OpenAI claims loading a narrative memory of the user, written offline from past chats, into the context window before the first prompt

2026-10-08Filings

United States Patent Application Publication No. 2026/0310729 A1, OpenAI OpCo, LLC, "Memory Platform for Generative Response Engines," Application No. 19/195,022, filed April 30, 2025, claiming priority to Provisional Application No. 63/784,739 filed April 7, 2025, published October 8, 2026

Claim 1 is a method in which a system server receives a prompt starting a conversation thread with a user account, inserts a personalized memory state into the context window before inference begins, and sends the prompt and the memory state together to a language model generative response engine. The memory state must contain two things: curated facts in natural language about the user account, and synthesized memories, which the claim defines as natural-language narrative descriptions of the user account generated by a language model that periodically analyzes past conversation threads in an asynchronous offline process before the prompt arrives, each summarizing conceptual connections across multiple threads and stored in a persistent user profile. The dependent claims add reasoning over which parts of the memory are relevant (claim 2), searching an archive of past interactions (claim 3), a short-term memory of recent threads (claim 4), a specialized model instance trained to decide what to synthesize (claim 6), demoting a stale summary to the archive (claim 7), keeping separate memory sets for two distinct usage profiles of one account (claim 8), trimming or summarizing the memory to free room in the context window (claim 10), and a personalization notepad the model writes to (claim 11). The published claim set shows claim 21 canceled and claims 22 and 23 added, so at least one amendment preceded publication; claim 1 as published also recites an "interference operation" where an inference operation is plainly meant. The provisional application was filed April 7, 2025, and OpenAI announced on April 10, 2025 that ChatGPT memory would draw on chat history as well as saved memories.

What it changesThe February 2024 version of ChatGPT memory, which stored facts the user asked it to keep, is more than one year older than the April 7, 2025 provisional, so it is OpenAI's own prior art under Section 102(a)(1) of Title 35 of the United States Code with no grace period, and Nvidia's Publication No. 2026/0288790 A1 on this page carries a 2022 priority date for storing what the user said and applying it later. That is why the claim puts its weight on the synthesized memories: narrative descriptions written offline by a model, summarizing connections across threads, and loaded before the first prompt. Anyone building assistant memory should compare their design to those limitations, because a product that stores only extracted facts, or that writes its summaries at inference time rather than in a background job, is outside claim 1 as published. The three-day gap between the provisional and the launch is the cadence an in-house program should copy: the provisional goes in when the launch date is set, not after the press coverage, because the launch starts the one-year clock in the United States and ends novelty everywhere else.

More on What the AI companies are filing

← All recent developments · Search all developments

Burbage Law is the practice of Jon Burbage, a registered patent attorney. 342 United States patents have issued from applications he wrote; the full list is on the Patents page. More about Jon Burbage.

This is general information about a published decision, agency guidance or piece of legislation. It is not legal advice and does not create an attorney-client relationship. Items concerning offices outside the United States are reported for information only and are not counsel on the law of those countries.