United States v. Curtis Jenkins

United States Court of Appeals
FOR THE DISTRICT OF COLUMBIA CIRCUIT

Argued May 11, 2022 Decided October 11, 2022

No. 21-3089

UNITED STATES OF AMERICA,
APPELLEE

v.

CURTIS JENKINS,
APPELLANT

Appeal from the United States District Court
for the District of Columbia
(No. 1:16-cr-00213-1)

Celia Goetzl, Assistant Federal Public Defender, argued
the cause for appellant. With her on the appellant’s
Memorandum of Law and Fact was A.J. Kramer, Federal
Public Defender.

Kevin Birney, Assistant U.S. Attorney, argued the cause
for appellee. With him on the appellee’s Memorandum of Law
and Fact were Chrisellen R. Kolb and Nicholas P. Coleman,
Assistant U.S. Attorneys.

Before: KATSAS and WALKER, Circuit Judges, and
GINSBURG, Senior Circuit Judge.
 2
Opinion for the Court filed by Circuit Judge KATSAS.

Opinion concurring in part, dissenting in part, and
concurring in the judgment filed by Senior Circuit Judge
GINSBURG.

KATSAS, Circuit Judge: A district court may grant an
inmate compassionate release only for “extraordinary and
compelling reasons.” 18 U.S.C. § 3582(c)(1)(A). Typically,
such reasons involve personal considerations such as the
inmate’s health, age, or family circumstances. Section 1B1.13
of the Sentencing Guidelines, which governs motions for
compassionate release filed by the Bureau of Prisons, addresses
when these considerations become sufficiently “extraordinary
and compelling” to warrant compassionate release.

In United States v. Long, 997 F.3d 342 (D.C. Cir. 2021),
we held that section 1B1.13 does not govern motions for
compassionate release filed by the inmate himself. This case
presents the question whether the district courts, in considering
such motions, may nonetheless rely on section 1B1.13 and its
commentary as persuasive authority. Following the view of
nine sister circuits, we hold that they may.

This case also presents the question whether certain
intervening legal changes, occurring after the sentence at issue
was imposed, can support compassionate release. One is a
statute that only prospectively reduces penalties for the
defendant’s offense. Another is a judicial decision that
retroactively establishes legal error at sentencing. A third is a
judicial decision that, if rendered earlier, might have affected
the negotiation of a plea bargain by reducing the defendant’s
exposure. We hold that none of these changes in sentencing
law can support the grant of compassionate release.
 3
I

A

Under the former regime of indeterminate sentencing, a
prisoner typically became eligible for release on parole after
serving the earlier of one third of his sentence or 10 years. 18
U.S.C. § 4205(a) (governing inmates sentenced before 1987).
Executive officials had broad discretion to grant or deny parole.
Id. § 4206(a). In addition, the Bureau of Prisons could at any
time and for any reason move in court to reduce the prisoner’s
sentence to time served. Id. § 4205(g).

The Sentencing Reform Act of 1984 abandoned
indeterminate sentencing. Pub. L. No. 98-473, tit. II, ch. II, 98
Stat. 1987. The SRA abolished parole, including release under
section 4205(g). Id. § 218(a)(5), 98 Stat. at 2027. It prohibits
modification of a sentence except in three defined
circumstances. First, as under the old section 4205(g), the
court may reduce a sentence upon motion of the Bureau of
Prisons. 18 U.S.C. § 3582(c)(1)(A). This avenue is widely
known as compassionate release. Second, the court may
modify a sentence “to the extent otherwise expressly permitted
by statute or by Rule 35 of the Federal Rules of Criminal
Procedure.” Id. § 3582(c)(1)(B). Third, the court may
resentence a defendant if the Sentencing Commission has
lowered the applicable guideline range and made the change
retroactive. Id. § 3582(c)(2).

The SRA imposed three limits on compassionate release
that were not present in the old section 4205(g). First, a district
court may grant the Bureau’s motion only if it finds that
“extraordinary and compelling reasons” warrant early release.
 4
18 U.S.C. § 3582(c)(1)(A)(i).1 Second, the reduced sentence
must be “consistent with applicable policy statements issued by
the Sentencing Commission.” 18 U.S.C. § 3582(c)(1)(A).
Finally, the court may grant relief only “after considering the
factors set forth in section 3553(a) to the extent that they are
applicable.” Id.

The First Step Act of 2018 created a new procedural
pathway to compassionate release. As before, a defendant
seeking compassionate release first must ask the BOP to file a
motion on his behalf. Now, if the Bureau refuses, the defendant
may file the motion on his own behalf. Pub. L. No. 115-391,
§ 602(b)(1), 132 Stat. 5194, 5239 (codified at 18 U.S.C.
§ 3582(c)(1)(A)). The First Step Act did not alter the other
statutory restrictions on compassionate release, including the
requirement of “extraordinary and compelling reasons.”

The Sentencing Commission has issued a policy
statement about compassionate release, U.S.S.G. § 1B1.13,
which it last amended before the passage of the First Step Act.
The statement provides that a court may grant compassionate
release “[u]pon motion of the Director of the Bureau of
Prisons” if “extraordinary and compelling reasons warrant the
reduction” and the inmate poses no danger to the safety of any
other person.

Application note 1 of the commentary on section 1B1.13
lists three kinds of reasons that the Commission deems
extraordinary and compelling: health, if the inmate suffers
from a terminal illness or other serious condition from which

1
Congress later permitted release under section 3582(c)(1)(A)
for certain elderly defendants sentenced to life imprisonment.
Violent Crime Control and Law Enforcement Act of 1994, Pub. L.
No. 103-322, § 70002(5), 108 Stat. 1796, 1985 (codified at 18 U.S.C.
§ 3582(c)(1)(A)(ii)). That provision is not at issue here.
 5
he is not expected to recover; age, if the inmate is over 65, has
seriously deteriorating health, and has served 10 years or 75%
of his sentence; and family circumstances, if the caregiver of
the inmate’s minor child dies or becomes incapacitated or if the
inmate’s spouse becomes incapacitated and has no other
caregiver. U.S.S.G. § 1B1.13 cmt. n.1(A)–(C). These
categories are not exclusive. A reduction may also be
warranted if, “[a]s determined by the Director of the Bureau of
Prisons, there exists in the defendant’s case an extraordinary
and compelling reason other than, or in combination with,” the
listed reasons. Id. cmt. n.1(D).

Section 1B1.13 binds courts in cases where it is
“applicable.” 18 U.S.C. § 3582(c)(1)(A). In United States v.
Long, 997 F.3d 342 (D.C. Cir. 2021), we held that section
1B1.13 “is not ‘applicable’ to defendant-filed motions for
compassionate release under the First Step Act,” because it
“applies only to motions … filed by the Bureau of Prisons.” Id.
at 355. The Sentencing Commission lacked a quorum from
January 2019 until August 2022, so it has been unable to
promulgate a policy statement that applies to such motions.

B

In November 2016, police conducted a traffic stop of
Curtis Jenkins’ car. They found a stolen handgun, 11 grams of
crack cocaine, and roughly $2,500 in cash. A federal grand
jury charged Jenkins with one count of possessing a firearm as
a felon, 18 U.S.C. § 922(g); one count of possessing cocaine
base with intent to distribute, 21 U.S.C. § 841(a)(1) &
(b)(1)(C); and one count of using a firearm during and in
relation to a drug trafficking crime, 18 U.S.C. § 924(c)(1)(A).
While Jenkins was on release pending trial, he fled after police
tried to stop his car a second time. When the police eventually
apprehended him, they discovered another gun, 3.5 grams of
 6
crack cocaine, and about $1,500 in cash. The grand jury once
again charged Jenkins with the same three offenses, for a total
of six counts.

At the time, Jenkins faced long mandatory minimum
sentences on the various firearms charges. A section 924(c)
offense carries a mandatory minimum sentence of five years’
imprisonment, which must run consecutively with any other
sentence. 18 U.S.C. § 924(c)(1)(A)(i), (D)(ii) (2018). And
section 924(c)’s stacking provision set a mandatory minimum
sentence for a second section 924(c) offense at 25 years, even
if the defendant had not yet been convicted of the first offense
at the time of the second. Id. § 924(c)(1)(C)(i) (2012).
Additionally, the Armed Career Criminal Act (ACCA)
imposes a 15-year mandatory minimum sentence for violating
section 922(g) if the defendant has three prior convictions for
a “violent felony” or “serious drug offense.” Id. § 924(e)(1)
(2018). Jenkins had two prior convictions for attempted
distribution of cocaine and one for assault with a dangerous
weapon under D.C. law. At the time, all three convictions
arguably qualified as predicate offenses to trigger ACCA.
Jenkins thus faced the substantial possibility of a 45-year
mandatory minimum sentence, with additional prison time for
the two drug charges.

Jenkins and the government entered a plea agreement.
Jenkins agreed to plead guilty to one section 924(c) charge and
one cocaine possession charge in exchange for dismissal of the
remaining four charges. Given Jenkins’ prior convictions, the
parties agreed that the career offender sentencing guideline,
U.S.S.G. § 4B1.1(a), applied. That gave Jenkins a guideline
range of about 22 to 27 years. Nonetheless, the parties agreed
to recommend a prison term of eight to 12 years. The district
court sentenced Jenkins to eight years. As part of the plea deal,
Jenkins waived any right to challenge the sentence on direct
 7
appeal or by motion under 28 U.S.C. § 2255, except to the
extent such a motion was based on newly discovered evidence
or a claim of ineffective assistance of counsel.

Three relevant legal developments then took place. First,
Congress prospectively narrowed section 924(c)’s stacking
provision. Now, its enhanced minimum sentence applies only
to offenses committed “after a prior conviction” under section
924(c). First Step Act § 403(a), 132 Stat. at 5221–22 (codified
at 18 U.S.C. § 924(c)(1)(C)) (emphasis added). But the
amendment applies only to defendants who had not yet been
sentenced at the time of its enactment. Id. § 403(b), 132 Stat.
at 5222. Had the change been in place when Jenkins was
charged, he would have faced a 10-year rather than a 30-year
minimum for the two section 924(c) charges. Second, United
States v. Winstead, 890 F.3d 1082 (D.C. Cir. 2018), held that
attempted drug offenses do not trigger the career offender
guideline. Id. at 1091. Without his prior convictions for
attempted distribution, Jenkins would not have been
considered a career offender, and his guideline range would
have been seven to seven and a half years—well below the
guideline range stipulated by the parties, and slightly below the
sentence ultimately imposed. Third, Borden v. United States,
141 S. Ct. 1817 (2021), held that offenses with a minimum
mens rea of recklessness are not “crimes of violence” under
ACCA. Id. at 1821–22 (plurality opinion); id. at 1835
(Thomas, J., concurring in the judgment). Under D.C. law,
recklessness is enough to support a conviction for assault with
a dangerous weapon. Frye v. United States, 926 A.2d 1085,
1097 (D.C. 2005). Without counting his conviction for that
offense, Jenkins would not have faced a 15-year minimum
sentence under ACCA.

Jenkins filed a motion for compassionate release. He
argued that the narrowed stacking provision, the commission
 8
of a Winstead error to trigger the career offender guideline, and
the pre-Borden threat of a 15-year minimum sentence under
ACCA were extraordinary and compelling circumstances
warranting early release. Jenkins also claimed that he was at
heightened risk of COVID-19 due to his borderline obesity,
that his elderly mother needed him as a caregiver, and that his
adult daughters needed him to be present in their lives.

The district court denied the motion. The court
acknowledged that section 1B1.13 “is not binding or
exhaustive” but still found it “useful” as “general guidance.”
ECF Doc. 42, at 8. The court then considered whether Jenkins
had “presented information that is similar, or is at least roughly
comparable to,” the reasons that section 1B1.13 and its
commentary identify as extraordinary and compelling. Id. at 9.
The court further held that the new statute, Winstead, and
Borden were irrelevant because the compassionate-release
statute does not permit courts to reexamine the lawfulness or
fairness of a sentence as originally imposed. Id. at 9–10. That
left only Jenkins’ health and family circumstances which, the
court concluded, did not themselves amount to extraordinary
and compelling reasons for early release. Id. at 11–13.

II

We review the denial of a motion for compassionate
release for abuse of discretion. Long, 997 F.3d at 352. An error
of law is necessarily an abuse of discretion. Koon v. United
States, 518 U.S. 81, 100 (1996). Here, the requirement of
“extraordinary and compelling reasons” establishes the
governing legal standard. Defining its scope—i.e.,
determining what kind of reasons can be “extraordinary and
compelling”—is a legal question that we review de novo. U.S.
Bank Nat. Ass’n v. Vill. at Lakeridge, LLC, 138 S. Ct. 960, 965
(2018). Whether the facts of a particular case satisfy the
 9
standard is a mixed question that we review deferentially. See
id. at 966–67.

III

Jenkins first takes aim at the district court’s reliance on
section 1B1.13 and its commentary. He claims the court
treated them as binding despite Long’s holding that section
1B1.13 does not apply to defendant-filed motions.
Alternatively, he argues that it was error to rely on the
administrative materials even as persuasive authority.

A

The district court did not consider itself bound by section
1B1.13. Citing Long, it stated that section 1B1.13 “is not
binding or exhaustive.” ECF Doc. 42, at 8. To be sure, the
court did frame its discussion in terms of whether Jenkins had
presented reasons comparable to the ones listed in section
1B1.13 and its commentary. But courts do not treat Sentencing
Commission guidance as controlling when they rely on it only
as the “starting point and … initial benchmark” of their
analysis. Gall v. United States, 552 U.S. 38, 49 (2007).

Jenkins cites United States v. Johnson, 858 F. App’x 381
(D.C. Cir. 2021), which vacated the denial of a defendant-filed
motion for compassionate release. There, the district court
asserted that extraordinary and compelling reasons were
“circumstances defined in the Guidelines.” Id. at 383 (cleaned
up). We held that the court had erroneously “consider[ed] itself
bound by the policy statement.” Id. In contrast, the district
court here was explicit that section 1B1.13 did not control its
decision.
 10
B

The district court appropriately considered the Sentencing
Commission’s guidance. Although not formally controlling,
the guidance interprets a statutory requirement common to both
BOP-filed and defendant-filed motions. Courts may consider
nonbinding agency interpretations of statutes to be persuasive
authority. Skidmore v. Swift & Co., 323 U.S. 134, 139–40
(1944). The level of weight such an interpretation carries
“var[ies] with circumstances,” including “the degree of the
agency’s care, its consistency, formality, and relative
expertness.” United States v. Mead Corp., 533 U.S. 218, 228
(2001) (cleaned up). Given these considerations, the district
court reasonably invoked the guidance.

To begin, the guidance is formal. The Sentencing
Commission has published it in the Guidelines Manual and
framed it as general guidance of nationwide scope. The
guidance also reflects an exercise of the Commission’s
rulemaking authority, which formally binds courts in cases
where it applies. 18 U.S.C. § 3582(c)(1)(A); 28 U.S.C.
§ 994(t). Likewise, the commentary is binding to the extent
that it reasonably interprets the policy statement. Stinson v.
United States, 508 U.S. 36, 44–45 (1993).

The guidance also falls within the Sentencing
Commission’s expertise. Congress has charged the
Commission with setting forth “what should be considered
extraordinary and compelling reasons” for compassionate
release. 28 U.S.C. § 994(t). And the guidance reflects a careful
exercise of that responsibility. The Commission adopted the
current version of application note 1 after “an in-depth review”
that included BOP data on compassionate-release applications,
two reports from the Department of Justice Office of Inspector
General, and a public hearing. U.S. Sentencing Comm’n,
 11
Guidelines Manual: Supplement to Appendix C, amend. 799, at
127 (2021). The Commission also explained in detail how the
facts it found led to its specific guidance. Id. at 127–28.

Finally, the Sentencing Commission’s guidance has been
consistent over time. The Commission first construed the
“extraordinary and compelling reasons” language in 2007.
Then, it listed three kinds of qualifying reasons—health, age,
and family circumstances—and acknowledged that other
reasons may also suffice. 3 U.S. Sentencing Comm’n,
Guidelines Manual: Appendix C, amend. 698, at 186 (2011).
Application note 1 retains this basic structure. The only change
has been to elaborate further on each category.

Given these considerations, the district court permissibly
relied on section 1B1.13 and its commentary to inform the
exercise of its discretion. As nine other circuits have held,
courts may consider these materials even in cases involving
defendant-filed motions for compassionate release. E.g.,
United States v. Ruvalcaba, 26 F.4th 14, 23 (1st Cir. 2022);
United States v. Andrews, 12 F.4th 255, 260 (3d Cir. 2021);
United States v. McCoy, 981 F.3d 271, 282 n.7 (4th Cir. 2020);
United States v. Thompson, 984 F.3d 431, 433 (5th Cir. 2021);
United States v. Tomes, 990 F.3d 500, 503 n.1 (6th Cir. 2021);
United States v. Gunn, 980 F.3d 1178, 1180 (7th Cir. 2020);
United States v. Marcussen, 15 F.4th 855, 859 (8th Cir. 2021);
United States v. Aruda, 993 F.3d 797, 802 (9th Cir. 2021);
United States v. Hald, 8 F.4th 932, 938 n.4 (10th Cir. 2021).

Jenkins’ contrary arguments are unconvincing. Citing
Long, Jenkins first contends that the guidance is unpersuasive
because it “does not reflect any … policy judgment … about how
compassionate release decisions should be made under the First
Step Act,” which “deliberately broadened its availability.” 997
F.3d at 359. But the Act’s sole change was to create a new
procedural avenue for relief. It did not alter the extraordinary-
 12
and-compelling-reasons standard, so it did not undermine the
Commission’s interpretation of that standard. And in Long, we
held only that section 1B1.13 is not formally “applicable” to
defendant-filed motions, so its guidance is not legally binding.
Id. at 347. We had no occasion to consider, and we did not
consider, the distinct question whether courts may rely on it as
persuasive authority.

Jenkins next claims that application note 1 unreasonably
limits “extraordinary and compelling reasons” to specific
circumstances related to the defendant’s health, age, and family
situation. See U.S.S.G. § 1B1.13 cmt. n.1(A)–(C). Jenkins
overlooks the note’s catchall provision, which states that a
“reason other than, or in combination with,” the listed
circumstances may be extraordinary and compelling. Id. cmt.
n.1(D). Admittedly, paragraph (D) also requires that any
unenumerated reason be raised by the BOP. Id. But the district
court properly looked beyond that requirement in the context
of a defendant-filed motion, instead asking whether Jenkins
had presented any reasons “at least roughly comparable” to the
ones enumerated in the guidance. ECF Doc. 42, at 9.

Finally, Jenkins argues that courts should give no weight
to application note 1 because it interprets a policy statement
that merely parrots the governing statute. As a general matter,
we defer to the Sentencing Commission’s official commentary
because it is analogous to an agency’s interpretation of its own
regulations. Stinson, 508 U.S. at 44–45; see Kisor v. Wilkie, 139
S. Ct. 2400 (2019); Auer v. Robbins, 519 U.S. 452 (1997). But
we do not extend Auer deference when the regulation being
interpreted “does little more than restate the terms of the statute
itself.” Gonzales v. Oregon, 546 U.S. 243, 257 (2006). Here,
section 1B1.13(1)(A) merely repeats verbatim the statutory
requirement of “extraordinary and compelling reasons.” Thus,
 13
Jenkins maintains, under the analogy of policy statements to
regulations, application note 1 is entitled to no deference.

We need not decide whether Jenkins is correct on this
point. Even when the parroting exception to Auer applies, an
agency’s interpretation of language common to the statute and
regulation still may receive Skidmore deference. Gonzales,
546 U.S. at 268–69. For Skidmore purposes, the only
difference between section 1B1.13 itself and application note 1
is that the policy statement is somewhat more formal. This
does not undermine our overall conclusion that the district
courts may permissibly rely on the Commission’s
interpretation as persuasive, for all the reasons noted above.

IV

We next consider Jenkins’ arguments based on the
stacking amendment, Winstead, and Borden. The district court
held that such intervening changes in sentencing law cannot
warrant compassionate release. Jenkins objects that this
holding is atextual. Congress has directed that “[r]ehabilitation
of the defendant alone” may “not be considered an
extraordinary and compelling reason.” 28 U.S.C. § 994(t). But
otherwise, Jenkins contends, Congress has placed no limits on
what reasons a court may consider in deciding whether to grant
compassionate release.

We disagree, for the proffered reasons must be
“extraordinary and compelling.” 18 U.S.C. § 3582(c)(1)(A).
An “extraordinary” reason must be “‘most unusual,’ ‘far from
common,’ and ‘having little or no precedent.’” United States
v. Hunter, 12 F.4th 555, 562 (6th Cir. 2021) (quoting Webster’s
Third New International Dictionary: Unabridged 807 (1971)).
And a “compelling” reason must be “both powerful and
convincing.” United States v. Canales-Ramos, 19 F.4th 561,
567 (1st Cir. 2021) (citing Webster’s Third, supra, at 462).
 14
Applying these requirements, we agree with the district court
that Jenkins’ three asserted grounds are neither extraordinary
nor compelling, whether considered in isolation or in
combination with other factors.

A

Start with the stacking provision. The First Step Act
eliminated the 25-year minimum sentence for a second section
924(c) offense if the defendant had not yet been convicted of
the first offense at the time of the second. Pub. L. No. 115-391,
§ 403(a), 132 Stat. at 5221–22. Although the amendment
applies only prospectively to defendants sentenced after the
date of its enactment, id. § 403(b), 132 Stat. at 5222, Jenkins
claims that courts may nonetheless consider it in deciding
whether to grant compassionate release to defendants
previously sentenced under the unamended provision.

The circuits have split on whether courts may consider
such intervening but expressly nonretroactive sentencing
statutes. Three circuits have held that such statutes may neither
support nor contribute to a finding that extraordinary and
compelling reasons warrant compassionate release. United
States v. Crandall, 25 F.4th 582, 586 (8th Cir. 2022); Andrews,
12 F.4th at 261 (3d Cir.); United States v. Thacker, 4 F.4th 569,
574 (7th Cir. 2021). Four circuits have held that such statutes
may be considered in connection with other factors. United
States v. Chen, No. 20-50333, 2022 WL 4231313, at *5 (9th
Cir. Sept. 14, 2022); Ruvalcaba, 26 F.4th at 28 (1st Cir.);
United States v. McGee, 992 F.3d 1035, 1047–48 (10th Cir.
2021); McCoy, 981 F.3d at 286 (4th Cir.); The Sixth Circuit is
internally split on the issue. Compare United States v.
McKinnie, 24 F.4th 583, 588 (6th Cir. 2022), with United States
v. McCall, 20 F.4th 1108, 1114 (6th Cir. 2021), reh’g en banc
granted, opinion vacated, 29 F.4th 816 (6th Cir. 2022) (mem.).
 15
We agree with the Third, Seventh, and Eighth Circuits. To
begin, there is nothing remotely extraordinary about statutes
applying only prospectively. In fact, there is a strong
presumption against statutory retroactivity, which is “deeply
rooted in our jurisprudence” and “embodies a legal doctrine
older than our Republic.” Landgraf v. USI Film Prods., 511
U.S. 244, 265 (1994). The Ex Post Facto Clause compels the
purely prospective application of criminal statutes that are
unfavorable to defendants. U.S. Const. art. 1, § 9, cl. 3. And
since 1871, federal law has codified the presumption against
retroactivity for statutes making criminal law more favorable
to defendants. Thus, a statute reducing the penalties for a
criminal offense—or even repealing the offense entirely—does
not apply to offenses committed prior to its enactment, “unless
the repealing Act shall so expressly provide.” 1 U.S.C. § 109;
see Dorsey v. United States, 567 U.S. 260, 273–75 (2012). The
SRA imposes its own prospectivity rule for sentencing
guidelines, by instructing courts to consider the guidelines “in
effect on the date the defendant is sentenced.” 18 U.S.C.
§ 3553(a)(4)(A)(ii). Thus, “in federal sentencing the ordinary
practice is to apply new [lower] penalties to defendants not yet
sentenced, while withholding that change from defendants
already sentenced.” Dorsey, 567 U.S. at 280. And what “the
Supreme Court views as the ‘ordinary practice’ cannot also be
an ‘extraordinary … reason’ to deviate from that practice.”
United States v. Wills, 997 F.3d 685, 688 (6th Cir. 2021). Nor
can such a settled, common practice be a compelling reason to
deviate.

Separation-of-powers considerations reinforce this
analysis. “It is the legislature, not the Court, which is to define
a crime, and ordain its punishment.” United States v.
Wiltberger, 18 U.S. (5 Wheat.) 76, 95 (1820). When Congress
enacted the original stacking provision, it deemed a 25-year
minimum sentence to be appropriate for all second section
 16
924(c) offenses. 18 U.S.C. § 924(c)(1)(C)(i) (2012). And by
making its ameliorative amendment expressly nonretroactive,
First Step Act § 403(b), 132 Stat. at 5222, Congress reaffirmed
that the 25-year minimum remained appropriate for defendants
already sentenced. We would usurp these quintessentially
legislative judgments if we used compassionate release as a
vehicle for applying the amendment retroactively, to
previously sentenced defendants who would not otherwise
qualify for compassionate release. As the Seventh Circuit
explained, we cannot allow the “general” and “discretionary”
authority conferred by the compassionate-release statute “to
upend the clear and precise limitation Congress imposed on the
effective date of the First Step Act’s amendment to § 924(c).”
Thacker, 4 F.4th at 574. Here as elsewhere, courts must
exercise their discretion consistent with all applicable statutory
constraints. See, e.g., Martin v. Franklin Cap. Corp., 546 U.S.
132, 139 (2005) (“Discretion is not whim.”); INS v. Pangilinan,
486 U.S. 875, 883 (1988) (“courts of equity can no more
disregard statutory and constitutional requirements … than can
courts of law” (cleaned up)).

Courts reaching the contrary conclusion have reasoned
that there is a significant difference between making an
intervening statute “categorically retroactive”—which all
agree Congress did not do here—and considering a purely
prospective statute as one relevant factor in an “individualized”
compassionate-release inquiry. McCoy, 981 F.3d at 286.
Making the statute retroactive would require “automatic
vacatur” of all past sentences to which the statute applied. Id.
In contrast, considering an intervening statute would merely
“allow[] for the provision of individual relief in the most
grievous cases.” Id. at 287 (cleaned up). For this reason, the
concurrence concludes that we have committed a “Fallacy of
Division” by “assuming what is true of the whole must be true
of each part.” Post at 2.
 17

The difficulty with this view is that the disputed
considerations do not vary from case to case—one might say
that they are not divisible. The Fourth Circuit would have
courts consider two assertedly “distinct features” of cases
involving defendants sentenced under the unamended stacking
provision—the “sheer and unusual length” of the 25-year
mandatory minimum and the “gross disparity” between that
provision and the five-year mandatory minimum that would
have applied had the amendment been made retroactive. See
McCoy, 981 F.3d at 285–86. Likewise, the concurrence would
have courts consider “a societal decision that the conduct for
which the prisoner was convicted incurs less moral opprobrium
than it previously had.” Post at 3. Such reasoning always runs
headlong into Congress’s judgment that the unamended statute
remains appropriate for previously sentenced defendants,
which is why even courts on the McCoy side of the split
recognize that a court may never grant compassionate release
based solely on prospective sentencing changes. See, e.g.,
Ruvalcaba, 26 F.4th at 27 (“if a district court were to reduce a
sentence solely because one of [a statute’s] non-retroactive
amendments would have lowered a defendant’s sentence, it
might be seen as substituting its own judgment on retroactivity
for Congress’s judgment”); id. at 32 (Barron, J., concurring)
(“the premise of our holding is not that a nonretroactive legal
change in and of itself can provide the ‘extraordinary and
compelling reason’ to reduce the sentence”). Of course, courts
still may consider individualized factors such as “the
defendants’ relative youth at the time of their offenses, their
post-sentencing conduct and rehabilitation, and the very
substantial terms of imprisonment they already served.”
McCoy, 981 F.3d at 288. But if those considerations do not
themselves warrant compassionate release, we should not tip
the balance by allowing courts to question whether the original
 18
mandatory minimum sentence was simply too long, either in
absolute terms or relative to the amendment. As Chief Judge
Sutton nicely summed up, “adding a legally impermissible
ground to three insufficient factual considerations does not
entitle a defendant to a sentence reduction.” United States v.
Jarvis, 999 F.3d 442, 444 (6th Cir. 2021).

Our analysis is consistent with the Supreme Court’s recent
decision in Concepcion v. United States, 142 S. Ct. 2389
(2022). That case involved section 404 of the First Step Act,
which authorizes district courts to reduce sentences for certain
offenses involving crack cocaine. The Supreme Court held that
the district courts, in deciding whether to do so, may consider
“intervening changes of law or fact.” Id. at 2404. The Court
reasoned that a court may “consider any relevant materials” for
sentencing, except as limited by statute or the Constitution. Id.
at 2400. As explained above, the compassionate-release statute
imposes just such a limit, in authorizing a reduced term of
imprisonment only for extraordinary and compelling reasons.
Moreover, Concepcion “mentioned the compassionate-release
statute only to support the proposition that Congress knows
how to limit which considerations may be used to reduce a
sentence”—an observation that “undermines rather than helps”
the position urged here by Jenkins. United States v. King, 40
F.4th 594, 596 (7th Cir. 2022) (holding that Thacker survives
Concepcion); see also United States v. Bledsoe, No. 22-2022,
2022 WL 3536493, at *2 (3d Cir. Aug. 18, 2022) (holding that
Andrews survives Concepcion).

B

We now turn to Winstead, under which Jenkins should not
have been classified as a career offender. As a “judicial
construction” of the career offender guideline, Winstead
establishes what that guideline meant “before as well as after”
 19
the date it was decided. Rivers v. Roadway Exp., Inc., 511 U.S.
298, 312–13 (1994). Winstead thus establishes that the district
court miscalculated Jenkins’ guideline range. But given the
availability of direct appeal and collateral review under section
2255 of title 28, we conclude that legal errors at sentencing—
including those established by the retroactive application of
intervening judicial decisions—cannot support a grant of
compassionate release. And even if they otherwise could, the
habeas-channeling rule of Preiser v. Rodriguez, 411 U.S. 475
(1973), would bar their consideration outside the context of
direct appeals or collateral review under section 2255.

1

Legal errors at sentencing are neither extraordinary nor
compelling. When they occur, they may be corrected on direct
review, including through the retroactive application of
intervening judicial decisions. Griffith v. Kentucky, 479 U.S.
314 (1987). Sentencing errors may also sometimes be
corrected on collateral review under section 2255. But because
collateral review significantly undermines the government’s
important interest in finality, Congress and the Supreme Court
have established many procedural rules limiting the availability
of such relief, including for errors made clear by intervening
judicial decisions. They have also established many exceptions
to these limitations, reflecting a careful balance between the
government’s interest in finality and the defendant’s interest in
obtaining relief from an unlawful sentence. The relevant
doctrines include retroactivity limits and exceptions, Teague v.
Lane, 489 U.S. 288, 311–12 (1989) (plurality opinion);
procedural-default rules and exceptions, Wainwright v. Sykes,
433 U.S. 72, 87 (1977); a statute of limitations and exceptions,
28 U.S.C. § 2255(f); and a bar on successive or abusive
motions and exceptions, id. § 2255(h). Congress “does not
alter the fundamental details of a regulatory scheme in vague
 20
terms.” Whitman v. Am. Trucking Ass’ns, 531 U.S. 457, 468
(2001). We must therefore interpret the compassionate-release
statute in light of this reticulated scheme for collateral review,
rather than invoke compassionate release to end-run its limits.
Thus, even if procedural hurdles would now prevent Jenkins
from raising his sentencing argument in collateral proceedings,
we cannot treat such a bar on relief under section 2255 as
supporting a finding of extraordinary and compelling
circumstances. As the Seventh Circuit explained: “There’s
nothing ‘extraordinary’ about new statutes or caselaw … ;
these are the ordinary business of the legal system, and their
consequences should be addressed by direct appeal or collateral
review under 28 U.S.C. § 2255.” King, 40 F.4th at 595
(emphasis added).

Treating sentencing errors as cognizable would also bring
the compassionate-release statute into conflict with other
provisions of the Sentencing Reform Act. As explained above,
the SRA enacted section 3582(c)(1)(B) of Title 18, which
permits a district court to “modify an imposed term of
imprisonment to the extent otherwise expressly permitted …
by Rule 35 of the Federal Rules of Criminal Procedure.” At
the same time, the SRA substantially narrowed the district
courts’ ability to modify illegal sentences under Rule 35.
Before the SRA was enacted, Rule 35 allowed the district
courts to “correct an illegal sentence at any time.” Fed. R.
Crim. P. 35(a) (1984). But the SRA amended Rule 35 to permit
sentence modifications in only two narrow circumstances—
(a) on remand from a court of appeals, or (b) to reflect a
defendant’s substantial assistance to the government, upon a
government motion filed within one year of the sentence. Pub.
L. No. 98-473, § 215(b), 98 Stat. at 2015–16. Rule 35 is now
only slightly broader: It also allows district courts to correct “a
sentence that resulted from arithmetical, technical, or other
clear error” within 14 days of sentencing. Fed. R. Crim. P.
 21
35(a) (2021). These post-SRA rules are far stricter than their
civil counterpart, which permits district courts to grant relief
from a judgment for a “mistake,” including a court’s error of
law, “within a reasonable time” of up to one year. Fed. R. Civ.
P. 60(b)(1), (c)(1); see Kemp v. United States, 142 S. Ct. 1856,
1861–62 (2022).

As this history indicates, Congress made a conscious
judgment, which Rule 35 still reflects, that a district court’s
authority to correct its own sentencing errors should be “very
narrow.” Fed. R. Crim. P. 35 advisory committee’s note to
1991 amendment. The Rule’s narrow terms do not “raise
doubts about the finality of determinate sentencing” that the
Sentencing Reform Act “attempted to resolve.” Id. But using
compassionate release to correct sentencing errors would blow
open these carefully crafted limits, violating the cardinal
principle that “specific statutory language should control more
general language when there is a conflict between the two.”
Nat’l Cable & Telecomms. Ass’n v. Gulf Power Co., 534 U.S.
327, 335 (2002).

As with intervening prospective statutes, it is no answer to
say that intervening judicial decisions may be considered only
as one factor among many, to be invoked only sparingly. For
one thing, because courts cannot pick needles from haystacks
ex ante, finality would be undermined whenever any judicial
decision, rendered any time after a sentence has become final
on direct review, at least arguably calls the sentence into
question. For another, judges would have widely divergent
views about when such decisions, considered only sometimes
and combined somehow with other factors, rise to the level of
an “extraordinary and compelling reason” for early release.
We doubt that the SRA—which effected a profound shift from
indeterminate to determinate sentencing—contained these
seeds of its own destruction. Finally, if an intervening judicial
 22
decision by itself could never support compassionate release,
we fail to see how such a “legally impermissible” consideration
could do so when combined with other “insufficient factual
considerations.” Jarvis, 999 F.3d at 444.

Jenkins claims that Johnson recognized Winstead errors as
cognizable on compassionate release. It did not. As discussed
above, the district court in Johnson erred in “considering itself
bound” by section 1B1.13 in a case involving a defendant-filed
motion. 858 F. App’x at 383. We determined that this error
was not harmless because the defendant had raised a Winstead
claim which did not fall within the grounds for compassionate
release recognized by the Sentencing Commission. Id. at 384.
But we did not affirmatively hold that a Winstead error could
support compassionate release. Rather than decide that
question, we remanded for the district court to consider it in the
first instance. Id.

2

The habeas-channeling rule of Preiser independently
forecloses using compassionate release to correct sentencing
errors. The writ of habeas corpus—including section 2255, the
habeas substitute for federal prisoners—traditionally “has been
accepted as the specific instrument to obtain release from
[unlawful] confinement.” Preiser, 411 U.S. at 486. As a result,
an inmate may not rely on a generally worded statute to attack
the lawfulness of his imprisonment, even if the terms of the
statute literally apply. Id. at 489. This includes both direct
attacks seeking “an injunction compelling speedier release”
and indirect attacks seeking “a judicial determination that
necessarily implies the unlawfulness of the State’s custody.”
Wilkinson v. Dotson, 544 U.S. 74, 81 (2005).

Jenkins’ assertion of Winstead error amounts to just such
an attack on the lawfulness of his sentence. For a court to grant
 23
compassionate release based on alleged Winstead error, it
would have to conclude that there in fact was Winstead error,
which would imply the invalidity of the sentence as originally
imposed. Because that is precisely what the habeas-channeling
rule forbids, we join the circuits holding that defendants cannot
seek compassionate release based on legal errors at sentencing,
including errors made clear through the retroactive application
of intervening precedent. King, 40 F.4th at 595–96 (7th Cir.);
Crandall, 25 F.4th at 586 (8th Cir.); Hunter, 12 F.4th at 567–
68 (6th Cir.); United States v. Mata-Soto, 861 F. App’x 251,
255 (10th Cir. 2021); United States v. Henderson, 858 F. App’x
466, 469 (3d Cir. 2021). Jenkins cites no circuit precedent to
the contrary, and we have not found any.

The arguments Jenkins does make are unpersuasive. First,
he claims that the government forfeited reliance on the habeas-
channeling rule by failing to raise it in its memorandum on
appeal. But as the appellee here, the government was not even
required to file a brief, see Fed. R. App. P. 31(c), much less to
take affirmative steps to preserve a ground for affirmance laid
out in the decision under review. In any event, the government
did argue that “appellant’s remedy” for his Winstead claim “is
to file a motion pursuant to 28 U.S.C. § 2255” rather than one
for compassionate release, and it collected cases from other
circuits holding as much. Appellee’s Mem. at 21–22.2

2
In his supplemental brief, Jenkins also argues that the
government failed to raise the habeas-channeling rule below. This
objection is itself forfeited: Although the government raised the
habeas-channeling rule in its memorandum on appeal, Jenkins did
not object in his reply memorandum that the government had
forfeited this point below. See Ass’n of Am. R.Rs. v. DOT, 821 F.3d
19, 25–26 (D.C. Cir. 2016); Reply Mem. at 10–11. In addition, the
district court rejected Jenkins’ Winstead argument on the ground that
 24
Second, Jenkins asserts that the habeas-channeling rule
governs only suits under 42 U.S.C. § 1983 or at least civil suits
generally. But we have applied the rule outside of section 1983
actions. See, e.g., Skinner v. DOJ, 584 F.3d 1093, 1099 (D.C.
Cir. 2009) (Privacy Act). Likewise, the civil-criminal
distinction is irrelevant, for the habeas-channeling rule simply
reflects the elementary principle that specific statutes qualify
general ones. Preiser, 411 U.S. at 489. After a conviction has
become final on direct review, habeas corpus is “the specific
instrument” for challenging assertedly unlawful confinement.
Id. at 486 (emphasis added). And as explained above, it
contains a bevy of retroactivity, procedural default, statute of
limitations, and successiveness rules specifically designed to
balance the competing interests in finality and error correction.
Measured in terms of “relative specificity” to habeas, post at
11, the compassionate-release statute is not distinguishable
from section 1983.

The concurrence seeks to distinguish the compassionate-
release statute as one “expressly designed to give judges the
discretion to modify sentences.” Post at 13. But regardless of
whether the statute literally extends to claims of sentencing
error established by intervening precedents, such claims are
hardly at its core. Most obviously, the compassionate-release
statute covers factors like those enumerated by the Sentencing
Commission—health, age, and family circumstances—which
turn on post-sentencing changes to a prisoner’s individual
situation, not on legal errors at sentencing. Thus, applying the
habeas-channeling rule to the compassionate-release statute
still gives the latter a broad range of meaningful application,

“the compassionate release statute was not intended to serve as a
second chance to address a defendant’s sentence.” ECF Doc. 42, at
10. That assertion—which fairly includes the points we address
here—teed the issue up in this court, for Jenkins makes no argument
that the district court abused its discretion in reaching it.
 25
just as applying the habeas-channeling rule to section 1983, a
statute expressly designed to provide redress for constitutional
violations, still gives that statute a broad range of meaningful
application.

Third, Jenkins reasons that because a grant of
compassionate release is discretionary, establishing legal error
at sentencing would not necessarily shorten his sentence. But
the habeas-channeling rule applies so long as “success in [the
inmate’s] action would necessarily demonstrate the invalidity
of confinement or its duration.” Wilkinson, 544 U.S. at 82; see
Edwards v. Balisok, 520 U.S. 641, 648 (1997). And for Jenkins
to prevail on his Winstead claim, the court would necessarily
have to conclude that his sentence was unlawfully imposed.
Otherwise, what could possibly be extraordinary and
compelling about an intervening judicial decision confirming
that a sentence was lawfully imposed, or simply not resolving
the question one way or the other? In sum, Jenkins highlights
that under his theory, establishing legal error at sentencing
would not be sufficient to warrant compassionate release. But
under Preiser, what matters is that establishing legal error at
sentencing would be necessary to granting compassionate
release based on an intervening judicial decision.

Fourth, Jenkins invokes Davis v. U.S. Sentencing
Commission, 716 F.3d 660 (D.C. Cir. 2013). That case
involved an equal-protection challenge to an amendment
retroactively lowering the guideline range for certain low-
volume drug offenses, but not for higher-volume offenses. Id.
at 661. We held the suit did not implicate the habeas-
channeling rule because “success on the merits” would not
“necessarily imply the invalidity of confinement or shorten its
duration.” Id. at 666 (cleaned up). That was so because the
promulgation of a guideline with retroactive effect does not
invalidate earlier sentences as originally imposed, but merely
 26
allows the affected inmates “to seek discretionary sentence
reductions.” Id. at 662. In contrast, a sentence issued after a
guideline calculation error is unlawfully imposed and subject
to vacatur on appeal unless the error was harmless. See Gall,
552 U.S. at 51.

One final point: The concurrence observes that Winstead
may not satisfy the Teague requirements for retroactive
application on habeas. Post at 12–13. Based on that premise,
the concurrence treats Winstead as no different from an
intervening “legislative change.” Post at 12. In our view, this
account misunderstands how judicial decisions work. See, e.g.,
Rivers, 511 U.S. at 312 (judicial decisions establish what the
law meant from date of enactment); Crandall, 25 F.4th at 586
(intervening judicial decision “did not change the law; it was
an interpretation of existing law”). But if the concurrence were
correct on this point, that would simply make Jenkins’
argument based on Winstead “conceptually the same” as his
argument based on the stacking amendment. Post at 12. If so,
then both arguments would fail for reasons explained above.

C

Finally, we address the argument based on Borden, which
establishes that Jenkins could not have received a 15-year
mandatory minimum under ACCA for his felon-in-possession
offenses. According to Jenkins, the possibility of such a
sentence pressured him to accept a less favorable plea deal than
he otherwise would have, and that is an extraordinary and
compelling reason for compassionate release.

To the extent Jenkins claims that his guilty plea was
involuntary, he mounts a due process challenge to the
conviction itself, see McCarthy v. United States, 394 U.S. 459,
466 (1969), which is subject to the habeas-channeling rule.
The challenge is also foreclosed by Brady v. United States, 397
 27
U.S. 742 (1970), which held that a guilty plea does not become
invalid simply because “later pronouncements of the courts …
hold that the maximum penalty for the crime in question was
less than was reasonably assumed at the time the plea was
entered.” Id. at 757.

To the extent Jenkins claims that his sentence, although
lawful, is harsher than it otherwise would have been, there is
nothing extraordinary about that. Of course, plea bargaining is
ubiquitous in the criminal justice system. See Missouri v. Frye,
566 U.S. 134, 144 (2012). And baked into the notion of plea
bargaining is that both parties forgo potentially meritorious
arguments to obtain a more certain, second-best result. The
defendant gives up his chance at acquittal, but he gains the
substantial likelihood of receiving a lower sentence than the
one he would have received had he been convicted at trial.
Brady, 397 U.S. at 751–52. The government forgoes pursuing
the maximum punishment available, but it gains the certainty
of a conviction and saves the time and expense of preparing for
trial. Id. at 752. In such a bargain, there is nothing unusual
about later decisions revealing which party would have
prevailed had the open issues been contested. See McMann v.
Richardson, 397 U.S. 759, 770 (1970).

Other considerations reinforce this analysis. For one
thing, to plead guilty, the defendant must affirmatively waive
any claim that the agreed-to sentence is legally or factually
inappropriate. See Fed. R. Crim P. 11(b). To grant early
release despite such a waiver, based on an intervening decision
suggesting that the defendant made a bad deal, would violate
the bedrock principle that even the “most basic rights of
criminal defendants” are waivable, Peretz v. United States, 501
U.S. 923, 936 (1991), and courts cannot grant relief based on
waived rights, United States v. Olano, 507 U.S. 725, 732–33
(1993). It would also deprive the government of the benefits
 28
for which it bargained—a guaranteed minimum level of
punishment without the risks and expense of trial. And if the
government cannot reliably obtain the benefit of its bargain,
then it will have less incentive to enter plea bargains in the first
place, making both itself and criminal defendants worse off in
the long run. See Frye, 566 U.S. at 144; Mabry v. Johnson, 467
U.S. 504, 508 (1984); Brady, 397 U.S. at 752.

D

The structure and history of the compassionate-release
statute reinforce our conclusion that the post-sentencing legal
changes discussed above are irrelevant to the question whether
extraordinary and compelling circumstances warrant early
release.

Consider the statutory exhaustion requirement. It requires
a defendant first to ask BOP to file a compassionate-release
motion on his behalf. 18 U.S.C. § 3582(c)(1)(A). If the Bureau
denies the request, the defendant then must exhaust all
administrative appeal rights within BOP, unless the Bureau
fails to act timely. Id.

Exhaustion requirements allow the agency “to apply its
special expertise” and to “produce a useful record for
subsequent judicial consideration.” McCarthy v. Madigan, 503
U.S. 140, 145 (1992). We should expect, then, that the
extraordinary-and-compelling-reasons inquiry is one that the
BOP is well suited—or at least institutionally competent—to
make. The Bureau is charged with holding federal prisoners
and preparing them for release. 18 U.S.C. § 4042(a). That
makes it well suited to evaluate the kinds of circumstances the
Sentencing Commission has identified—an inmate’s age,
health, and family circumstances. But the BOP plays no role
in plea bargaining, calculating guideline ranges, or determining
the lawfulness of individual sentences. Within the Department
 29
of Justice, that is the province of the United States Attorneys,
who represent the government at sentencing. 28 U.S.C.
§ 547(1); DOJ, Justice Manual § 9-27.730 (2018). If Jenkins’
asserted grounds for release were cognizable, the
compassionate-release statute would require the BOP to make
thousands of determinations every year in areas well beyond its
core mission and expertise.

Jenkins’ reading is also at odds with the concept of
compassionate release. When “pertaining to allowances, leave,
etc.,” “compassionate” means “[g]ranted out of compassion,
without legal or other obligation.” Compassionate, Oxford
English Dictionary (2d ed. 1989); see, e.g., Compassionate,
Webster’s Ninth New Collegiate Dictionary (1990) (“granted
because of unusual distressing circumstances affecting an
individual”); Thomas v. Farley, 31 F.3d 557, 558–59 (7th Cir.
1994) (inmate sought “compassionate leave” “to attend his
mother’s funeral”); Gwin v. Snow, 870 F.2d 616, 618 (11th Cir.
1989) (inmate sought “compassionate leave when his mother
died”); United States v. Lawson, 39 C.M.R. 726, 727–28
(1968) (soldier given “compassionate leave” because of
“problems at home”); cf. Compassionate Leave, Cambridge
Business English Dictionary (2011) (“paid time off work that
someone is allowed to have because a member of their family
has died, or because they have a serious personal problem”).

Statutes authorizing “compassionate release” thus
typically permit release only for age, health, family, and similar
personal circumstances. See, e.g., D.C. Code § 24-403.04(a)
(age, medical condition, family circumstances); Ill. Stat. Ch.
730 § 5/3-3-14(b) (medical condition); Mich. Comp. L.
§ 771.3h(1) (same); N.J. Stat. § 30:4-123.51e(f)(1) (same); see
also Compassionate Release, Black’s Law Dictionary (11th ed.
2019) (equating the term with “medical parole”). In contrast,
 30
Jenkins seeks release based on changes in the legal landscape,
which are not ordinarily cognizable under such statutes.

Jenkins objects that section 3582(c)(1)(A) does not itself
use the name “compassionate release.” True enough, but the
phrase nonetheless has interpretive significance.

The government bodies charged with implementing
section 3582(c)(1)(A) consistently have described it as a
mechanism for “compassionate release.” The BOP’s first
regulations implementing the statute used that term. See
Control, Custody, Care, Treatment and Instruction of Inmates;
Compassionate Release, 59 Fed. Reg. 1238 (Jan. 7, 1994). So
do less formal BOP guidance documents. See Memorandum
of Kathleen M. Hawk, Director, BOP, to Executive Staff (July
22, 1994), reprinted in DOJ, Office of the Inspector General,
The Federal Bureau of Prisons’ Compassionate Release
Program, No. I-2013-006, at 67, 67 (2013) (BOP “has
historically taken a conservative approach to filing a motion
with the courts for the compassionate release of an inmate
under … § 3582(c)(1)(A)”). The Sentencing Commission and
the DOJ Office of the Inspector General also use the term
“compassionate release” to refer to section 3582(c)(1)(A).
Guidelines Manual: Supplement to Appendix C, supra, amend.
799, at 126; Office of the Inspector General, supra, at i n.3. So
do courts and litigants. See, e.g., United States v. Jackson, 26
F.4th 994, 996 (D.C. Cir. 2022); Long, 997 F.3d at 347. It is
hard to see how the name “compassionate release” could gain
such universal and unquestioned acceptance absent a shared
understanding that section 3582(c)(1)(A) concerns
compassionate grounds for release.

In the First Step Act, Congress itself used “Compassionate
Release” to refer to section 3582(c)(1)(A). Pub. L. No. 115-
391, § 603(b), 132 Stat. at 5239. Like a statutory title or section
 31
heading, this name is “a short-hand reference to the general
subject matter involved,” and so may be relied upon to “shed
light” on it. Brotherhood of R.R. Trainmen v. Balt. & Ohio
R.R., 331 U.S. 519, 528–29 (1947); see A. Scalia & B. Garner,
Reading Law: The Interpretation of Legal Texts 221 (2012).
And because the First Step Act amends section 3582(c)(1)(A),
the two should be read in pari materia. See, e.g., United States
v. Stewart, 311 U.S. 60, 64 (1940); Scalia & Garner, supra, at
254–55.

Finally, Jenkins’ reading is inconsistent with Sentencing
Commission commentary, which is binding for BOP-filed
motions and at least persuasive here. As explained above, that
commentary flags considerations such as a defendant’s health,
age, and need to care for family members. U.S.S.G. § 1B1.13
cmt. n.1(A)–(C). The commentary also states that a “reason
other than, or in combination with, the reasons” specifically
identified may also qualify. Id. cmt. n.1(D). Under the canon
of ejusdem generis, this catchall provision “embrace[s] only
objects similar in nature to those objects enumerated by the
preceding specific words.” Circuit City Stores, Inc. v. Adams,
532 U.S. 105, 114–15 (2001) (cleaned up). That would include
other kinds of sympathetic personal circumstances, but not
intervening changes in sentencing law.

V

In closing, Jenkins argues that the district court abused its
discretion by failing to consider in combination the various
grounds he had asserted. Jenkins is correct that factors may
sometimes become extraordinary and compelling when
considered together. See U.S.S.G. § 1B1.13 cmt. n.1(D). And
here the district court did not explicitly address the combined
weight of Jenkins’ arguments. Still, the court did not abuse its
discretion. It correctly determined that Jenkins’ arguments
 32
about the intervening changes in sentencing law were legally
irrelevant to the compassionate-release determination. That
left only arguments about his own health and family
circumstances. The court reasonably found that these
circumstances were minimally significant, so it did not need to
say explicitly that their combined force did not rise to the level
of extraordinary and compelling circumstances.

VI

The district court properly denied Jenkins’ motion for
compassionate release.

Affirmed.
 GINSBURG, Senior Circuit Judge, concurring in part,
dissenting in part, and concurring in the judgment: Although I
agree with the result in this case, I write separately because I
believe the Court errs in some of its analyses.

I. The Narrowed Stacking Provision

Had Jenkins been convicted and sentenced for both of his
charges under 18 U.S.C. § 924(c), he would have faced a
mandatory 30-year prison sentence under the then-applicable
“stacking” provision. 18 U.S.C. § 924(c)(1)(C)(i)(2012). The
First Step Act has since narrowed the stacking provision
prospectively, however, such that his two § 924(c) charges
would now lead to a mandatory 10-year sentence. See First
Step Act, § 403(a), 132 Stat. 5194, 5221–22 (codified at 18
U.S.C. § 924(c)(1)(C)). This discrepancy, Jenkins contends,
helps support a finding of “extraordinary and compelling
reasons” that warrant a reduction in his sentence under the
compassionate release statute, 18 U.S.C. § 3582(c)(1)(A)(i).
This argument implicates the question whether a non-
retroactive change to the law can ever be considered an
“extraordinary and compelling reason[],” a question that has
divided the courts of appeal. The Court today joins those courts
that have answered the question in the negative. The Court’s
arguments notwithstanding, I would not adopt a rule that
categorically precludes consideration of a non-retroactive
change in the law. I nevertheless agree that the narrowed
stacking provision does not establish “extraordinary and
compelling reasons” in the circumstances Jenkins presents.

A. Merits of the Court’s Reasoning

Central to my disagreement with the Court’s reasoning is
that the statute provides no basis for categorically precluding
consideration of a non-retroactive change to the law. The
statute does not define the term “extraordinary and
compelling,” and I see no reason to impose upon it a categorical
 2
exclusion. The Congress well knows how to preclude
consideration of certain factors. That is exactly what it did
when it stated that “[r]ehabilitation of the defendant alone shall
not be considered an extraordinary and compelling reason.” 28
U.S.C. § 994(t). But that is the only limit the Congress placed
upon the term. “In light of [this] specific statutory exclusion,”
I am “reluctant to infer that Congress intended” categorically
to exclude the consideration of a non-retroactive change to the
law. United States v. Ruvalcaba, 26 F.4th 14, 26 (1st Cir. 2022)
(citation omitted).

The Court dismisses this straightforward, textualist point
by asserting that it is the ordinary practice to apply new, lower
penalties only prospectively, and “what the Supreme Court
views as the ‘ordinary practice’ cannot also be an
‘extraordinary reason’ to deviate from that practice.” Ct. Op.
at 15 (cleaned up). Likewise, the Court asserts that “common
practice” can never be a compelling reason. Id. The Court has
thus committed the Fallacy of Division — assuming what is
true of the whole must be true of each part. Just because the
general practice of applying new penalties only prospectively
is “ordinary” or “common” does not mean that a non-
retroactive change in the law can never so much as contribute
to a finding of “extraordinary and compelling reasons” even
after considering the idiosyncratic circumstances of a particular
defendant. A compassionate release motion requires a district
court to “assess interactions among a myriad of factors” as part
of an “individualized consideration of a defendant’s
circumstances,” and because judges are not soothsayers, they
cannot possibly “predict how this mix of factors — including
non-retroactive changes in sentencing law — will play out in
every case.” Ruvalcaba, 26 F.4th at 27.

Following the lead of some other circuits, the Court further
says that allowing a district court to consider a non-retroactive
 3
change to the law “would usurp these quintessentially
legislative judgments” not to make laws retroactive. Ct. Op. at
16. That claim would be quite formidable had anyone argued
that a non-retroactive change in the law should always, or even
often, be considered an “extraordinary and compelling” reason
for a sentence reduction. But no one has made that argument.
The contrary approach simply contends that a district judge
should have the discretion to consider a non-retroactive change
to the law as part of an individualized assessment of a
defendant’s unique circumstances. This is particularly relevant
to a change that reflects a societal decision that the conduct for
which the prisoner was convicted incurs less moral opprobrium
than it previously had. Under that reasonable approach, cases
in which legislative change contribute to “a finding of
‘extraordinary and compelling reasons’ should be relatively
rare.” United States v. Maumau, 993 F.3d 821, 838 (10th Cir.
2021) (Tymkovich, J., concurring). “[A]llowing for the
provision of individual relief in the most grievous cases” does
not in any way usurp the legislative determination of the
Congress to eschew “automatic vacatur and resentencing of an
entire class of sentences” and the attendant logistical
nightmares. United States v. McCoy, 981 F.3d 271, 286–87
(4th Cir. 2020).

I am equally unmoved by the Court’s suggestion that my
reading of the statute is inconsistent with that of the Sentencing
Commission. Although I agree with the Court’s conclusion
that a district judge does not abuse his or her discretion by
looking to section 1B1.13 of the Sentencing Guidelines and its
commentary as “persuasive” authority, I think the Court
assigns the policy statement of the Guidelines more weight
than a fair reading of our decision in United States v. Long can
support. 997 F.3d 342 (2021). The Court asserts:
 4
[T]he Act’s sole change was to create a new
procedural avenue for relief. It did not alter the
extraordinary-and-compelling-reasons standard,
so it did not undermine the Commission’s
interpretation of that standard.

Ct. Op. at 11–12. The Court contends that this proceduralist
reading of the First Step Act is consistent with our decision in
Long, as there “we did not consider[] the distinct question
whether courts may rely on” the Sentencing Commission’s
interpretation of the standard “as persuasive authority.” Ct. Op.
at 12.

To the contrary, Long did much to undermine the
Commission’s interpretation of the standard, at least as applied
to a motion filed by a prisoner. In Long we first explained that
the change the First Step Act made to the compassionate
release statute (allowing a prisoner to file a motion on his or
her own behalf) was spurred by the Congress’s displeasure
with the Bureau of Prisons’ sparing use of that statute, as
evidenced by the title of the relevant section of the Act:
“Increasing the Use and Transparency of Compassionate
Release.” 997 F.3d at 348. Later, in explaining why the policy
statement of the Sentencing Guidelines did not apply to a
motion filed by a prisoner, the court stated that “[a]ny change
by Congress to the substantive reach of the statutory sentencing
scheme may rightly be expected to result in a change to the
policy statements guiding those statutes’ implementation.” Id.
at 356. Thus, the Long court held that the policy statement was
of limited value for interpreting the scope of “extraordinary and
compelling reasons” as applied to a motion filed by a prisoner,
as opposed to one filed by the Bureau of Prisons, because the
Congress had evinced an intent to expand use of compassionate
release, and the Sentencing Commission could be expected to
update its Guidelines to bring them into alignment with the
 5
Congress’s more permissive attitude. In other words, the
change to the statute was not a mere procedural change, as the
Court would have us believe; it was a “paradigm shift,” and
“[i]t would blink reality to assume that the Sentencing
Commission would think that the only modifications necessary
to the existing policy statement would be to disregard the
references to motions brought by the” Bureau of Prisons.
Ruvalcaba, 26 F.4th at 22.

The Supreme Court’s opinion in the recent case of
Concepcion v. United States buoys my conviction that the
statute should not be read as barring the district court from
considering a non-retroactive change in the law. 142 S. Ct.
2389 (2022). In Concepcion, the Supreme Court lauded the
“venerable tradition of discretion” in sentencing and sentence-
modification proceedings. Id. at 2401 n.4. Based upon that
tradition, the Supreme Court cautioned the lower courts against
reading exceptions into statutes dealing with the discretion of
judges to consider information during sentencing or sentence-
modification proceedings. “The only limitations on a court’s
discretion to consider any relevant materials . . . are those set
forth by Congress in a statute or by the Constitution.” Id. at
2400.

To be sure, unlike the Supreme Court in Concepcion, we
are not dealing here with an implied exception to a judge’s
discretion; the question, rather, is the scope of the phrase
“extraordinary and compelling reasons.” But the same
tradition that gave rise to Concepcion’s admonition against
creating non-explicit exceptions to a judge’s discretion
counsels against a parsimonious reading of the statute at hand.
The Court’s reasoning is profoundly inconsistent with the
Supreme Court’s teaching in Concepcion. The Congress has
created only two relevant limitations that control the district
courts’ reading of the standard: One for controlling Sentencing
 6
Commission policy statements, and the other for
“[r]ehabilitation . . alone.” 28 U.S.C. § 994(t). Because the
“Congress has not legislated to create a third limitation on
extraordinary and compelling reasons prohibiting district
courts from considering non-retroactive changes in sentencing
law, we” should “decline to create one now.” United States v.
Chen, No. 20-50333, 2022 WL 4231313, at *6 (9th Cir. Sept.
14, 2022) (relying upon Concepcion, 142 S. Ct. at 2396).

B. Alternative Ground

Despite my disagreement with the Court’s reasoning, I
agree that Jenkins’s argument based upon the narrowing of the
stacking provision does not support a finding of “extraordinary
and compelling reasons” in this case. Because Jenkins pleaded
guilty and was sentenced for only one § 924(c) charge, he was
not sentenced under the stacking provision. Therefore, Jenkins
argues that the plea negotiations were tainted by the possibility
of being sentenced under the then-existing stacking provision,
which caused him to enter a harsher plea agreement than he
otherwise would have done. This is of a piece with the
argument Jenkins made based upon Borden v. United States,
regarding the 15-year mandatory minimum in the Armed
Career Criminal Act. 141 S. Ct. 1817 (2021). The Court’s
treatment of the Borden issue, in Part IV.C, is therefore
sufficient to decide the issue presented by the narrowing of the
stacking provision. Finding myself in agreement with the
Court’s treatment of the Borden issue, I also agree that the
narrowing of the stacking provision does not support a finding
of “extraordinary and compelling” circumstances in this case.

II. Winstead Argument

In United States v. Winstead, we held that an inchoate
offense does not qualify as a predicate offense for career
 7
offender status under the Sentencing Guidelines. 890 F.3d,
1082, 1091 (2018). Had that standard been applied to Jenkins,
he would not have been considered a career offender. In a pre-
Winstead world, however, Jenkins and the Government both
assumed his inchoate offenses were predicate offenses and that
he therefore was a career offender. This assumption had a
tangible effect upon Jenkins’s sentence. As the district judge
acknowledged, Jenkins’s “exposure to a significantly higher
Guidelines range due to his then-undisputed status as a career
offender was apparently a significant factor behind . . . the
agreed sentence presented to the Court.” ECF Doc. 42, 9–10
(cleaned up).

Jenkins argues that having been sentenced under the
assumption he was a career offender helps support a finding of
“extraordinary and compelling reasons.” His argument
presents the question whether a change in the legal landscape
established by an intervening judicial decision can ever support
a finding of “extraordinary and compelling reasons.” The
Court answers no for two reasons, both of which I find wanting:
(1) This type of change is neither extraordinary nor compelling;
and (2) even if a claim based upon this type of change literally
fits the words of the statute, the “habeas-channeling” rule set
forth in Preiser v. Rodriguez, recognizes an implicit exception
for such claims. 411 U.S. 475, 489–90 (1973). Because I
disagree with these reasons, I would not announce a categorical
rule precluding the consideration of a Winstead argument, or
any similar argument based upon a judicial ruling, in assessing
whether there are “extraordinary and compelling reasons” that
warrant compassionate release. I nevertheless agree that under
the circumstances presented by Jenkins, his Winstead argument
does not help establish “extraordinary and compelling reasons”
for compassionate release.
 8
My disagreement with the Court’s first reason, that a
change to the legal landscape established by judicial decision
is never an “extraordinary and compelling” reason largely
tracks my disagreement with the Court’s similar conclusion
regarding non-retroactive changes in the law. My comments
here are therefore confined to the Court’s second reason — that
the habeas-channeling rule precludes consideration of
Jenkins’s Winstead argument — and to explaining why I
nevertheless would reject Jenkins’s Winstead argument.

The Court’s second argument is wrong for two reasons: (1)
the Government forfeited this argument, and (2) even if it had
not, the habeas-channeling rule, properly understood, does not
apply to motions for compassionate release.

A. Forfeiture

“[T]he habeas channeling rule . . . can be forfeited if a
defendant fails to assert it.” Dufur v. United States Parole
Comm’n, 34 F.4th 1090, 1095 (D.C. Cir. 2022). There is no
doubt the Government failed to assert it before the district
court, so it is forfeit. The Court, however, says that
because the Government raised the argument before us in its
Memorandum of Law and Fact and Jenkins failed to argue
forfeiture in his Reply Memorandum of Law and Fact, he
forfeited the forfeiture.

The Court sees in the Government’s Memorandum of Law
and Fact far more than I can find there. The Court points to
pages 21–22 where, after having argued that Jenkins’s claims
do not establish “extraordinary and compelling reasons,” the
Government concluded: “Instead . . . appellant’s remedy is to
file a motion pursuant to 28 U.S.C. § 2255, challenging the
validity of his decision to plead guilty and resulting sentence.”
Fairly read, all that says is that because sentencing error, in the
 9
view of the Government, is not an “extraordinary and
compelling” reason, Jenkins’s only recourse is a § 2255
proceeding. That is not the habeas-channeling argument,
which, as explained below in Part II.B, asserts that even if a
change in the law made by a judicial decision fits the terms of
the compassionate release statute, it is implicitly excluded
therefrom. Indeed, the Government’s memorandum makes no
mention of Preiser or any of the other seminal cases upon
which the habeas-channeling rule is based. Their conspicuous
absence makes perfectly clear what the Government did and
did not argue.

The several citations that follow the Government’s
statement of its position reinforces this point. One of those
cases, Ivy v. Pontesso, is completely irrelevant, as it deals with
the availability of relief for a federal prisoner under § 2241 as
opposed to § 2255. 328 F.3d 1057, 1059 (9th Cir. 2003). The
other cases make not a single mention of the habeas-channeling
rule or any of the cases that expound upon it. Rather, the cited
cases mention habeas relief to the same end as does the
Government, namely, to make the point that for legal claims
that are not “extraordinary and compelling” as a matter of law,
a prisoner’s only recourse is in a § 2255 proceeding.

The Court also asserts (in a footnote) that certain wording
in the district court’s opinion “fairly includes” the habeas-
channeling argument and thus “teed the issue up in this court.”
Ct. Op. at 23–24 n.2. All the district court said, however, was
that “the compassionate release statute was not intended to
serve as a second chance to address a defendant’s sentence.”
What follows that broad statement is a quotation from another
opinion rejecting the use a non-retroactive legislative change
to support a finding of “extraordinary and compelling reasons,”
an issue to which the habeas-channeling rule does not apply.
Indeed, throughout its opinion, the district court deals with
 10
Winstead and legislative changes as presenting the very same
issue. Clearly, then, all the district court meant in the above-
cited passage was that neither a claim based upon Winstead nor
a claim based upon legislative change can ever fit into the
wording of the compassionate release statute. Nary a whiff of
the habeas-channeling argument is to be found in this passage
or anywhere else in the district court’s opinion.

“In our adversarial system of adjudication, we follow the
principle of party presentation.” United States v. Sineneng-
Smith, 140 S. Ct. 1575, 1579 (2020). While “[t]here are no
doubt circumstances in which a modest initiating role for a
court is appropriate,” by adding a wholly new argument to the
Government’s presentation on appeal, the Court exceeds its
modest role as an “essentially passive instrument[] of [the
parties].” Id.

In short, the Government forfeited any habeas-channeling
argument, and a fair reading of the Government’s
Memorandum of Law and Fact provides no basis for
concluding that Jenkins’ forfeited the Government’s forfeiture.

B. Merits of the Habeas-Channeling Argument

Worse yet, the habeas-channeling argument is also plain
wrong. In adopting it, the Court makes a grave mistake. The
habeas-channeling rule simply does not apply to claims such as
Jenkins’s under the compassionate release statute.

The habeas-channeling rule made its debut in Preiser v.
Rodriguez. There, the Supreme Court held that state prisoners
alleging the unconstitutional deprivation of their good-time
credits could seek redress only under the federal habeas corpus
statute applicable to state prisoners, 28 U.S.C. § 2254, and not
under the Civil Rights Act of 1871, 42 U.S.C. § 1983. 411 U.S.
 11
at 489. “Despite the literal applicability of [§ 1983’s] terms”
to the prisoners’ claim, id., the Preiser Court found “an implicit
exception from § 1983’s otherwise broad scope for actions that
lie ‘within the core of habeas corpus.’” Wilkinson v. Dotson,
544 U.S. 74, 79 (2005) (quoting Preiser, 411 U.S. at 487).
Claims that involve “an attack by a person in custody upon the
legality of that custody” are therefore excluded from the scope
of § 1983. Preiser, 411 U.S. at 484.

This implied exception is rooted in the significant overlap
between the federal habeas statute and § 1983. While § 1983
creates a general cause of action for “the deprivation [under the
color of state law] of any rights, privileges, or immunities
secured by the Constitution and laws,” 42 U.S.C. § 1983, the
federal habeas statute deals with the similar but more specific
situation of a state prisoner who claims that “he is in custody
in violation of the Constitution or laws or treaties of the United
States.” 28 U.S.C. § 2254. This overlap created a conflict: The
habeas statute requires the exhaustion of state remedies, but
§ 1983 does not, so authorizing overlapping § 1983 suits would
practically nullify the habeas statute’s exhaustion requirement.
Preiser, 411 U.S. at 489–90. Given this conflict and the
relative specificity of the habeas statute, the Supreme Court
held that when a state prisoner attacks the legality of his
confinement and “seeks a determination that he is entitled to
immediate release or a speedier release from that
imprisonment, his sole federal remedy is a writ of habeas
corpus.” Id. at 500.

To state the rule is to see why it does not apply to a
Winstead argument. A prisoner making that argument is not
claiming his sentence is “invalid” or “unlawful.” Rather, the
prisoner concedes, at least for the purpose of his motion for
compassionate release, that the sentence is currently valid and
lawful, but nevertheless appeals to the equitable discretion of
 12
the judge for a sentence modification. See United States v.
Trenkler, 47 F.4th 42, 2022 WL 3711709, at *4 (1st Cir. 2022)
(“[H]abeas and compassionate release are distinct vehicles for
relief.”). Therefore, a motion for compassionate release, even
one based upon Winstead, does not go to the “core of habeas,”
which is to “claim[] the right to be released upon the ground
that the sentence was imposed in violation of the Constitution
or laws of the United States.” 28 U.S.C. § 2255(a).

The Court says this is wrong because, “[a]s a ‘judicial
construction’ of the career offender guideline, Winstead
establishes what that guideline meant ‘before as well as after’
the date it was decided.” Ct. Op. at 18 (quoting Rivers v.
Roadway Exp., Inc., 511 U.S. 298, 312 (1994)). But the
question whether a judicial construction has retroactive effect
is irrelevant to Jenkins’s argument. His argument does not rest
upon the contention that the sentencing judge committed a
legal error. Rather, in his motion for compassionate release, he
asks the judge, in the exercise of her discretion, to consider the
discrepancy between his sentence and a sentence he would
receive were he resentenced under current law. In this respect,
an argument based upon Winstead is no different from an
argument based upon a legislative change. See United States
v. Hunter, 12 F.4th. 555, 565 (6th Cir. 2021) (explaining that
changes based upon judicial decisions and legislative changes
are conceptually the same with respect to a motion for
compassionate release). In other words, even if, contrary to
Rivers, judicial opinions did not apply retrospectively, Jenkins
still would have made the same argument: That he would have
received a lower sentence had he been sentenced after the
Winstead decision, rather than before it, helps establish
grounds for a sentence reduction.

In any event, with respect to the habeas-channeling rule,
the only retroactivity that could matter is retroactivity with
 13
respect to a habeas petition. This is not the place for an
extended treatment of whether Winstead applies retroactively
to a habeas petition. Suffice it to say that it is far from clear
that it satisfies the demanding requirements for retroactivity set
forth by Teague v. Lane, 489 U.S. 288 (1989). The Court does
not even attempt to show that it does, although the Court’s
reliance upon the retroactivity of judicial opinions is
meaningless without that showing.

Finally, as mentioned above, the habeas-channeling rule of
Preiser is based upon the “implicit [habeas] exception” to
§ 1983. Wilkinson, 544 U.S. at 79. The Court in effect argues
that the compassionate release statute includes a similar
implicit exception. The Supreme Court has never applied the
habeas-channeling rule to a non § 1983 case. This court has
done so, see Davis v. U.S. Sentencing Commission, 716 F.3d
660, 662–63 (2013); Skinner v. DOJ, 584 F.3d 1093, 1099
(2009), but not when dealing with a statute expressly designed
to give judges the discretion to modify sentences. Reading an
implicit habeas exception into “a statute whose very purpose is
to open up final judgments,” Concepcion, 142 S. Ct. at 2398
n.3, is a far cry from what the Supreme Court did in Preiser. It
also runs afoul of Concepcion’s clear admonition against
reading a limitation into a statute providing judges with
sentencing discretion. 142 S. Ct. at 2400.

C. Alternative Grounds

Despite my misgivings about the Court’s reasoning, I
agree that Jenkins’s Winstead argument does not support a
finding of “extraordinary and compelling reasons” in his
circumstances. Without career-offender status, the Guidelines’
range for the two offenses to which Jenkins pleaded guilty
would have been seven to seven-and-one-half years. Jenkins
was instead sentenced to eight years. Because the other reasons
 14
he provided (his health and his need to care for his mother) are
legally insignificant and the discrepancy in his sentence is
modest, I do not believe Jenkins has presented “extraordinary
and compelling reasons” for early release. Although I believe
the district judge erred, and hence abused her discretion, by
concluding that she was not permitted even to consider
Jenkins’s Winstead argument, remanding the case to her would
be pointless, as she would inevitably reach the same conclusion
even if she applied what I believe to be the correct standard.

III. Conclusion

This case should never have come before us. Had the
Sentencing Commission not been left without a quorum for
years, it would by now have published a policy statement
providing guidance on the relevant questions. In a long
overdue development, the Senate recently confirmed seven
nominees to the Commission, allowing it to resume the
important work for which the Congress created it. I look
forward to the Commission soon clearing up the confusion
wrought by the protracted absence of a quorum.

As explained above, I respectfully dissent from much of
the Court’s reasoning and the broad implications that flow from
it, but I concur in the judgment of the Court because Jenkins’s
specific circumstances do not support a finding of
“extraordinary and compelling reasons” that warrant a
reduction of his sentence.