The Federal Circuit’s “Constellation” Prize

Last year, in response to Recentive Analytics, Inc. v. Fox Corp. [1], we argued that the preemption doctrine should be reinstated as the central inquiry in any 101 subject matter eligibility determination. [2] In Recentive, the Federal Circuit provided its first analysis of claims directed to machine learning models: “[t]oday, we hold only that patents that do no more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied, are patent ineligible under § 101.” In its petition for Writ of Certiorari to appeal the decision, [3] Recentive warned that “[d]eparting from preemption has its consequences. The Federal Circuit’s patent-eligibility test no longer provides clear guidance to parties on whether a patent claim is directed to an ‘abstract idea’ and instead has led to unpredictable and inconsistent results. This has allowed courts to invalidate key technological innovations that do not threaten to preempt fundamental scientific tools.” [4] As expected, the Supreme Court refused to provide clarity to continued inconsistencies in the legal landscape for patent eligibility.

Although the Supreme Court turned a deaf ear (Justice is blind anyway) to Recentive’s preemption plea, it appears that the Federal Circuit may have taken that valid criticism to heart. In Constellation Designs, LLC v. LG Electronics Inc., [5] the dispute involved patents for “communication systems that use non-uniform constellations that have increased capacity compared to conventional, uniform constellations operating within a similar signal-to-noise ratio band.” [6] The District Court had found LG to willfully infringe several claims, including claim 17 of U.S. Patent No. 8,842,761 and claim 21 of U.S. Patent No. 11,019,509. [7] LG challenged the eligibility of these claims, and the Federal Circuit found claim 17 of the ‘761 patent to be ineligible and claim 21 of the ‘509 patent to be eligible. Since both claims were designated as representative claims, the court’s ruling applied to the entire set of claims each represented.

In finding claim 17 of the ‘761 patent ineligible, the Federal Circuit relied on the Supreme Court’s reasoning in O’Reilly v. Morse [8] (and, yes, that is Samuel Morse of Morse code fame) and reasoned that “the breadth of claim 17 demonstrates that the claim is directed to the abstract idea of ‘optimizing’ a constellation for PD capacity. … (‘[P]reemption may signal patent ineligible subject matter.’). While claim 17 does not recite every way to optimize a constellation, it does recite every way to optimize a constellation for PD capacity. The claim is written so broadly that it covers the mathematical concept of optimizing a constellation for PD capacity to accomplish improved capacity.” [9] The Federal Circuit used the preemption doctrine to anchor its analysis in both steps of the Alice/Mayo framework: (1) at step 1, the court held that claiming an “optimized” result without the “how” effectively preempts the entire field of optimizing a constellation for capacity, and (2) at step 2, it found no “inventive concept” because the alleged innovation, the optimal constellation, was indistinguishable from the abstract idea being preempted.

The Court found claim 21 of the ‘509 patent eligible by finding that claim 21 “covers a concrete implementation of a technological improvement to solve a technological problem” [10] and cited “better coding gains,” and a “counterintuitive conclusion.” The Court again used the preemption doctrine: “[r]ather than broadly claiming all means of ‘optimizing,’ claim 21 is an application of the described process with parameters representing the described invention, including that ‘the plurality of unique point locations are unequally spaced,’ ‘each have a location and a different label,’ and ‘the locations of at least two of the constellation points are the same.’” [11]

If this is a shift toward bringing back the preemption doctrine as the centerpiece of patent eligibility determination, Constellation is a breath of fresh air. This case also provides a “navigation chart” for surviving Section 101:

1. Escape the generic label: In Recentive, the claims were ineligible because they were seen as using generic ML in a particular environment. Although Constellation does not relate to ML, it suggests that claims directed to a machine learning model can be eligible by claiming how training datasets, loss functions, and/or mathematical parameters customize the model.
   
2. The specification is not a lifeboat: In Constellation, the Federal Circuit clarified that the specification may include enough detail to describe the invention, but the § 101 inquiry must focus on the language of the claims. [12] As a result, even if a disclosure describes a large language model (LLM) fine-tuned via smart prompting, or a particular training dataset that optimizes the model, if the claim only recites an optimized output without significantly more, it may be deemed ineligible.
3. Define the practical application: To overcome the 101 threshold, innovation must integrate models into practical applications. Constellation teaches that claims directed to solving a particular technological problem (like overcoming capacity constraints to improve coding gains) using a particular technological solution (like specific, non-uniform constellations with overlapping constellation point locations) would be eligible under the preemption doctrine.

The Constellation ruling is a positive for inventors in the software space trying to avoid 101 issues as long as the applications are crafted carefully in line with the “navigation chart” laid out above. We hope that the Federal Circuit will continue down this positive path and use the preemption doctrine as a central inquiry for 101 subject matter eligibility.

[1] Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025), cert. denied, No. 25-505 (U.S. Dec. 8, 2025). (Available at: www.cafc.uscourts.gov/opinions-orders/23-2437.OPINION.4-18-2025_2500790.pdf)

[2] Manav Das, “The Preemption Doctrine: A Necessary Course Correction After Recentive v. Fox,” July 9, 2025. (Avaliable at: https://patentdocs.org/2025/07/09/the-preemption-doctrine-a-necessary-course-correction-after-recentive-v-fox/)

[3] Petition for Writ of Certiorari, Recentive Analytics, Inc. v. Fox Corp., No. 25-505 (U.S. Oct. 21, 2025). (Available at: www.supremecourt.gov/DocketPDF/25/25-505/380165/20251021131338718_25-___PetitionForAWritOfCertiorari.pdf)

[4] Id. at 18.

[5] Constellation Designs, LLC v. LG Electronics Inc., No. 2024-1822 (Fed. Cir. Apr. 28, 2026) (Available at: www.cafc.uscourts.gov/opinions-orders/24-1822.OPINION.4-28-2026_2683894.pdf)

[6] Id. at 2

[7] Id.

[8] Id. at 17-18. (In Morse, the Supreme Court “upheld claims related to the details of Samuel Morse’s invention of the electromagnetic telegraph, but invalidated a claim for the use of ‘electromagnetism, however developed for marking or printing intelligible characters, signs, or letters, at any distances.’” ChargePoint, 920 F.3d at 769 (quoting Morse, 56 U.S. at 112). “The Court [in Morse] expressed concern that such a broad claim would cover any application of printing at a distance via electromagnetism regardless of whether those applications used the invention in the patent.” Id. (citing Morse, 56 U.S. at 113). Morse’s invalidated claim “encompassed all solutions for achieving a desired result” and was “drafted in such a result-oriented way that [it] amounted to encompassing the ‘principle in the abstract’ no matter how implemented.” Interval Licensing, 896 F.3d at 1343).

[9] Id. at 18.

[10] Id at 23.

[11] Id. at 24 (citing ‘509 patent col. 19 11. 35–47).

[12] Id. at 20 (quoting Synopsys, Inc. v. Mentor Graphics Corp., 839 F.3d 1138, 1149 (Fed. Cir. 2016)).