Prosecution Insights
Last updated: November 04, 2026
Application No. 18/444,235

ARTIFICIAL INTELLIGENCE MODEL THAT ACTS AS USER INTERMEDIARY TO EXTERNAL ARTIFICIAL INTELLIGENCE MODEL API

Non-Final OA §101§103§112
Filed
Feb 16, 2024
Priority
Feb 21, 2023 — provisional 63/486,191
Examiner
HICKS, AUSTIN JAMES
Art Unit
2142
Tech Center
2100 — Computer Architecture & Software
Assignee
Auddia Inc.
OA Round
1 (Non-Final)
75%
Grant Probability
Favorable
1-2
OA Rounds
6m
Est. Remaining
99%
With Interview

Examiner Intelligence

Grants 75% — above average
75%
Career Allowance Rate
323 granted / 430 resolved
+20.1% vs TC avg
Strong +26% interview lift
Without
With
+25.7%
Interview Lift
resolved cases with interview
Typical timeline
3y 2m
Avg Prosecution
45 currently pending
Career history
489
Total Applications
across all art units

Statute-Specific Performance

§101
13.5%
-26.5% vs TC avg
§103
53.5%
+13.5% vs TC avg
§102
16.1%
-23.9% vs TC avg
§112
14.4%
-25.6% vs TC avg
Black line = Tech Center average estimate • Based on career data from 430 resolved cases

Office Action

§101 §103 §112
Notice of Pre-AIA or AIA Status The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA . Claim Rejections - 35 USC § 101 35 U.S.C. 101 reads as follows: Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title. Claims 1-7 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claims do not fall within at least one of the four categories of patent eligible subject matter because the claimed system has no structural recitation and therefore amounts to software per se. Claims 1, 2, 4-9, 11-16 and 18-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea of a mental concept without significantly more. The claims recite the abstract idea of reviewing and reformulating a question based upon domain knowledge before submitting the question to someone else. This judicial exception is not integrated into a practical application because the additional element of a user interface, general purpose AI, training, domain-specific AI and communication between AI models merely links the abstract idea to the technical field of machine learning. The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements such as a user interface and music playlist amount to insignificant extra solution activity. MPEP 2016.05(g). Claim Rejections - 35 USC § 112 The following is a quotation of the first paragraph of 35 U.S.C. 112(a): (a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention. Claims 1-20 rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention. In claims 1, 8 and 15, Applicant claims “predicting next characters”. This is not described, but it’s also not how these models usually work. Usually, a next token is predicted. Delete this element from the claims. This is same rejection applies to “limited purpose framework”. This is not a term of art and it’s not defined in the specification. The following is a quotation of 35 U.S.C. 112(b): (b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention. Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. In claims 1, 8 and 14, Applicant claims “limited purpose framework”. This is not a term of art and it’s not defined in the specification. Claims 7, 14 and 20 recite, “general-purpose AI includes a propensity to respond to command sets that refer to the predetermined domain of subject matter with a predetermined class of output and revision of command set prompts by the domain-specific AI is configured to correct the propensity.” Propensity is indefinite. Likewise “configured to correct the propensity” is indefinite. “predetermined class of output” is not defined in the specification either. Claim 20 recites, “the general-purpose AI includes a propensity to respond to command sets that requesting musical playlists with repeat songs within the output and revision of command set by the domain-specific AI is configured to replace repeats.” This is unclear, maybe a typo. Claim 15 recites, “revision of the command set to includes incorporating additional parameters…” The “includes” seems like a typo. Claim Rejections - 35 USC § 103 The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action: A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made. Claims 1, 2, 6-9, 13 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Language Models are Few-Shot Learners by Brown et al and Optimizing Prompts for Text-to-Image Generation by Hao et al. Claims 3-5, 10-12 and 15-20 are rejected under 35 U.S.C. 103 as being unpatentable over Language Models are Few-Shot Learners by Brown et al, Optimizing Prompts for Text-to-Image Generation by Hao et al and US20110040707A1 to Theisen et al. Brown teaches claims 1 and 8. A system for revision of generative artificial intelligence (AI) model operators using domain specific command set modification connected to independent AI models, comprising: a user interface that is communicatively connected to a general-purpose AI, the user interface including a command set prompt, wherein the general-purpose AI includes a training regimen that causes the general-purpose AI to respond to command sets by predicting next characters that conform to a command set; and (Brown abs “GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model.” GPT-3 was trained on next character prediction that conforms to a command set. The claim to “predicting next characters” is not described in the specification and usually a next token is predicted. Examiner interprets “next characters” to mean any number of characters, but usually more than one character representing a chunk of a work. GPT3 teaches this.) Brown doesn’t teach a domain specific AI. However, Hao teaches a domain-specific AI that receives data entered into the command set prompt and an output of the general-purpose AI, wherein the domain-specific AI includes a training regimen that causes the domain-specific AI to generate output within a limited purpose framework, the limited purpose framework including revision of the command set based on a predetermined domain of subject matter, (Hao fig. 1 below, “Figure 1: Overview of PROMPTIST training: (1) supervised fine-tuning (SFT) on manually engineered prompts; (2) reinforcement learning (RL) to increase the rewards of generated images after prompt optimization.”) PNG media_image1.png 298 538 media_image1.png Greyscale wherein the user interface is configured to transmit data from the command set prompt to the domain-specific AI and the domain-specific AI is configured to generate revised command set data that is executed on by the general-purpose AI. (Hao teaches this as an optimized prompt in fig. 1.) Hao, Brown and the claims are all using AI to process prompts. It would have been obvious to a person having ordinary skill in the art, at the time of filing, to optimize prompts for the domain because “the performant prompts are often model-specific and misaligned with user input. Instead of laborious human engineering, we propose prompt adaptation, a general framework that automatically adapts original user input to model-preferred prompts.” Hao abs. Hao teaches claims 2 and 9. The system of claim 1, wherein the domain-specific AI is further configured to revise output of the general-purpose AI within the predetermined domain of subject matter. (Hao fig. 1 “results show that the optimized prompts outperform human-engineered ones and the original inputs. Human preference ratings also show consistent improvements across in-domain and out-of-domain prompts. Moreover, we find that reinforcement learning is more favorable than supervised fine-tuning, especially on out-of-domain user inputs. Overall, we show that language models can serve as a prompt interface that optimizes user input into model-preferred prompts.”) Hao teaches claims 3 and 10. The system of claim 2… the domain-specific AI revise the command set to include additional parameters within the command set. (Hao fig. 1 optimized prompt) Hao doesn’t teach a playlist generator. However, Theisen teaches wherein the predetermined domain of subject matter is music playlists, the system further comprising: (Theisen abs “intelligent music selection in a vehicle…”) a media player communicatively connected to the user interface and configured to return a plurality of audio files found on a text-music playlist and execute on those audio files; and (Theisen fig. 2 para 16 “The system comprises a music artificial intelligence module for selecting music and a context aware music player (CAMP) configured to play the selected music.” Theisen para 17 “the context aware music player may be further configured to play music in accordance with user commands.”) wherein the command set requests generation of a musical playlist (Theisen para 17 “the context aware music player may be further configured to play music in accordance with user commands.”) Hao, Theisen and the claims all take user prompts. It would have been obvious to a person having ordinary skill in the art, at the time of filing, to apply Hao to playlist prompts because “[l]istening to music on the radio while driving is common practice for drivers in the United States; however, this can become a safety concern when a driver's attention is diverted from the road to the radio controls.” Theisen para 5. Hao teaches claims 4 and 11. The system of claim 1, wherein the training regimen of the domain-specific AI further includes user-specific data (Hao fig. 1 user prompts) Hao doesn’t teach a play history of a user. However, Theisen teaches user-specific data of a first user. (Theisen para 12 “The system identifies the features the user likes and refines its choices based on the history of responses from the user.”) Hao, Theisen and the claims all take user prompts. It would have been obvious to a person having ordinary skill in the art, at the time of filing, to apply Hao to playlist prompts because “[l]istening to music on the radio while driving is common practice for drivers in the United States; however, this can become a safety concern when a driver's attention is diverted from the road to the radio controls.” Theisen para 5. Theisen teaches claims 5 and 12. The system of claim 4, wherein the user-specific data includes a play history of music associated with the first user. (Theisen para 12 “The system identifies the features the user likes and refines its choices based on the history of responses from the user.”) Hao teaches claims 6 and 13. The system of claim 1, wherein the training regimen of the domain-specific AI further includes a user-informed feedback loop. (Hao sec. 1 “The aesthetic score indicates what degree the generated images are aesthetically pleasing.” The aesthetic score is the user-informed feedback. Hao sec. 2.2 “we employ the aesthetic predictor3 to quantify aesthetic preferences. The predictor builds a linear estimator on top of a frozen CLIP model, which is trained by human ratings in the Aesthetic Visual Analysis [MMP12] dataset.”) Hao teaches claims 7 and 14. The system of claim 1, wherein the general-purpose AI includes a propensity to respond to command sets that refer to the predetermined domain of subject matter with a predetermined class of output and revision of command set prompts by the domain-specific AI is configured to correct the propensity. (Examiner interprets this to mean that the domain-specific AI optimized prompt is better in some way than the original prompt. Hao fig. 2, below, teaches this, “Figure 2: Reward comparison of optimized prompts with other baselines on in-domain and out-of domain data, which indicates that the text-to-image model benefits a lot from our method.” PNG media_image2.png 258 572 media_image2.png Greyscale Brown teaches claim 15. A method for revision of generative artificial intelligence (AI) model operators using domain specific command set modification connected to independent AI models, comprising: receiving, by a user interface that is communicatively connected to a general-purpose AI, a command set that requests generation of a musical playlist, wherein the general-purpose AI includes a training regimen that causes the general-purpose AI to respond to command sets by predicting next characters that conform to the command set; (Brown abs “GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model.” GPT-3 was trained on next character prediction that conforms to a command set. The claim to “predicting next characters” is not described in the specification and usually a next token is predicted. Examiner interprets “next characters” to mean any number of characters, but usually more than one character representing a chunk of a work. GPT3 teaches this.) Brown doesn’t teach a domain specific AI. However, Hao teaches providing, the command set to a domain-specific AI, wherein the domain-specific AI includes a training regimen that causes the domain-specific AI to generate output within a limited purpose framework, the limited purpose framework including revision of command sets based on a predetermined domain of subject matter (Hao fig. 1 below, “Figure 1: Overview of PROMPTIST training: (1) supervised fine-tuning (SFT) on manually engineered prompts; (2) reinforcement learning (RL) to increase the rewards of generated images after prompt optimization.”) PNG media_image1.png 298 538 media_image1.png Greyscale revising, by the domain-specific AI, the command set based on the predetermined domain of subject matter wherein revision of the command set to includes incorporating additional parameters within the command set; and (Hao teaches this as an optimized prompt in fig. 1.) providing the revised command set to the general-purpose AI for execution. (Hao fig. 1 shows giving the optimized prompt to a image generator AI, above.) Hao, Brown and the claims are all using AI to process prompts. It would have been obvious to a person having ordinary skill in the art, at the time of filing, to optimize prompts for the domain because “the performant prompts are often model-specific and misaligned with user input. Instead of laborious human engineering, we propose prompt adaptation, a general framework that automatically adapts original user input to model-preferred prompts.” Hao abs. Hao doesn’t teach a playlist generator. However, Theisen teaches connected to musical playlist requests; (Theisen para 17 “the context aware music player may be further configured to play music in accordance with user commands.”) Hao, Theisen and the claims all take user prompts. It would have been obvious to a person having ordinary skill in the art, at the time of filing, to apply Hao to playlist prompts because “[l]istening to music on the radio while driving is common practice for drivers in the United States; however, this can become a safety concern when a driver's attention is diverted from the road to the radio controls.” Theisen para 5. Hao teaches claim 16. The method of claim 15, further comprising: receiving, by the domain-specific AI, output of the general-purpose AI; and revising, by the domain-specific AI, the output within the predetermined domain of subject matter. (Hao fig. 1 “results show that the optimized prompts outperform human-engineered ones and the original inputs. Human preference ratings also show consistent improvements across in-domain and out-of-domain prompts. Moreover, we find that reinforcement learning is more favorable than supervised fine-tuning, especially on out-of-domain user inputs. Overall, we show that language models can serve as a prompt interface that optimizes user input into model-preferred prompts.” The output from the prompts are rewards in fig. 1.) Theisen teaches claim 17. The method of claim 15, further comprising: requesting, from a media player, a plurality of audio files corresponding to songs included on a text-music playlist output by either of the general-purpose AI or the domain-specific AI; and (Theisen fig. 8 and para 150 “block 166 depicts selecting music based on the predicted next music selection.”) executing, by the media player, the plurality of audio files found on a text-music playlist. (Theisen fig. 8 play the music 168.) Hao teaches claim 18. The method of claim 15, wherein the training regimen of the domain-specific AI (Hao fig. 1 “results show that the optimized prompts outperform human-engineered ones and the original inputs. Human preference ratings also show consistent improvements across in-domain and out-of-domain prompts. Moreover, we find that reinforcement learning is more favorable than supervised fine-tuning…”) Hao doesn’t teach music. However, Theisen teaches the data further includes user specific data of a first user including a play history of music associated with the first user. (Theisen para 12 “The system identifies the features the user likes and refines its choices based on the history of responses from the user.”) Hao teaches claim 19. The method of claim 15, further comprising: training the domain-specific AI via a user-informed feedback loop. (Hao sec. 1 “The aesthetic score indicates what degree the generated images are aesthetically pleasing.” The aesthetic score is the user-informed feedback. Hao sec. 2.2 “we employ the aesthetic predictor3 to quantify aesthetic preferences. The predictor builds a linear estimator on top of a frozen CLIP model, which is trained by human ratings in the Aesthetic Visual Analysis [MMP12] dataset.”) Theisen teaches claim 20. The method of claim 15, wherein the general-purpose AI includes a propensity to respond to command sets that requesting musical playlists with repeat songs within the output and revision of command set by the domain-specific AI is configured to replace repeats. (Examiner interprets this to mean getting rid of repeats somehow. Theisen para 52 “For instance they can say “Next” neutrally indicating they might like the song but just do not want to listen to it right now, or they could say “Next” angrily indicating they do not like the song and don't want to hear it again. This can aid in building up the user's preferences quickly.”) Conclusion Any inquiry concerning this communication or earlier communications from the examiner should be directed to Austin Hicks whose telephone number is (571)270-3377. The examiner can normally be reached Monday - Thursday 8-4 PST. Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Mariela Reyes can be reached at (571) 270-1006. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300. Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000. /AUSTIN HICKS/Primary Examiner, Art Unit 2142
Read full office action

Prosecution Timeline

Feb 16, 2024
Application Filed
Aug 17, 2026
Non-Final Rejection mailed — §101, §103, §112 (current)

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12748999
EXPECTATION VALUE ESTIMATION METHOD AND APPARATUS IN QUANTUM SYSTEM, DEVICE, AND SYSTEM
3y 10m to grant Granted Sep 29, 2026
Patent 12718943
FACILITATING INTERPRETABILITY OF CLASSIFICATION MODEL
4y 6m to grant Granted Aug 25, 2026
Patent 12711096
OPTICAL CO-PROCESSOR ARCHITECTURE USING ARRAY OF WEAK OPTICAL PERCEPTRON
3y 10m to grant Granted Aug 18, 2026
Patent 12711355
SYSTEM AND METHOD FOR AN ADJUSTABLE NEURAL NETWORK
3y 4m to grant Granted Aug 18, 2026
Patent 12705474
REDUCED POWER CONSUMPTION ANALOG OR HYBRID MAC NEURAL NETWORK
4y 6m to grant Granted Aug 11, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

Strategy Recommendation AI-generated — please review before filing

Get a prosecution strategy drawn from examiner precedents, rejection analysis, and claim mapping.
Typically takes 5-10 seconds — AI-generated, attorney review required before filing

Prosecution Projections

1-2
Expected OA Rounds
75%
Grant Probability
99%
With Interview (+25.7%)
3y 2m (~6m remaining)
Median Time to Grant
Low
PTA Risk
Based on 430 resolved cases by this examiner. Grant probability derived from career allowance rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month