IP Library › Granted Patent US 12,602,549
Granted Patent B1
US 12,602,549 · App. 19/321,173 · Granted Apr 14, 2026

Persistent cognitive machine with curated long term memory

Inventor: Brian Galvin (Silverdale, WA)
Assignee: ATOMBEAM TECHNOLOGIES INC.
G06F40/30G06F16/3325G06F16/3329
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,602,549
App. No.
19/321,173
Filed
Sep 6, 2025
Granted
Apr 14, 2026
Kind
B1
Art Unit
2657
USPC
704/9
Abstract

A system and method for implementing persistent cognitive computation through geometric representation of thought in a dynamic latent manifold. The system encodes inputs into a curved space characterized by time-evolving metric tensors, compression pressure fields derived from Ricci curvature, and goal potential fields that shape attention flow. Cognition occurs through geodesic traversal of this manifold, with attention following paths that minimize cognitive action while balancing semantic density and goal relevance. A Cognitive Dynamics Engine maintains manifold geometry, computing optimal trajectories and managing thought bundle operations including consolidation, expansion, and higher-order abstraction. During idle periods, autonomous dreaming processes reorganize the manifold through perturbation, recombination, and topological surgery. This architecture enables persistent memory through geometric encoding, where frequently accessed concepts develop high-curvature regions and cognitive shortcuts emerge from usage patterns, transforming artificial intelligence from stateless computation to structured motion through shaped memory space.

Claims (46)

1 . A computer system comprising a hardware memory, wherein the computer system is configured to execute software instructions stored on nontransitory machine-readable storage media that:

maintain a latent manifold as a geometric substrate for cognitive operations, wherein the latent manifold evolves through use;

encode inputs into geometric structures within the latent manifold, wherein semantic relationships are represented through geometric properties including distance and curvature;

compute paths through the latent manifold for cognitive processing, wherein the paths are influenced by the geometric structure of the manifold;

store persistent representations as geometric regions within the latent manifold, wherein frequently accessed representations develop characteristic geometric properties that facilitate future access;

modify the geometric structure of the latent manifold based on cognitive operations, wherein successful reasoning patterns create persistent modifications to the manifold geometry; and

generate outputs by traversing the latent manifold and decoding geometric information into user-interpretable responses.

2 . The computer system of claim 1 , wherein the software instructions further:

compute compression pressure fields derived from local curvature of the latent manifold, wherein regions of high semantic density exhibit higher compression pressure that influences path computation.

3 . The computer system of claim 1 , wherein the software instructions further:

organize persistent representations into thought bundles comprising coherent submanifolds of semantically related concepts, wherein the thought bundles support operations including consolidation, expansion, and recombination.

4 . The computer system of claim 1 , wherein the software instructions further:

execute autonomous reorganization of the latent manifold during idle periods, including perturbation of existing structures, synthesis of new connections between disparate regions, and removal of unused or redundant structures.

5 . The computer system of claim 1 , wherein the software instructions further:

implement a distributed thought cache that stores frequently accessed geometric structures, wherein cache hits enable direct response generation without full path computation through the latent manifold.

6 . The computer system of claim 1 , wherein the software instructions further:

track activation energy for each persistent representation, wherein representations with low activation energy undergo thermodynamic decay and eventual removal from the latent manifold.

7 . The computer system of claim 1 , wherein the software instructions further:

maintain bidirectional attention fields that support both forward exploration toward goals and reverse traversal along previously computed paths, enabling backtracking and path revision.

8 . The computer system of claim 1 , wherein the software instructions further:

establish goal potential fields that create attractive forces within the latent manifold, guiding path computation toward semantically relevant regions for achieving specific objectives.

9 . The computer system of claim 1 , wherein the software instructions further:

implement hierarchical organization with multiple nested latent manifolds operating at different levels of abstraction, wherein paths can traverse between abstraction levels through geometric bridges.

10 . A method for a persistent cognitive computation through geometric representation of thought in a dynamic latent manifold, comprising the steps of:

maintaining a latent manifold as a geometric substrate for cognitive operations, wherein the latent manifold evolves through use;

encoding inputs into geometric structures within the latent manifold, wherein semantic relationships are represented through geometric properties including distance and curvature;

computing paths through the latent manifold for cognitive processing, wherein the paths are influenced by the geometric structure of the manifold;

storing persistent representations as geometric regions within the latent manifold, wherein frequently accessed representations develop characteristic geometric properties that facilitate future access;

modifying the geometric structure of the latent manifold based on cognitive operations, wherein successful reasoning patterns create persistent modifications to the manifold geometry; and

generating outputs by traversing the latent manifold and decoding geometric information into user-interpretable responses.

11 . The method of claim 10 , further comprising the step:

computing compression pressure fields derived from local curvature of the latent manifold, wherein regions of high semantic density exhibit higher compression pressure that influences path computation.

12 . The method of claim 10 , further comprising the step:

organizing persistent representations into thought bundles comprising coherent submanifolds of semantically related concepts, wherein the thought bundles support operations including consolidation, expansion, and recombination.

13 . The method of 10 , further comprising the step:

executing autonomous reorganization of the latent manifold during idle periods, including perturbation of existing structures, synthesis of new connections between disparate regions, and removal of unused or redundant structures.

14 . The method of claim 10 , further comprising the step:

implementing a distributed thought cache that stores frequently accessed geometric structures, wherein cache hits enable direct response generation without full path computation through the latent manifold.

15 . The method of claim 10 , further comprising the step:

tracking activation energy for each persistent representation, wherein representations with low activation energy undergo thermodynamic decay and eventual removal from the latent manifold.

16 . The method of claim 10 , further comprising the step:

maintaining bidirectional attention fields that support both forward exploration toward goals and reverse traversal along previously computed paths, enabling backtracking and path revision.

17 . The method of claim 10 , further comprising the step:

establishing goal potential fields that create attractive forces within the latent manifold, guiding path computation toward semantically relevant regions for achieving specific objectives.

18 . The method of claim 10 , further comprising the step:

implementing hierarchical organization with multiple nested latent manifolds operating at different levels of abstraction, wherein paths can traverse between abstraction levels through geometric bridges.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Feb 25, 2026
From: GALVIN, BRIAN
To: ATOMBEAM TECHNOLOGIES INC.
Reel/Frame 073897/0687 →
Continuity (6)
Continuation In Part 19284115 · Jul 29, 2025
Continuation 19051193 · Feb 12, 2025
Provisional Application 63847082 · Jul 19, 2025
Provisional Application 63847096 · Jul 19, 2025
Provisional Application 63847091 · Jul 19, 2025
Provisional Application 63847101 · Jul 19, 2025
References Cited (116)
US 9477783B2 · Barbieri · 2016 [cited by examiner]
US 10009466B2 · Marrelli · 2018 [cited by examiner]
US 11436246B2 · Lee · 2022 [cited by examiner]
US 12141527B2 · Olmstead · 2024 [cited by examiner]
US 12223456B1 · Manohar · 2025 [cited by examiner]
US 12481517B1 · Zhang · 2025 [cited by examiner]
US 20020091801A1 · Lewin · 2002 [cited by examiner]
US 20050197992A1 · Kipersztok · 2005 [cited by examiner]
US 20150363386A1 · Song · 2015 [cited by examiner]
US 20160034305A1 · Shear · 2016 [cited by examiner]
US 20170337287A1 · Gill · 2017 [cited by examiner]
US 20180174055A1 · Tirumale · 2018 [cited by examiner]
US 20190174514A1 · Ramesh · 2019 [cited by examiner]
US 20200034776A1 · Peran · 2020 [cited by examiner]
US 20200034914A1 · Boss · 2020 [cited by examiner]
US 20200073983A1 · Sen · 2020 [cited by examiner]
US 20200097554A1 · Rezagholizadeh · 2020 [cited by examiner]
US 20200285704A1 · Rajani · 2020 [cited by examiner]
US 20200336562A1 · Luft · 2020 [cited by examiner]
US 20200351344A1 · Das Gupta · 2020 [cited by examiner]
US 20200387809A1 · Banerjee · 2020 [cited by examiner]
US 20210027862A1 · Wei · 2021 [cited by examiner]
US 20210073808A1 · Gu · 2021 [cited by examiner]
US 20210110306A1 · Krishnan · 2021 [cited by examiner]
US 20210406224A1 · Neufeld · 2021 [cited by examiner]
US 20220138156A1 · Wang · 2022 [cited by examiner]
US 20220207362A1 · Meyerson · 2022 [cited by examiner]
US 20230142467A1 · Kaplanyan · 2023 [cited by examiner]
US 20230177281A1 · Kamath · 2023 [cited by examiner]
US 20230229722A1 · Ishii · 2023 [cited by examiner]
US 20230316006A1 · Tunstall-Pedoe · 2023 [cited by examiner]
US 20230362091A1 · Michaud · 2023 [cited by examiner]
US 20230365138A1 · Park · 2023 [cited by examiner]
US 20240046318A1 · Muriqi · 2024 [cited by examiner]
US 20240095460A1 · Xu · 2024 [cited by examiner]
US 20240104391A1 · Higgins · 2024 [cited by examiner]
US 20240160955A1 · Zhao · 2024 [cited by examiner]
US 20240176805A1 · Liu · 2024 [cited by examiner]
US 20240256965A1 · Chung · 2024 [cited by examiner]
US 20240303496A1 · Nair · 2024 [cited by examiner]
US 20240338393A1 · Malon · 2024 [cited by examiner]
US 20240354320A1 · Procter · 2024 [cited by examiner]
US 20240386015A1 · Crabtree · 2024 [cited by examiner]
US 20240411809A1 · Najafirad · 2024 [cited by examiner]
US 20240419905A1 · Mai · 2024 [cited by examiner]
US 20240428008A1 · Abraham · 2024 [cited by examiner]
US 20250028882A1 · Ataei · 2025 [cited by examiner]
US 20250094455A1 · Bista · 2025 [cited by examiner]
US 20250148203A1 · Pan · 2025 [cited by examiner]
US 20250156684A1 · Liu · 2025 [cited by examiner]
US 20250165718A1 · Seo · 2025 [cited by examiner]
US 20250191369A1 · Huang · 2025 [cited by examiner]
US 20250225160A1 · Heisler · 2025 [cited by examiner]
US 20250259041A1 · Crabtree · 2025 [cited by examiner]
US 20250259042A1 · Crabtree · 2025 [cited by examiner]
US 20250259043A1 · Crabtree · 2025 [cited by examiner]
US 20250259044A1 · Crabtree · 2025 [cited by examiner]
US 20250259082A1 · Crabtree · 2025 [cited by examiner]
US 20250259085A1 · Crabtree · 2025 [cited by examiner]
US 20250259696A1 · Crabtree · 2025 [cited by examiner]
US 20250259724A1 · Crabtree · 2025 [cited by examiner]
US 20250306680A1 · Zhou · 2025 [cited by examiner]
US 20250307672A1 · Esencan · 2025 [cited by examiner]
US 20250342344A1 · Gao · 2025 [cited by examiner]
US 20250348707A1 · Myers · 2025 [cited by examiner]
US 20250349407A1 · Crabtree · 2025 [cited by examiner]
CN 110489087A · 2019 [cited by examiner]
CN 106156083B · 2020 [cited by examiner]
CN 113255364A · 2021 [cited by examiner]
CN 113590836A · 2021 [cited by examiner]
CN 114020885A · 2022 [cited by examiner]
CN 114509825A · 2022 [cited by examiner]
CN 115525768A · 2022 [cited by examiner]
CN 114818740B · 2023 [cited by examiner]
CN 116775846A · 2023 [cited by examiner]
CN 116795958A · 2023 [cited by examiner]
CN 116860987A · 2023 [cited by examiner]
CN 117056524A · 2023 [cited by examiner]
CN 117435923A · 2024 [cited by examiner]
CN 117557425A · 2024 [cited by examiner]
CN 117610586A · 2024 [cited by examiner]
CN 117667918A · 2024 [cited by examiner]
CN 118052289A · 2024 [cited by examiner]
CN 118211649A · 2024 [cited by examiner]
CN 118296115A · 2024 [cited by examiner]
CN 114880429B · 2024 [cited by examiner]
CN 118611907A · 2024 [cited by examiner]
CN 118657201A · 2024 [cited by examiner]
CN 118821800A · 2024 [cited by examiner]
CN 118964538A · 2024 [cited by examiner]
CN 119005308A · 2024 [cited by examiner]
CN 119179786A · 2024 [cited by examiner]
CN 119336900A · 2025 [cited by examiner]
CN 119443288A · 2025 [cited by examiner]
CN 119539026A · 2025 [cited by examiner]
CN 119918654A · 2025 [cited by examiner]
CN 119988690A · 2025 [cited by examiner]
CN 120030114A · 2025 [cited by examiner]
CN 119150237B · 2025 [cited by examiner]
CN 120163226A · 2025 [cited by examiner]
CN 120298880A · 2025 [cited by examiner]
CN 120670651A · 2025 [cited by examiner]
GB 2563183A · 2018 [cited by examiner]
JP H04205030A · 1992 [cited by examiner]
KR 20160121997A · 2016 [cited by examiner]
KR 102721100B1 · 2024 [cited by examiner]
KR 20240157317A · 2024 [cited by examiner]
TW 202526694A · 2025 [cited by examiner]
WO WO2023200762A1 · 2023 [cited by examiner]
WO WO2025100557A1 · 2025 [cited by examiner]
WO WO2025136527A1 · 2025 [cited by examiner]
Author={Sarfati, Rapha{\″e}l and Liu, Toni JB and Boull{\′e}, Nicolas and Earls, Christopher J}, title={Lines of thought in large language models}, journal={arXiv preprint arXiv:2410.01545}, year={2024}, pp. 1-20 (Year:… [cited by examiner]
Gao, Hang & Zhang, Yongfeng; “Memory Sharing for Large Language Model Based Agents”, arXiv:2404.09982v2, Jul. 5, 2024. [cited by applicant]
Gim, In, et al; “Prompt Cache: Modular Attention Reuse for Low-Latency Inference”, arXiv:2311.04934v2, Apr. 2024. [cited by applicant]
Ramirez, Guillem, et al; “Cache & Distil: Optimising API Calls to Large Language Models”, arXiv:2310.13561v1, Oct. 20, 2023. [cited by applicant]
Schroeder, Luis Gaspar; “VectorQ: Advanced Semantic Prompt Caching with Dynamic Thresholds and Performance-Based Clustering”, Technical University of Munich, Nov. 26, 2024. [cited by applicant]
Cited By (1)
US 12,710,935