Version 3

OpenAI launched Dreaming V3 for ChatGPT on June 4, 2026 — a background synthesis system that continuously updates user memory, described by OpenAI as 'more capable and compute-efficient' than prior static memory [3][4]. The update adds a user-facing memory summary page alongside the architecture change [3]. OpenAI initially described the rollout as limited to Plus and Pro users, but The Verge reported it 'rolling out to everyone' [6], leaving the actual scope ambiguous. Startup Anuma positions cross-model portable memory as a structural alternative to platform-locked approaches [12][13], and security researchers have documented real-world LLM-agent exploitation [16] alongside persistent memory's susceptibility to indirect prompt injection [15].

How AI assistants handle persistent memory is becoming a key axis of platform differentiation. OpenAI's Dreaming locks richer context into ChatGPT specifically, while Anuma's portable model treats memory as user-owned infrastructure spanning all models — the two approaches carry different implications for user lock-in and platform economics. The security risks are no longer theoretical: documented real-world LLM agent exploitation and the observation that continuously self-updating systems may not match the version originally safety-reviewed make persistent memory both more capable and harder to audit.

On June 4, 2026, OpenAI announced Dreaming — internally versioned as Dreaming V3 — a redesigned memory architecture for ChatGPT that runs background synthesis processes to keep user context current rather than relying on explicit, static memory entries [1][2]. OpenAI described it as 'a more capable and compute-efficient' architecture [3] that periodically reviews conversation history and updates what it knows about a user's preferences and goals. The update also adds a user-facing memory summary page, giving subscribers visibility into what the system has retained [3][4]. OpenAI framed the change as treating memory as an architecture problem rather than a settings toggle [5]. The initial announcement described the rollout as limited to Plus and Pro users, but The Verge subsequently reported it 'rolling out to everyone' [6], leaving the actual scope ambiguous. Social amplification of the announcement was substantial across Twitter and technology press by June 5 [6][7][8].

The Dreaming launch sits in a broader context of competing approaches to AI memory. Developer practitioners have documented 'compaction amnesia and context rot' — the pattern where models progressively lose track of prior decisions in complex workflows [9] — and argued that structured selective memory retrieval is more reliable than extending context windows [10]. MIT's MeMo research reported a 26% LLM performance improvement by keeping memory architecturally separate from the base model [11]. Anuma, a startup building cross-model portable AI context, positions itself as the infrastructure-layer response to what OpenAI addresses with a platform-locked approach: Rohan Paul described the core problem as 'Most AI workflows break because the user has to carry the context manually' [12][13]. A third-party tool called AgentMemory also appeared aimed at giving Claude Code persistent cross-session memory [14], showing developer communities building their own solutions to the same problem in parallel to platform-native approaches.

The security dimension of persistent AI memory has moved from theoretical to operational. Palo Alto Networks documented that indirect prompt injection can poison AI long-term memory, causing persistent behavioral changes across sessions [15]. Sysdig documented the first publicly known LLM-agent-driven post-exploitation chain in May 2026 [16], and SecurityWeek published a formal benchmark ranking the security posture of 100 AI agents [17]. A structural concern has emerged in practitioner commentary: Danny Livshits observed that a self-evolving agent voids its own safety review, since the version tested is not the version running a month later [18] — a problem that background-synthesis systems like Dreaming make more acute. None of the major providers have publicly addressed how they intend to handle persistent attack surfaces or the safety-review validity problem for continuously updating memory systems.