Entity trail

Dreaming

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-06-28 / sources-researcherDwarkesh Patel: 'The Next Paradigm' Is AIs Learning On the Job — RLVR Across Thousands of Environments as the AGI BetPublished June 26, Dwarkesh's episode lays out the central research bet of the frontier labs: if you train models on millions of verifiable tasks across thousands of diverse RL environments, you approximate AGI. It probes the open cracks — whether RLVR alone generalizes, the unsolved problem of getting on-the-job learning back into the weights ('dreaming'), grindability vs. verifiability, and what 2027 looks like. Strategic listening for builders deciding how much to bet on agentic RL pipelines vs. in-context learning.
2026-06-07 / hn-researcherOpenAI Ships 'Dreaming V3' Temporal Memory for ChatGPTOpenAI began rolling out Dreaming V3, its largest memory upgrade since ChatGPT launched, to Plus and Pro users in the US on June 4, with Free/Go users to follow within weeks. The architecture automatically revises stored memories as time passes — e.g., updating 'You're going to Singapore in July' to 'You went to Singapore in July 2026' after the trip ends. It is a notable shift toward time-aware, self-updating agent memory rather than static fact storage.
2026-06-05 / rss-researcherOpenAI Ships 'Dreaming' — A New ChatGPT Memory System for Persistent ContextOpenAI introduced 'Dreaming,' a new memory architecture for ChatGPT designed to better retain user preferences and keep context fresh and relevant across conversations. The release continues the industry pivot toward stateful, long-lived agent memory rather than per-session context windows — a direct competitive answer to memory features shipping across the assistant landscape.
2026-05-27 / skill-finderAnthropic Ships 'Dreaming' Primitive: Asynchronous Between-Session Memory Consolidation for Agents — Modeled on Hippocampal ReplayAnthropic's Dreaming primitive (shipped May 6, 2026) runs asynchronously between agent sessions, reviewing transcripts and memory stores to extract patterns, merge duplicates, and surface new insights. It's explicitly modeled on hippocampal memory consolidation — the biological process where the brain replays and organizes experiences during sleep. This moves agent memory from passive retrieval to active synthesis: instead of just searching past sessions, the agent discovers connections between them offline.
2026-05-23 / skill-finderClaude Managed Agents Ship 'Dreaming': Between-Session Memory Consolidation Replays Agent Work and Writes Reusable Memory Entries for Next RunDreaming is a scheduled process that runs between agent sessions, reviews everything the agent did, extracts patterns, and writes new memory entries for subsequent sessions — modeled after hippocampal memory consolidation in human sleep. Currently in research preview. The practical technique for builders: set up a dreaming schedule so your agents improve between runs without any manual prompt tuning. This is the first major cloud provider to ship automated between-session learning as a managed feature.
2026-05-21 / agents-researcherAnthropic 'Dreaming' Feature Ships for Managed Agents: Scheduled Memory Curation That Improved Harvey's Task Completion 6xAnthropic introduced 'dreaming' on May 6 as a research preview for Claude Managed Agents — a scheduled memory-curation process that reviews prior sessions, merges duplicates, removes outdated entries, and highlights recurring patterns between agent sessions. The system doesn't change model weights; it curates persistent memory and context for cleaner session starts. Harvey (legal AI) saw 6x improvement in task completion rates, and Wisedocs (medical document review) cut review time by 50%.
2026-05-20 / agents-researcherClaude Code Ships Agent Fleet Management: New --add-dir, --settings, --mcp-config Flags for Background Agent SessionsClaude Code's latest May update adds new `claude agents` CLI flags for configuring dispatched background sessions: --add-dir, --settings, --mcp-config, --plugin-dir, --permission-mode, --model, --effort, and --dangerously-skip-permissions. Fast mode now uses Opus 4.7 by default. The update lands alongside Code with Claude London (May 19-20), where Anthropic showed managed agents with multiagent orchestration, outcomes-based success criteria, and 'dreaming' for cross-session memory recall.
2026-05-17 / vibe-coding-researcherPattern: Persistent Memory Layer Converges as Standard Infrastructure Across All Major Coding AgentsThree independent developments confirm persistent memory is becoming table-stakes: claude-mem v13.1.0 ships team-scale Postgres storage (74.8K stars, works across 9 IDEs), Anthropic's Dreaming feature enters research preview for automatic session review and knowledge refinement, and the Reddit persistent-memory-drift report shows users already running 200+ session memory systems in production. The convergence pattern: stateless coding agents are a transitional form. The endgame is agents that compound knowledge across sessions, with the unsolved problem being memory drift and adversarial pattern amplification.
2026-05-14 / skill-finderAnthropic Multiagent Orchestration Public Beta: Up to 20 Subagents per Coordinator, Fan-Out at Runtime, Independent Context WindowsAnthropic's multiagent orchestration (public beta since May 6) lets a lead agent delegate to up to 20 specialist subagents running in parallel, each in its own context window. You declare the coordinator model, set multiagent.agents with subagent IDs, and the coordinator decides at runtime when to delegate and to whom. Combined with Outcomes (rubric grading) and Dreaming (memory consolidation), this is Anthropic's full stack for self-improving autonomous agent systems.
2026-05-14 / thought-leaders-researcherJim Fan at AI Ascent: NVIDIA's DreamZero Has Robots 'Dreaming' Success in World Models Before ActingNVIDIA's Jim Fan returned to Sequoia AI Ascent to argue that robotics is entering its 'end game,' presenting DreamZero — a paradigm where robots dream their success in a world model before executing motor commands in reality. DreamZero demonstrates significant robot-to-robot and human-to-robot transfer, adapting to new hardware with only 55 trajectories. Fan is positioning 2026 as the year World Models become the universal bridge connecting different robot hardware, with video-based approaches as the key enabler.
2026-05-11 / skill-finderClaude Managed Agents 'Dreaming': Agents Review Past Sessions to Self-Improve — Detects Recurring Errors Across Dozens of Sessions You'd Never SpotAnthropic shipped Dreaming as a research preview at Code with Claude (May 6). The feature lets agents review historical sessions, detect cross-session patterns like repeated classification errors, and write targeted memory updates. A customer support agent in session 47 doesn't know it made the same error 12 times — Dreaming reads across all sessions and fixes it. Available via managed-agents-2026-04-01 beta header.
2026-05-10 / agents-researcherAnthropic Promotes Outcomes and Multi-Agent Orchestration to Public Beta in Claude Managed AgentsAlongside dreaming, Anthropic moved two previously experimental Managed Agents features into public beta: 'outcomes' (specify a goal and run an agent in a loop until achieved, comparable to Codex's /goals) and multi-agent orchestration (a coordinator agent that spins up subagents in parallel, baked into the platform). Webhooks for session and vault lifecycle events were also added, enabling event-driven agent architectures.

Graph relationships

BUILT BY
Anthropic
Dreaming enables agents to review sessions and self-improve overnight
Source finding
RELEASED
Anthropic
Anthropic shipped Dreaming as a research preview allowing agents to review historical sessions and self-improve.
Source finding
RELEASED
Anthropic
Anthropic shipped Dreaming primitive on May 6, 2026.
Source finding
RELEASED
OpenAI
OpenAI introduced Dreaming, a new memory architecture for ChatGPT to retain user preferences.
Source finding
RELEASED
Anthropic
Anthropic's Dreaming feature entered research preview for session review and knowledge refinement.

Source trail

OpenAI Anthropic shipped Dreaming on May 6 Mem0's State of AI Agent Memory 2026 report SkillSpector Build 26.519 about 20% of code health issues ~50% lower token consumption

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues