joined by LUCERO, TYMKOVICH, HOLMES, BACHARACH, and MORITZ, Circuit Judges.
ON REHEARING EN BANC
New statutes have proven as enigmatic as 18 U.S.C. § 924(c). Everyone knows that, generally speaking, the statute imposes heightened penalties on those who use guns to commit violent crimes or drug offenses. But the details are full of devils. Originally passed in 1968, today the statute says that “any person who, during and in relation to any crime of violence or drug trafficking crime ... uses or carries a firearm, or who, in furtherance of any such crime, possesses a firearm, shall, in addition to the punishment provided for such crime ... be sentenced to a term of imprisonment of not less than 5 years.” 18 U.S.C. § 924(c)(1)(A). That bramble of prepositional phrases may excite the grammar teacher but it’s certainly kept the federal courts busy. What does it mean to “use” a gun “during and in relation to” a drug trafficking offense? The question rattled around for years until Bailey v. United States, 516 U.S. 137, 116 S.Ct. 501, 133 L.Ed.2d 472 (1995), and even now isn’t fully resolved. What does and doesn’t qualify as a “crime of violence”? The better part of five decades after the statute’s enactment and courts are still struggling *1107to say. Cf. United States v. Castleman, — U.S.-, 134 S.Ct. 1405, 188 L.Ed.2d 426 (2014); United States v. Serafin, 562 F.3d 1105, 1110-14 (10th Cir.2009). And then there’s the question posed by this case: What is the statute’s proper unit of prosecution? The parties before us agree that Philbert Rentz “used” a gun only once but did so “during and in relation to” two separate “crimes of violence” — by firing a single shot that hit and injured one victim but then managed to strike and kill another. In circumstances like these, does the statute permit the government to charge one violation or two?
This circuit and virtually every other has held that for each separate § 924(c)(1)(A) charge it pursues the government must prove a separate crime of violence or drug trafficking crime.1 The government admits this burden and no one asks us to revisit it. But what about the statute’s discussion of uses, carries, and possessions? Must the government also prove a separate one of those for each separate § 924(c)(1)(A) charge it brings? Though they vary in their approach to the question, the Second, Fifth, Seventh, and D.C. Circuits all seem to suggest the answer is yes, while the Eighth Circuit has (but only arguably, as we shall see) said no.2 We agreed to hear this case en bane to answer the question for ourselves.
The answer is consequential. If Mr. Rentz properly faces only a single § 924(c)(1)(A) charge he could receive between 5 and 10 years in prison for it. See 18 U.S.C. § 924(c)(l)(A)(i)-(iii). But if, as the government contends, it may bring a second § 924(c)(1)(A) charge premised on his single use of a firearm, Mr. Rentz could face a second mandatory sentence of 25 years to life — time he must serve in addition to, not instead of, the years he must serve for his initial § 924(c)(1)(A) conviction. See id. § 924(c)(l)(C)(i)-(ii), (e)(l)(D)(ii). Neither, of course, is that the end of it. Any . and all § 924(c)(1)(A) sentences Mr. Rentz receives must themselves be served consecutively to, not concurrently with, any sentences associated with his underlying crimes of violence (assaulting the first victim, murdering the second). See id. § 924(c)(l)(D)(ii).
Cases like Mr. Rentz’s are hardly unusual. In an age when the manifest of federal criminal offenses stretches ever longer, a parsimonious pleader can easily describe a defendant’s single use of a firearm as happening “during and in relation to” multiple qualifying crimes. Like when a defendant shoots a potential witness against him— committing at once the separate crimes of murder and the killing of a witness. Wilson, 160 F.3d at 736. Or when a defendant brandishes a weapon to induce his victim to surrender a car and come with *1108him — committing in the process the crimes of car-jacking and kidnapping. Phipps, 319 F.3d at 180-81. Or when a defendant points a gun at his victim and demands that she call relatives for cash — giving rise to both an unlawful ransom demand and attempted extortion. Cureton, 739 F.3d at 1035-37. In all these circumstances and many more besides, deciding what is required to prove each § 924(c)(1)(A) charge matters greatly, determining whether the defendant will face five or ten years in prison or more like thirty years to life (all, again, on top of the time he must serve for the underlying predicate offenses).
This court has not yet clearly decided whether a separate use, carry, or possession is necessary to support each count of conviction. To be sure, panels have touched on tangential matters. In Blockburger v. United States, the Supreme Court held that the double jeopardy clause prohibits punishing a defendant for the same conduct under “two distinct statutory provisions ” unless “each provision requires proof of a fact which the other does not.” 284 U.S. 299, 304, 52 S.Ct. 180, 76 L.Ed. 306 (1932) (emphasis added). Analogizing to Blockburger, some defendants have sought to suggest that a double jeopardy problem also arises when the government seeks to punish them multiple times under a single statutory provision like § 924(c)(1)(A) for a single use of a gun resulting in two predicate crimes. Panels of this court have rejected appeals along these lines, reasoning that the defendants at hand were charged with two different predicate offenses and that each charged predicate offense required proof of a fact that the other did not. But whether and how multiple punishments under a single statutory provision like § 924(c)(1)(A) could ever pose a Blockburger double jeopardy problem are questions that simply aren’t presented in this appeal. See Sanabria v. United States, 437 U.S. 54, 70, 98 S.Ct. 2170, 57 L.Ed.2d 43 n.24 (1978) (“Because only a single violation of a single statute is at issue here, we do not analyze this case under the [Blockburger ] same-evidence test”).3
Instead, the question presented by this appeal is whether, as a matter of statutory interpretation, § 924(c)(1)(A) authorizes multiple charges when everyone admits there’s only a single use, carry, or possession. Ours is a unit of prosecution case, an inquiry into the “minimum amount of activity for which criminal liability attaches” for each charge under a single criminal statute. Cureton, 739 F.3d at 1041 (internal quotation marks omitted); see also Callanan v. United States, 364 U.S. 587, 597, 81 S.Ct. 321, 5 L.Ed.2d 312 (1961) (unit of prosecution questions concern “whether conduct constitutes one or several violations of a single statutory provision”). And, as other circuits have recognized, “[t]he absence of a [Blockburger ] Double Jeopardy problem does not end [this] inquiry.” Cureton, 739 F.3d at 1040; see also Phipps, 319 F.3d at 184, 188 n. 10 (Fifth Circuit noting that neither its previous double jeopardy cases nor ours “addressees] the precise statutory question whether § 924(c)(1) authorizes multiple convictions for a single use of a single *1109firearm based on multiple predicate offenses”).4
When seeking a statute’s unit of prosecution — -when asking what the minimum amount of activity a defendant must undertake, what he must do, to commit each new and independent violation of a criminal statute — the feature that naturally draws our immediate attention is the statute’s verb. This" comes as no surprise, of course, as the verb supplies the action or doing part of most any sentence, statutory or otherwise. See United States v. Rodriguez-Moreno, 526 U.S. 275, 279-80, 119 S.Ct. 1239, 143 L.Ed.2d 388 (1999) (in deciding what the statute defines as an offense, the “ ‘verb test’ certainly has value as an interpretive tool”); The Chicago Manual of Style § 5.97 (15th ed.2003). True, in the business of statutory interpretation we do not always bow to linguistic rules. A court’s job, after all, is to discern the statute’s meaning not grade its grammar, and sometimes a law’s meaning can be clear even when the grammar’s downright awful. But until a clue emerges suggesting otherwise, it’s not unreasonable to think that Congress used the English language according to its conventions. And in § 924(c)(1)(A) we find three relevant verbs: uses, carries, and possesses. This alone supplies some evidence that each § 924(c)(1)(A) charge must involve an independent act of using, carrying, or possessing. After all, if a law’s verb says it’s a crime to kill someone, we usually think a defendant must kill more than one person to be found guilty of more than one offense. That’s the action necessary to support each and every unit of prosecution. Section 924(c)(l)(A)’s verbs make it a crime to use, carry, or possess a firearm in certain circumstances. So reading § 924(c)(1)(A) like our homicide statute and in accord with the normal rules of statutory (and sentence) construction goes some way to suggest that every new conviction requires a new act falling into one of those three categories.
Another linguistic clue points in the same direction. Section 924(c)(1)(A) doesn’t prohibit using or carrying or possessing a gun in isolation. Nor could it for guns often may be lawfully used, carried, or possessed: the Constitution guarantees as much. Instead, the statute prohibits using or carrying a gun during and in relation to any crime of violence or drug trafficking crime, or possessing a gun in furtherance of any such crime. These adverbial prepositional phrases modify the verbs uses, carries, and possesses. See The Chicago Manual of Style, supra, §§ 5.143, 5.166; Bryan A. Garner, Garner’s Modem American Usage 879, 911 (3d ed.2009). They tell us which acts of using, carrying, or possessing Congress sought to punish — explaining that the statute doesn’t seek to make illegal all such acts, only the narrower subset the phrases specify. Simplified somewhat, the language looks like this:
*1110[[Image here]]
Visualized this way it’s hard to see how the total number of charges might ever exceed the number of uses, carries, or possessions. Just as you can’t throw more touchdowns during the fourth quarter than the total number of times you have thrown a touchdown, you cannot use a firearm during and in relation to crimes of violence more than the total number of times you have used a firearm. So it is we now have in hand a pair of textual clues, clues suggesting that each § 924(c)(1)(A) charge must involve both an act of using, carrying, or possessing and that such an act must come during and in relation to (or in furtherance of) a qualifying crime.
Verbs and adverbial modifiers usually work in exactly this way. Consider another example. Imagine Congress passed a statute that in terms of its syntax is nearly identical to § 924(c)(1)(A), a statute imposing special penalties for “any person who, during a holiday, murders another.” In a year when Hanukkah and Christmas happen to fall on the same day, a man is arrested for shooting a single victim. Does the statute allow the government to bring two charges because of the calendar’s curiosities? Without some further direction from the legislature, we wouldn’t think so. We wouldn’t because murders is the statute’s operative verb and it focuses our attention on how many times the defendant performed that act — not on how many holidays that act happened to fall. The adverbial prepositional phrase acts only to impose additional limits on the availability of a charge, beyond what the verb already does, suggesting that for each charge there must be both a murder (use) and one that occurs during a holiday (during and in relation to any crime of violence or drug trafficking crime).
While this reading of the statute — like most good ones — flows from plain old grade school grammar, the government’s contrary interpretation requires some sophisticated syntactical somersaults. To bring an initial § 924(c)(1)(A) charge the government admits it must prove an act of using, carrying, or possessing that comes during and in relation to (or in furtherance of) a qualifying crime. But after that, the government says, it should be entitled to pile on additional § 924(c)(1)(A) charges without proving any further uses, carries, or possessions. A single use, carry, or possession can thus give rise to one or one hundred counts. The only limiting factor the government sees is the number of *1111qualifying crimes the prosecutor can describe as having occurred during and in relation to (or in furtherance of) that initial act of using, carrying, or possessing. Under the government’s reading, an act of using, carrying, or possessing is part of the government’s burden for the first conviction. The statute’s verb counts then. But the government would have us ignore that requirement for any additional charges. In later charges our focus would be limited myopically to the sentence’s adverbial phrases without even a stolen glance at the verbs those phrases modify. A sort of appearing and then disappearing elemental burden, a world in which verbs vanish but their modifiers float freely and commandingly alone. We don’t doubt that Congress could have found the words to accomplish this result — or even that it could have done so pretty easily. But as it stands, we just don’t see anything in § 924(c)(1)(A) that obviously does so much.5
Looking beyond the specific language Congress used in § 924(c)(1)(A) to take in a broader view of the statute does nothing to alter our initial impressions. In § 924(c)(1)(C) we encounter the punishments Congress specified for second and subsequent convictions, a provision that speaks directly to cases involving multiple lawful convictions and so might naturally appear to be a source of guidance. As it turns out, it, too, contains nothing to suggest that the government’s burden morphs between an initial charge and later ones, no hint that our focus shifts from the verbs and their modifiers to their modifiers alone. Instead, the subsection simply specifies certain punishments for a “second or subsequent conviction under this subsection” — language that, if anything, seems to anticipate that the government’s burden remains constant between initial and subsequent convictions.
Neither is this the only clue § 924(c)(1)(C) yields. The provision proceeds to impose mandatory sentences for second convictions many times higher than those associated with first convictions. As we’ve seen, a defendant can expect to receive between 25 years and life for a second conviction compared with between 5 and 10 years for most initial convictions. This massive sentencing differential suggests a legislative judgment that second violations are something different in kind from initial violations. Such a difference is obvious enough if a second conviction requires the defendant to make a second choice to use, carry, or possess a gun to further a crime — say, by firing a gun at different people on different occasions. In cases like that, the defendant hasn’t made one intentionally bad decision but two. But if a second conviction doesn’t require a second blameworthy choice to use, carry, or possess a firearm in aid of a predicate act, the logic behind the leap in punishment becomes less apparent. For, as *1112we’ve seen, a single shot can be easily described as assault and witness tampering of the same victim; a single brandishing can be both kidnapping and robbery of a lone victim; the examples could go on and on. The severity of the punishment associated with second convictions seems to suggest that Congress saw a second charge less as some sort of discretionary pleading choice or a nearly foregone conclusion in every case and more a highly unusual act, one wholly different from and more reprehensible than an initial violation. Yet it’s pretty hard to reconcile the government’s interpretation with this textually manifest judgment. See, e.g., Cureton, 739 F.3d at 1045 (describing the “statute’s purpose” as punishing “the choice to use or possess a firearm in committing a predicate offense”) (emphasis added); see also Phipps, 319 F.3d at 187 (the statute is “aimed ... at the choice to use a firearm during and in relation to a predicate offense”) (emphasis added).6
The government replies that the legislative history compels its contrary understanding of the statute. This may be a shaky enough claim on which to stake any statutory interpretation in the face of adverse textual evidence. But however that may be, the government’s legislative history proves nothing even on its own terms. The government points only to a single paragraph in a committee report prepared in 1984 when Congress amended the statute. That paragraph says just this:
The Committee has concluded that subsection 924(c)(1) should be completely revised to ensure that all persons who commit Federal crimes of violence, including those crimes set forth in statutes which already provide for enhanced sentences for their commission with a dangerous weapon, receive a mandatory sentence, without the possibility of the sentence being made to run concurrently with that for the underlying offense or for any other crime and without the possibility of a probationary sentence or parole.
S.Rep. No. 98-225, at 313 (1983) (footnote omitted), reprinted in 1984 U.S.C.C.A.N. 3182, 3491. As this language clearly explains, the 1984 amendments added a new statutory subsection (§ 924(c)(l)(D)(ii)) to ensure that § 924(c)(1)(A) sentences are served consecutively to, not concurrently with, sentences for the defendant’s underlying offense. Neither the amendment effecting that result nor the committee report discussing it had any occasion to address the statute’s proper unit of prosecution — or even the language of § 924(c)(1)(A) itself.
The government replies that the phrase “who commit Federal crimes of violence” proves that the number of convictions depends only on the number of crimes of violence a prosecutor can charge. But here (again) the government would have us read into a sentence fragment a good deal more than it can bear. Under any plausible view about the proper unit of prosecution a defendant convicted under § 924(c)(1)(A) has “commit[ted] a Federal crime[] of violence” and will receive a sentence consecutive to his underlying crime of violence sentence, just as the committee report says he should. Both Mr. Rentz and the government agree on that much. The question we face in this case is how many § 924(c)(1)(A) sentences *1113Mr. Rentz and those like him must face. And that’s a question the 1984 amendment and the government’s legislative history-clipping simply don’t address.
Perhaps the government means to suggest that because the 1984 amendment and its legislative history display an intent to punish harshly in certain respects (mandating consecutive rather than concurrent sentences) we should interpret the statute harshly in other respects too (including when it comes to defining the unit of prosecution). But if that’s the argument, it’s a poor one. Legislation is compromise and it’s rare to find a statute that pursues a single purpose unrelentingly. Assuming that whatever seems to further a statute’s perceived purpose must be the law commits the fallacy of overgeneralization. See United States v. Smith, 756 F.3d 1179, 1191 (10th Cir.2014); Bd. of Governors v. Dimension Fin. Corp., 474 U.S. 361, 374, 106 S.Ct. 681, 88 L.Ed.2d 691 (1986) (“Invocation of the ‘plain purpose’ of legislation at the expense of the terms of the statute itself ... prevents the effectuation of congressional intent.”). Indeed, from the fact that Congress wished § 924(c)(1)(A) sentences to run consecutively rather than concurrently one could as easily draw the very opposite conclusion the government does — surmising that Congress thought consecutive sentences appropriate precisely because each charge arises from something as weighty as an independent intentional choice to use, carry, or possess a gun. Certainly other circuits that have scrutinized the legislative history have found it somewhere between utterly unfathomable and modestly more helpful to the defense than the government. See, e.g., Phipps, 319 F.3d at 187; United States v. Anderson, 59 F.3d 1323, 1327-28 (D.C.Cir.1995) (en banc).
This observation leads to another. To the extent any ambiguity remains at this point about the meaning of § 924(c)(1)(A) — -after we have exhausted all the evidence of congressional meaning identified by the parties — we don’t default to the most severe possible interpretation of the statute but to the rule of lenity. United States v. Bass, 404 U.S. 336, 347-49, 92 S.Ct. 515, 30 L.Ed.2d 488 (1971). Our job is always in the first instance to follow Congress’s directions. But if those directions are unclear, the tie goes to the presumptively free citizen and not the prosecutor. Here that means the government must prove both a use, carry, or possession as well as a qualifying crime. The rule of lenity seeks to ensure legislatures, not prosecutors, decide the circumstances when people may be sent to prison. It seeks to ensure, too, that if a legislature wishes to attach criminal consequences to certain conduct — to deprive persons of their property, liberty, or even lives — it provides fair warning. Of course, Congress is free if it wishes to amend § 924(c)(1)(A) to state that a second conviction doesn’t require a second use, carry, or possession. But unless and until it does, we will not relegate men and women to prison (or to decades more time in prison) because they did something that might — or might not — -have amounted to a violation of the law as enacted. “In our legal order it is not the job of independent courts to bend ambiguous statutory subsections in procrustean ways to fit the prosecutor’s bill.” Smith, 756 F.3d at 1191; see also Orin S. Kerr, A Rule of Lenity for National Security Surveillance Law, 100 Va. L.Rev. 1513, 1533-34 (2014).
Tellingly we think, the Supreme Court has applied the rule of lenity in highly similar circumstances. In Bell v. United States, 349 U.S. 81, 75 S.Ct. 620, 99 L.Ed. 905 (1955), the Court faced a statute establishing federal penalties for “[wjhoever knowingly transports in interstate or foreign commerce ... any woman or girl for the purpose of prostitution.” The question presented was whether a single transpor*1114tation of multiple women amounted to one violation of the statute or many. The Court held the statute ambiguous, applied the rule of lenity, and.ruled that a single violation had occurred. In Ladner v. United States, 358 U.S. 169, 79 S.Ct. 209, 3 L.Ed.2d 199 (1958), the statute at issue prohibited assaults against law enforcement officers. The Court faced a single assault (a single discharge of a weapon) that happened to hurt two officers. Here again the Court found ambiguity as to the proper unit of prosecution, applied the rule of lenity, and held one violation existed. We fail to see how the Supreme Court could have found for the defendants in Bell and Ladner and this court might yet fairly find against Mr. Rentz.
If anything, the case for lenity seems even more persuasive in our case than it was in Bell or Ladner. While the government tries to persuade this en banc court that the number of § 924(c)(1)(A) charges depends only on the number of predicate crimes it can conjure and not the number of uses it can prove, in another en banc case before another circuit court the government argued exactly the other side of the debate. In Anderson the government contended that the number of uses alone, not the number of predicate crimes, limits the number of available § 924(c)(1)(A) charges. See 59 F.3d at 1325. In highlighting this pretty-difficult-to-ignore flip-flop we do not mean to suggest the government or any other party isn’t permitted to amend its view over time. The willingness to change your mind is often an admirable trait. Neither do we mean to suggest that the existence of different views about how best to untangle a statute’s knots necessarily and by itself proves the law is ambiguous. Instead, we suggest only that if a federal criminal statute is so enigmatic that the government has experienced such difficulty settling on its meaning maybe that goes some way toward showing that ordinary citizens lack reasonable notice about its meaning and the time has come to let lenity in.
Most other circuits to have come this way before us have reached the same destination we do. See, e.g., Finley, 245 F.3d at 206; Cureton, 739 F.3d at 1044; Wilson, 160 F.3d at 749-50; Phipps, 319 F.3d at 185-88. Neither does the one circuit that most apparently seems to have gone a different way (the Eighth) cause us to doubt our path. In fact, it’s not even clear that court would disagree with anything we’ve said. In Sandstrom, the Eighth Circuit did allow multiple charges premised on a single gun use to proceed under § 924(c)(1)(A). But it did so only after relying on our own panel decisions rejecting Blockburger-tyge double jeopardy challenges to multiple § 924(c)(1)(A) convictions. As we’ve seen, those cases do not directly answer the question what, as a matter of statutory interpretation, the government must prove for each successive charge under a single statute. See supra note 3. When presented with that question squarely, it remains entirely possible the Eighth Circuit would agree with us and most everyone else.7
*1115We don’t mean to suggest we think we’ve somehow cracked § 924(c)’s code. Even now plenty of hard questions remain. Not least this: what exactly suffices to constitute a unique and independent use, carry, or possession sufficient to support a second or successive § 924(c)(1)(A) charge? In light of our holding that each charge requires an independent use, carry, or possession, that question will no doubt take on particular prominence in the future. We fully expect parties to dispute when one use, carry, or possession ends and another begins. Some courts already do. Compare Finley, 245 F.3d at 206-08 (suggesting that a single possession existed over a continuous period of time), with Phipps, 319 F.3d at 188 n. 11 (suggesting this might count as two acts of possession); United States v. Cejas, 761 F.3d 717, 729 (7th Cir.2014) (upholding two convictions for two acts of possession separated by nearly a week). But to resolve this appeal, we don’t have to wade into such troubled waters. Throughout this appeal both sides have proceeded on the premise that this ease involves only one use, carry, or possession of a firearm. That much is shared ground. And to know that much is to know the government may seek and obtain no more than one § 924(c)(1)(A) conviction. Resolving that question is work enough for the day and tomorrow’s questions may be left to tomorrow. The panel opinion is vacated and the district court is affirmed.
. See, e.g., United States v. Lindsay, 985 F.2d 666, 672-74 (2d Cir.1993); United States v. Diaz, 592 F.3d 467, 470-75 (3d Cir.2010); United States v. Privette, 947 F.2d 1259, 1262-63 (5th Cir.1991); United States v. Henry, 878 F.2d 937, 942-45 (6th Cir.1989); United States v. Cappas, 29 F.3d 1187, 1189 (7th Cir.1994); United States v. Smith, 924 F.2d 889, 894-95 (9th Cir.1991); United States v. Henning, 906 F.2d 1392, 1398-99 (10th Cir.1990); United States v. Ross, 920 F.2d 1530, 1538-39 (10th Cir.1990); United States v. Johnson, 977 F.2d 1360, 1377 (10th Cir.1992); United States v. Hamilton, 953 F.2d 1344, 1345-46 (11th Cir.1992). But see United States v. Camps, 32 F.3d 102, 106-09 & n. 9 (4th Cir.1994) (suggesting otherwise but also suggesting “that a defendant may not be punished multiple times for simultaneous firearm use or carriage” no matter how many predicate offenses are involved).
. Compare United States v. Finley, 245 F.3d 199, 206-08 (2d Cir.2001); United States v. Phipps, 319 F.3d 177, 186-89 (5th Cir.2003); United States v. Cureton, 739 F.3d 1032, 1043 (7th Cir.2014); United States v. Wilson, 160 F.3d 732, 748-49 (D.C.Cir.1998), with United States v. Sandstrom, 594 F.3d 634, 656-59 (8th Cir.2010).
. For examples of panel decisions rejecting analogies to Blockburger, see, e.g., United States v. Chalan, 812 F.2d 1302, 1315-17 (10th Cir.1987); United States v. Abreu, 962 F.2d 1425, 1433 (10th Cir.1992); United States v. Floyd, 81 F.3d 1517, 1527 (10th Cir.1996); United States v. Renteria, 720 F.3d 1245, 1256 (10th Cir.2013). Other panel decisions of this court have purported to address the meaning of § 924(c)(1)(A) itself but often in the context of plain error review and then only briefly and still influenced by a sort of analogy to Blockburger's double jeopardy inquiry. See, e.g., United States v. Malone, 222 F.3d 1286, 1292-94 (10th Cir.2000); United States v. Bairett, 496 F.3d 1079, 1095-96 (10th Cir.2007).
. That's not to suggest that a decision about the proper unit of prosecution as a matter of statutory interpretation never bears constitutional double jeopardy implications. It goes without saying, for example, that once a defendant is acquitted of an offense he can’t be retried on the same offense. So deciding as a matter of statutory interpretation the scope of what it takes to prove one or more offenses under a single statute naturally may affect whether retrial is or isn't constitutionally permissible in particular cases. See, e.g., Sanabria, 437 U.S. at 69-70, 98 S.Ct. 2170. Still and again, though, our inquiry in this appeal isn’t into potential knock-on double jeopardy consequences; it is into the meaning of the statute itself.
. Our concurring colleague Judge Matheson notes that folded within the statute's adverbial prepositional phrase lies the word "any”: the statute penalizes using, carrying, or possessing a gun during "any” crime of violence or drug trafficking crime. Our colleague considers the possibility this word might help the government’s cause, but ultimately concludes that if it does help it doesn’t help much. Matheson Concurrence at 1109-11. We agree for reasons by now clear. The word "any” doesn't float freely — say, punishing "any” crime of violence or drug trafficking crime or "any” use of a gun. Instead, it appears in an adverbial prepositional phrase whose job is to narrow the universe of uses, carries, or possessions criminalized by § 924(c), focusing our attention on only those uses, carries, or possessions that occur during "any” of certain specified conditions. So "any” doesn't tell us anything about the number of uses, carries, or possessions required to justify each independent charge — just about the sorts of uses, carries, or possessions that violate the statute.
. Judge Matheson examines a separate sentencing provision that applies when the gun used is a machine gun (18 U.S.C. § 924(c)(l)(B)(ii)) but fails to find any definitive support there for the government either, and again we agree. See Matheson Concurrence at 1110-11. The fact that a particular sentencing enhancement follows if you use, .carry, or possess a machine gun during or in relation to a crime of violence or drug trafficking crime doesn't speak one way or the other to the question whether an independent use, carry, or possession' is required for each such special enhancement.
. The government says the case for lenity so many circuits have found persuasive should be reexamined in light of Rosemond v. United States, - U.S. -, 134 S.Ct. 1240, 188 L.Ed.2d 248 (2014). But in Rosemond the Court addressed the elements of aiding and abetting liability under § 924(c)(1)(A), not what must be proven for principal liability. The government acknowledges this but says we should take a note from the Court's description of § 924(c)(1)(A) as a “combination crime,” one that “punishes the temporal and relational conjunction of two separate acts, on the ground that they together pose an extreme risk of harm.” Id. at 1248. Our dissenting colleague echoes this argument, reading Rosemond to "suggest[ ] that the unit of prosecution is based upon pairing the crime of violence with the use of the firearm.” Dissent at 3. We of course agree with these *1115observations but fail to see how they decide our case. Mr. Rentz no less than the government agrees that § 924(c)(1)(A) punishes the conjunction — the pairing, if you like — of a gun use and a qualifying crime. The question isn't whether § 924(c)(1)(A) is a combination crime but the nature of that combination, more particularly whether there must be a combination of an independent use/carry/possession and an independent crime of violence or drag trafficking offense for each § 924(c)(1)(A) charge. On that, the dispositive question in our case, Rosemond offers not a hint, let alone suggests that the statute unambiguously favors the government's position in this case.