, and 'visible so you can review' only helps people who review, which is nobody

@rohanpaul_ai Calling it 'dreaming' is generous. 41% to 67% is the real story. Everything else is packaging.

@rohanpaul_ai the honest answer is, we're only just starting to understand how people want ai systems to remember us. this new dreaming memory system might be the start of an entirely new kind of user experience with chatbots

@rohanpaul_ai