OpenAI claims the buffer that turns a reasoning model's hidden chain of thought into step summaries, and a continuation claims the timeline that displays them

2026-10-22Filings

Listen · Thinking Makes It So, 2026-10-26 · 10 min

All episodes · Subscribe: Apple Podcasts · Spotify

United States Patent Application Publication Nos. 2026/0330193 A1 and 2026/0330194 A1, OpenAI OpCo, LLC, both titled "Visual Reasoning and Multimodal Chain-of-Thought Back-End," Application No. 19/364,606, filed October 21, 2025, claiming priority to Provisional Application No. 63/789,762 filed April 16, 2025, and its continuation, Application No. 19/389,708, filed November 14, 2025, both published October 22, 2026

Claim 1 of the first publication is a method of artificial intelligence reasoning: map an input to tokens, process them with a model that generates reasoning tokens and tool calls, perform the tool calls and queue their responses, buffer the reasoning tokens, determine when to release the buffered tokens and the queued responses to a summarization model, have that model write a step summary of the multi-step reasoning, and return the summary to an application programming interface or to the conversation thread of a user interface. The dependent claims add an ordered queue that holds an output back until its summary is finished (claim 5), a delta threshold for the release based on a difference in token count or an elapsed time (claim 6, which depends from claim 5), chunk sizes set by a latency threshold, a token-generation rate or available summarization resources (claim 3), visibility flags that mark reasoning as displayable or as protected intermediate reasoning (claim 8), and summaries that draw on earlier summaries (claim 10). The specification describes an analysis channel marked non-persistent, whose messages are discarded when the reasoning turn ends. The continuation claims the other end of the wire: a client that receives reasoning messages tagged by type, a text output, an image output, a search result or a code execution result, chooses a visual style, an animation or a display duration from the type, and renders the messages as a timeline that updates in real time, with a collapsed mode that shows only the latest step (claim 2) and search citations that arrive one at a time (claim 7). Both publications are classified only in Cooperative Patent Classification group G06N 3/0475, generative networks, and neither carries a user interface class. OpenAI wrote on September 12, 2024 that it would not show users the raw chain of thought and that the o1 series would show a model-generated summary instead; the provisional was filed April 16, 2025, the day OpenAI released o3 and o4-mini.

What it changesShowing users a model-written summary of a hidden chain of thought was public seven months before the provisional, in OpenAI's own words. Whether that post counts against OpenAI depends on the inventor's one-year grace period under Section 102(b)(1)(A) of Title 35 of the United States Code, but it counts against anyone who files on the idea now, so an application on reasoning summaries filed today has to claim how the summaries are produced, not that they are shown. That is what OpenAI did. Claim 1 recites the plumbing: a buffer, a determination of when to release it, a queue for tool results, and a summarization model that receives a subset of the reasoning tokens at each release. Upon review, I conclude that the release determination is likely where examination will turn, because claim 1 says that the system determines when to release the tokens without saying how, and I would expect the claim to end up closer to claim 6, which ties the release to a difference in token count or an elapsed time. The second lesson is in the continuation, filed twenty-four days after the parent. The server that generates a reasoning feed and the application that displays it are often built and sold by different companies, and a method claim whose steps are split between them risks having no single infringer; the parent claims the server side and the continuation the client side. When drafting for a system that splits work between a server and a client, consider an independent claim for each side.

More on What the AI companies are filing · Prior art, public disclosure and obviousness

← 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.