Imagine That IP Law Blog

Introduction
The person of ordinary skill in the art is central to many patent-law inquiries, but not in the same way across the Patent Act. Section 103 expressly asks whether the claimed invention would have been obvious to a POSITA. Section 112 often turns on what the skilled artisan would have understood from the disclosure. Claim construction likewise depends on how claim language would be understood in view of the intrinsic record. Section 101 is different. Unlike § 103, § 101 does not ask what would have been obvious to a POSITA. The POSITA enters eligibility indirectly, through factual predicates that bear on claim construction, conventionality, technological improvement, mental-process characterization, and practical application.

That derivative role has become more important because eligibility disputes increasingly turn on technical characterizations. Under Alice, courts first ask whether the claims are directed to a judicial exception and, if so, whether the claim elements, individually or as an ordered combination, add enough to transform the claim into patent-eligible subject matter. Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 573 U.S. 208, 217–18 (2014). Although that framework is legal in structure, it often depends on factual propositions: whether an asserted improvement is actually technological, whether a claim changes computer functionality, whether a limitation can practically be performed in the human mind, whether an ordered combination was well-understood, routine, and conventional, and whether the specification would have conveyed the relevant improvement to a skilled artisan.

The USPTO’s recent Subject Matter Eligibility Declaration (“SMED”) memoranda attempt to formalize that factual dimension in prosecution. The December 4, 2025 examiner-facing memorandum states that a SMED may show how “one of ordinary skill in the art would interpret a specification that describes a technological improvement,” may provide facts about the state of the art, and may supply objective evidence of how the invention improved upon that state of the art. USPTO Memorandum to Patent Examining Corps, Subject Matter Eligibility Declarations, at 3 (Dec. 4, 2025) (“SMED Examiner Memo”). The April 30, 2026 practitioner memorandum reframes the same idea as a “constructive model” for technologies whose real-world operation may not be readily visible from the paper record, including software, artificial intelligence, quantum technologies, and diagnostics. USPTO Memorandum to All Patent Applicants and Patent Practitioners, Best Practices for Submission of Rule 132 Subject Matter Eligibility Declarations, at 2 (Apr. 30, 2026) (“April 2026 SMED Memo”).

The guidance is important, but it should not be overstated. Federal Circuit law recognizes factual issues in eligibility, especially after Berkheimer and Aatrix. See Berkheimer v. HP Inc., 881 F.3d 1360, 1368–70 (Fed. Cir. 2018); Aatrix Software, Inc. v. Green Shades Software, Inc., 882 F.3d 1121, 1128–30 (Fed. Cir. 2018). But the Federal Circuit remains claim-centric. A declaration cannot rewrite the claims, cannot import unclaimed technological mechanisms from the specification, cannot supply missing disclosure, and cannot convert a result-oriented claim into a patent-eligible technological improvement.

The POSITA therefore has a real but derivative role in § 101. The skilled artisan may illuminate what the filed specification and claims already convey. The skilled artisan may help rebut factual assertions about conventionality, mental processes, or the state of the art. But the POSITA cannot supply the “something more” that the claims themselves fail to recite.

The SMED Memoranda: Status, Chronology, and Hierarchy
The SMED materials should be understood chronologically. On December 4, 2025, the USPTO issued a public-facing best-practices memorandum for applicants and practitioners and a separate examiner-facing memorandum. The public memorandum recommended that SMEDs be filed as separate Rule 132 declarations rather than combined with declarations addressing other statutory issues. USPTO Memorandum to All Patent Applicants and Patent Practitioners, Best Practices for Submission of Rule 132 Subject Matter Eligibility Declarations, at 1–4 (Dec. 4, 2025) (“December 2025 Public SMED Memo”). The examiner memorandum instructed examiners that applicants may voluntarily submit SMEDs under existing Rule 132 practice and that such evidence may be useful where eligibility is not readily apparent. SMED Examiner Memo, at 2.

The April 30, 2026 practitioner memorandum then superseded the December 4, 2025 Director’s public best-practices memorandum, but did not supersede the examiner-facing memorandum. The April memorandum expressly states that it supersedes the December 4, 2025 Director’s “Best Practices” memorandum, while also stating that no substantive changes were intended to the December 4, 2025 examiner memorandum or updated MPEP guidance. April 2026 SMED Memo, at 1–3. Thus, the December public memorandum is best treated as historical and superseded; the December examiner memorandum remains relevant as examiner-facing guidance; and the April 2026 memorandum is the operative practitioner-facing best-practices document.

That hierarchy matters because the memoranda do not change governing law. The examiner memorandum says it is consistent with existing USPTO guidance and does not announce new USPTO practice or procedure. SMED Examiner Memo, at 2. Nor could the memoranda bind the Federal Circuit. The Federal Circuit has repeatedly made clear that USPTO guidance does not control the judicial eligibility inquiry. See In re Rudy, 956 F.3d 1379, 1382–83 (Fed. Cir. 2020); ENCO Sys., Inc. v. DaVincia, LLC, 845 F. App’x 953, 957–58 (Fed. Cir. 2021). The SMED memoranda are therefore best understood as prosecution guidance: they explain how applicants may create an evidentiary record within the Office, not how courts must decide eligibility.

The USPTO’s SMED Framework Gives the POSITA an Evidentiary Role in § 101
The examiner memorandum places the POSITA into the eligibility record in a defined way. It states that any person with knowledge of the asserted facts may sign a SMED, including an inventor, an inventor’s co-worker, an independent expert, or another knowledgeable witness. SMED Examiner Memo, at 2. It also identifies the core relevance requirement: there must be a nexus between the invention as claimed and the evidence provided. Id. at 3.

In the eligibility context, the memorandum states that a SMED may show how a POSITA would interpret a specification that describes a technological improvement. Id. The memorandum further explains that a SMED may provide facts describing the state of the art at the time of filing, objective evidence showing how the invention improved upon the state of the art, or a factual basis for determining that a POSITA would have concluded that the invention improved the underlying technology. Id.

That language gives the POSITA three related functions. First, the POSITA may serve as an interpretive witness, explaining what the filed specification would have conveyed to a skilled artisan. Second, the POSITA may serve as a technological-context witness, explaining the state of the art at the time of filing. Third, the POSITA may serve as a factual rebuttal witness, challenging examiner assertions that claim elements are conventional, mental, generic, or lacking technological effect.

The April 2026 memorandum strengthens that framing by describing a SMED as a “constructive model” for less tangible technologies. April 2026 SMED Memo, at 2. The memorandum analogizes SMEDs to a modern substitute for a working model, particularly in fields where innovation may be reflected in software architecture, AI training behavior, data structures, or diagnostic workflows rather than visible mechanical structure. Id. It states that a SMED may show how the claimed invention, supported by the specification, is “better, cheaper, faster, and/or more efficient.” Id.

But the same memoranda impose important limits. The evidence must have a nexus to the claimed invention. The declaration cannot improperly supplement the specification. And the April memorandum states directly that “[c]laims always define the invention.” April 2026 SMED Memo, at 3. Thus, even under the USPTO’s own guidance, the POSITA is not a free-ranging eligibility expert. The POSITA is a factual witness to what the claims and disclosure already support.

Federal Circuit Law Supports a Limited POSITA Role Through Factual Predicates in Alice
Federal Circuit precedent supports a limited role for POSITA evidence in § 101. Berkheimer is the central case. There, the Federal Circuit held that whether a claim element or ordered combination is well-understood, routine, and conventional to a skilled artisan is a question of fact. Berkheimer, 881 F.3d at 1368–70. The court also emphasized that the mere fact that something appears in the prior art does not establish that it was well-understood, routine, and conventional. Id. at 1369.

Aatrix reinforced the same principle at the pleading stage. The Federal Circuit held that factual allegations that claim elements were not routine or conventional, or that the claimed invention improved computer functionality, could preclude dismissal under Rule 12. Aatrix, 882 F.3d at 1128–30. Together, Berkheimer and Aatrix create the doctrinal opening for POSITA evidence in eligibility disputes. If conventionality is factual, then evidence about what skilled artisans understood, used, or regarded as routine at the relevant time can matter.

Other Federal Circuit cases likewise leave room for skilled-artisan evidence where eligibility depends on whether the claims recite a technological improvement. In Enfish, the Federal Circuit held that claims directed to a self-referential table for a computer database were not abstract because they focused on a specific improvement in computer functionality. Enfish, LLC v. Microsoft Corp., 822 F.3d 1327, 1335–39 (Fed. Cir. 2016). A SMED modeled on Enfish could explain the state of database technology at filing, identify the claimed data structure, and show why a POSITA would understand that structure to improve computer operation.

McRO provides a similar model. There, the claims used particular rules to automate lip synchronization and facial expression animation. The Federal Circuit held that the claims were not directed merely to an abstract idea because they recited a specific asserted improvement in computer animation. McRO, Inc. v. Bandai Namco Games Am. Inc., 837 F.3d 1299, 1313–16 (Fed. Cir. 2016). A POSITA declaration could be relevant to explain why the claimed rule structure changed the way computer animation was performed, rather than merely invoking a computer to implement an abstract result.

BASCOM is especially important at Alice step two. The Federal Circuit held that an inventive concept may reside in a nonconventional and non-generic arrangement of otherwise known components. BASCOM Glob. Internet Servs., Inc. v. AT&T Mobility LLC, 827 F.3d 1341, 1349–52 (Fed. Cir. 2016). That holding maps directly onto SMED practice. If an examiner asserts that each element was conventional, a SMED may address whether the ordered combination, as claimed, was conventional to a POSITA.

SRI and CardioNet also illustrate the point. In SRI, the Federal Circuit held that claims directed to network monitoring using multiple network monitors were not abstract where they improved computer network technology. SRI Int’l, Inc. v. Cisco Sys., Inc., 930 F.3d 1295, 1303–04 (Fed. Cir. 2019). In CardioNet, the Federal Circuit held that claims directed to detecting atrial fibrillation and atrial flutter using beat-to-beat timing variability were eligible because they recited a specific improvement in cardiac monitoring technology. CardioNet, LLC v. InfoBionic, Inc., 955 F.3d 1358, 1368–71 (Fed. Cir. 2020). In each instance, skilled-artisan evidence may help explain the technological problem, the claimed mechanism, and the improvement over prior practice.

The USPTO’s examiner memorandum expressly models several SMED examples on cases like SRI, Enfish, McRO, Classen, and BASCOM. SMED Examiner Memo, at 4–7. That is significant. The Office is not purporting to create an independent eligibility regime. It is attempting to use Rule 132 declarations to demonstrate that a pending claim fits within eligibility pathways already recognized in Federal Circuit law.

Desjardins and Recentive as Competing Poles for AI Eligibility
The tension between USPTO guidance and Federal Circuit law is most visible in artificial intelligence. The USPTO’s preferred path is reflected in Ex parte Desjardins. The Federal Circuit’s limiting rule is reflected in Recentive Analytics.

Desjardins involved claims directed to training a machine-learning model on a second task while protecting performance on a first task. The claim required determining measures of importance for parameters of a model trained on a first task, obtaining second training data, and training the model on the second task by adjusting parameter values to optimize performance on the second task while protecting performance on the first task. Ex parte Desjardins, Appeal No. 2024-000567, slip op. at 2–3 (P.T.A.B. Sept. 26, 2025) (Appeals Review Panel), designated precedential Nov. 4, 2025. The specification identified the relevant technical benefits: using a single model for multiple tasks, reducing storage requirements, reducing system complexity, and protecting knowledge about previous tasks. Id. at 3.

The Appeals Review Panel vacated the Board’s sua sponte § 101 rejection. The panel acknowledged that the claims recited at least one abstract idea in the form of a mathematical calculation, but concluded that the claims integrated that abstract idea into a practical application. Id. at 6–10. The critical point was not that the claims merely used AI. Rather, the claims reflected an improvement in how the machine-learning model itself operated. The panel identified, in particular, the claim limitation requiring adjustment of parameter values to optimize performance on the second task while protecting performance on the first task. Id. at 9.

Desjardins therefore represents the USPTO’s preferred AI-eligibility pathway: claims that improve the model, training architecture, computational performance, storage requirements, system complexity, or data structures themselves. The decision is also careful to maintain the claim-centric limit. The panel stated that an assertion in the specification alone is insufficient unless the claim itself reflects the disclosed improvement. Id. at 8. That point aligns Desjardins with Enfish, not with a broad rule that AI claims are categorically favored.

Recentive Analytics supplies the counterweight. There, the Federal Circuit held that claims applying generic machine-learning techniques to scheduling and event planning were patent ineligible. Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025). Recentive limits any broad reading of the USPTO’s AI-friendly posture. An invention is not eligible merely because it uses machine learning, nor because machine learning is applied in a new business, media, or scheduling context. The claims must recite a specific improvement in machine-learning technology, computer functionality, training architecture, data structure, inference efficiency, or another technological mechanism.

Desjardins and Recentive thus mark competing poles. Desjardins illustrates how an AI claim may be eligible when it improves the functioning of the model itself. Recentive illustrates how an AI claim may be ineligible when it applies generic machine learning to an otherwise abstract field of use. The practical lesson is that AI eligibility should not turn on the label “AI.” It should turn on whether the claims recite a technological improvement in the AI system or computational architecture itself.

The Core Tension: USPTO Guidance Invites Contextual Evidence, While the Federal Circuit Remains Claim-Centric
The SMED memoranda contemplate that a declaration may explain how a skilled artisan would understand the specification, why the invention improves technology, and how the claimed invention operates in the real world. SMED Examiner Memo, at 3–7; April 2026 SMED Memo, at 2–4. Federal Circuit law permits such evidence in some circumstances, but only where the claim itself recites the relevant technological improvement.

ChargePoint illustrates the limitation. The claims involved network-controlled electric vehicle charging stations, but the Federal Circuit held that the claims were directed to the abstract idea of communication over a network for device interaction. ChargePoint, Inc. v. SemaConnect, Inc., 920 F.3d 759, 766–74 (Fed. Cir. 2019). The technological environment did not save the claims. The claims needed to recite an improvement in charging-station technology or network operation, not merely use networking to control charging stations.

Simio reflects the same concern. Although the claims involved simulation-modeling software, the Federal Circuit concluded that making simulation modeling easier through a graphical process did not supply eligibility where the claims were directed to an abstract idea implemented on a generic computer. Simio, LLC v. FlexSim Software Prods., Inc., 983 F.3d 1353, 1360–65 (Fed. Cir. 2020). A POSITA might well have understood the software to be useful, but usefulness is not eligibility. The claim must recite the technological means of improvement.

American Axle presents the tension in perhaps its starkest form. The claimed method involved manufacturing a shaft assembly to attenuate vibration, yet the Federal Circuit held that the claims were directed to a natural law without sufficiently reciting the means of achieving the claimed result. Am. Axle & Mfg., Inc. v. Neapco Holdings LLC, 967 F.3d 1285, 1292–1301 (Fed. Cir. 2020). Even in a mechanical manufacturing context, the court required the claims to do more than state a desired result. A skilled artisan’s understanding of how to practice an invention cannot replace claim limitations that apply the natural law or abstract idea in a specific technological way.

The USPTO memoranda are therefore best understood as prosecution tools, not as a relaxation of Federal Circuit eligibility doctrine. The April 2026 memorandum’s “better, cheaper, faster, and/or more efficient” language can be powerful when those benefits are grounded in claim limitations and supported by the specification. April 2026 SMED Memo, at 2. But under ChargePoint, Simio, American Axle, and Recentive, a declaration that merely characterizes the invention as faster or more efficient will not rescue claims that recite only a result.

The POSITA Can Explain Technological Meaning, But Cannot Supply Missing Claim Limitations
The proper reconciliation is that the POSITA can explain technological meaning, but cannot supply legal sufficiency.

That distinction appears in the SMED memoranda themselves. The examiner memorandum states that a declaration cannot supply information that was required to be present in the original disclosure. SMED Examiner Memo, at 3. It further states that a claimed invention may be an eligible technological improvement when the specification provides sufficient detail such that a POSITA would recognize the claimed invention as providing a technological improvement and the claims reflect the disclosed improvement. Id. The final clause is critical. The claims must reflect the improvement.

The April 2026 memorandum makes the same point more directly: “Claims always define the invention.” April 2026 SMED Memo, at 3. That statement harmonizes the guidance with Federal Circuit law. A SMED may explain what the original disclosure conveyed to a skilled artisan, but it cannot convert an unclaimed embodiment into a claimed technological improvement.

MyMail provides a useful but narrow example of how POSITA evidence may matter. The Federal Circuit vacated a Rule 12 eligibility ruling because the district court failed to resolve a claim-construction dispute before deciding § 101. MyMail, Ltd. v. ooVoo, LLC, 934 F.3d 1373, 1378–80 (Fed. Cir. 2019). Because claim construction is assessed from the standpoint of a POSITA, MyMail confirms that skilled-artisan understanding can affect eligibility where claim meaning is genuinely disputed. But MyMail does not hold that expert testimony can create eligibility untethered to the claims. It confirms only that a court must know what the claims mean before deciding whether they are eligible.

CardioNet is similar. The eligibility outcome depended on the claim’s recitation of a specific cardiac-monitoring improvement, not merely on the specification’s description of cardiac technology. CardioNet, 955 F.3d at 1368–71. POSITA evidence may explain why the claimed beat-to-beat variability analysis improved cardiac monitoring, but the claim still had to recite the technological technique.

This is the key boundary for SMEDs. The declaration may explain the significance of limitations already in the claim. It may not write those limitations into the claim.

POSITA Evidence and the Evidentiary-Weight Problem
A second tension concerns evidentiary weight. The USPTO guidance allows a SMED to be signed by an inventor, co-worker, independent expert, or another person with knowledge of the relevant facts. SMED Examiner Memo, at 2. But a declaration’s admissibility does not guarantee its probative force. This is particularly important when the declaration is inventor-only and rests on uncorroborated, prosecution-driven assertions.

Patent law has long been skeptical of interested testimony offered to establish technically significant historical facts. In The Barbed Wire Patent, the Supreme Court cautioned against relying on oral testimony concerning prior use, noting the dangers of faded memory, exaggeration, and self-interest. The Barbed Wire Patent, 143 U.S. 275, 284–85 (1892). Eibel Process likewise emphasized the superiority of contemporaneous records over later testimonial reconstruction. Eibel Process Co. v. Minn. & Ont. Paper Co., 261 U.S. 45, 60–61 (1923).

The Federal Circuit carried that concern forward in modern corroboration doctrine. In Price v. Symsek, the court held that inventor testimony standing alone is insufficient to prove conception, derivation, or priority, and that corroboration is assessed under a rule of reason. Price v. Symsek, 988 F.2d 1187, 1193–95 (Fed. Cir. 1993). Woodland Trust and Finnigan likewise reflect skepticism toward uncorroborated testimony concerning historical patent facts. Woodland Tr. v. Flowertree Nursery, Inc., 148 F.3d 1368, 1371–73 (Fed. Cir. 1998); Finnigan Corp. v. U.S. Int’l Trade Comm’n, 180 F.3d 1354, 1367–70 (Fed. Cir. 1999). Singh v. Brake states the policy concern directly: an interested person may be tempted to describe events in an unjustifiably self-serving way. Singh v. Brake, 222 F.3d 1362, 1367 (Fed. Cir. 2000).

EmeraChem provides a close analogue. There, the Federal Circuit rejected a declaration that amounted to an unsupported “naked assertion” and distinguished declarations supported by explanatory detail or circumstantial evidence. EmeraChem Holdings, LLC v. Volkswagen Grp. of Am., Inc., 859 F.3d 1341, 1345–48 (Fed. Cir. 2017). Although EmeraChem was not a § 101 case, its evidentiary lesson applies naturally to SMEDs: conclusory testimony carries little weight when it lacks factual context, documentary support, or objective corroboration.

The SMED memoranda themselves are consistent with this evidentiary principle. The examiner memorandum states that objective evidence should be supported by actual proof and that legal conclusions on eligibility are not entitled to weight. SMED Examiner Memo, at 3. Thus, a declaration that merely states that the claims “improve computer functionality,” “are not abstract,” “cannot be performed in the human mind,” or “recite an unconventional arrangement” is vulnerable. Those assertions identify the desired conclusion, but they do not prove the underlying facts.

A persuasive SMED should instead identify the specific claim limitations at issue, tie them to the specification, explain the state of the art at filing, and provide objective support for the alleged improvement. In software and AI cases, that support may include comparative benchmarks, architecture diagrams, training results, model-performance data, source-code excerpts, design documents, system logs, or contemporaneous technical records. The declaration should function as a foundation for evidence, not as a substitute for evidence.

The POSITA’s Role Across the Alice Framework
At Alice step one, the POSITA may be relevant when the issue is whether the claims are directed to a specific technological improvement rather than an abstract result. Enfish, McRO, SRI, CardioNet, and Desjardins are the principal examples. Enfish, 822 F.3d at 1335–39; McRO, 837 F.3d at 1313–16; SRI, 930 F.3d at 1303–04; CardioNet, 955 F.3d at 1368–71; Desjardins, slip op. at 7–10. A declaration may help explain why a claimed data structure, animation rule set, distributed network-monitoring arrangement, cardiac-monitoring technique, or machine-learning training architecture changes technology rather than merely applying an abstract idea.

At Alice step two, the POSITA’s role is more direct because Berkheimer makes conventionality a factual issue. Berkheimer, 881 F.3d at 1368–70. A SMED can be particularly useful where an examiner asserts that additional elements are well-understood, routine, and conventional without adequate evidentiary support. BASCOM supplies the doctrinal hook: even if individual elements are conventional, their ordered arrangement may be nonconventional and may supply an inventive concept. BASCOM, 827 F.3d at 1349–52.

Under the USPTO’s practical-application framework, the POSITA may also help show integration of a judicial exception into a practical application. The examiner memorandum provides examples involving mental-process rejections, computer-functionality improvements, neural-network performance improvements, computer animation, treatment or prophylaxis, and unconventional arrangements. SMED Examiner Memo, at 4–7. But this is also where the divergence between USPTO guidance and Federal Circuit law becomes most visible. The USPTO’s examination guidance is not binding on courts. Rudy, 956 F.3d at 1382–83; ENCO, 845 F. App’x at 957–58.

That creates a prosecution-litigation gap. For examiners, a SMED may persuade under MPEP § 2106, practical-application analysis, and the Office’s internal guidance. For courts, the same record may help only if it creates a material factual dispute, informs claim construction, or supports a claim-tethered technological improvement under Federal Circuit doctrine. A declaration that works before an examiner because it supplies practical context may be much less useful in district court if the claim language remains functional, result-oriented, or generic.

Practical Implications for Patent Prosecutors
The first practical implication is that the technological-improvement narrative should be built into the application at filing. The SMED guidance permits a declaration to explain what a POSITA would understand, but it does not allow a declaration to add missing disclosure. SMED Examiner Memo, at 3. The specification should identify the technical problem, describe the technical mechanism, and explain the technical effect.

Second, the claims should recite the improvement. The April 2026 memorandum’s statement that claims define the invention should be treated as the drafting discipline that governs SMED practice. April 2026 SMED Memo, at 3. If the claim merely states a desired result, a declaration explaining how an embodiment works may not solve the § 101 problem.

Third, the SMED should prove facts rather than argue law. It should not merely recite that the claims are eligible, technical, unconventional, or not mental. It should identify claim limitations, connect them to the specification, describe the relevant state of the art, and supply objective evidence showing the technological improvement or nonconventional arrangement.

Fourth, the SMED should be separate from § 103 and § 112 declarations. Both the December 2025 public memorandum and the April 2026 memorandum recommend separate declarations to preserve record clarity and avoid confusing eligibility evidence with obviousness or disclosure evidence. December 2025 Public SMED Memo, at 1–4; April 2026 SMED Memo, at 3–6.

Fifth, inventor testimony should be corroborated when possible. The corroboration cases do not create a formal rule that inventor-signed SMEDs are insufficient. But they warn against uncorroborated, self-interested testimony offered to establish historical technical facts. The Barbed Wire Patent, 143 U.S. at 284–85; Price, 988 F.2d at 1193–95; Singh, 222 F.3d at 1367; EmeraChem, 859 F.3d at 1345–48. The strongest SMED will generally be supported by objective records or independent testimony.

Finally, a SMED should be drafted not merely to obtain allowance, but to survive a later Rule 12, summary-judgment, or post-grant eligibility challenge. The declaration should identify which factual propositions would matter under Federal Circuit doctrine, not merely under the Office’s practical-application rubric. For example, the declaration should make clear whether it is supporting claim construction, rebutting conventionality, showing that a limitation cannot practically be performed in the human mind, proving a specific technological improvement, or demonstrating that an ordered combination was unconventional. That discipline makes the SMED useful both to the examiner and to a later tribunal applying Federal Circuit law.

Conclusion
The POSITA has an important but bounded role in subject matter eligibility. The skilled artisan may explain what the filed specification would have conveyed, describe the state of the art, rebut unsupported assertions of conventionality, and show why claimed limitations reflect a technological improvement or practical application. The USPTO’s SMED memoranda properly recognize that eligibility often depends on technical facts and that Rule 132 declarations can help develop those facts during prosecution.

But Federal Circuit law prevents the POSITA from becoming a substitute for the claims. The claims must recite the technological improvement or unconventional arrangement. The specification must support that improvement as of filing. The declaration must have a nexus to the claimed invention. And conclusory or uncorroborated testimony, particularly from an interested inventor, may receive little weight.

The tension between USPTO guidance and Federal Circuit law is therefore real, but manageable. The USPTO invites applicants to use POSITA evidence to demonstrate practical technological application. Federal Circuit law permits such evidence only insofar as it illuminates what the claims and specification already establish. Desjardins shows the Office’s preferred path for AI claims that improve the model or computational architecture itself. Recentive shows the Federal Circuit’s limiting rule for claims that merely apply generic machine learning to a new field of use. The best SMED practice is not to ask the POSITA to rescue eligibility. It is to use the POSITA to prove, with objective and claim-tethered evidence, that the claimed invention was technological from the beginning.