dissenting:
It is hard to say whether a medium or a fortune-teller would be best suited for this case. The Court’s opinion seems to be one part Back to the Future, requiring trial counsel to possess a DeLorean, a flux capacitor and the inventiveness of the fictional Doc Brown in order to render competent assistance to a client. But it is another part Zombie Apocalypse, intent on disinterring the grisly remains of the long defunct approach of purposive judicial interpretation. Because neither approach is consistent with precedent and our judicial responsibilities, I respectfully dissent.
It seems silly to recite again the familiar ineffective-assistance standard of Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), but circumstances demand clarity. Under the Strickland framework, a defendant must first establish “counsel made errors so serious that counsel was not functioning as the ‘counsel’ guaranteed the defendant by the Sixth Amendment.” Id. at 687, 104 S.Ct. 2052. If so, the defendant must then demonstrate the deficient performance was prejudicial by showing “there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have-been different. A reasonable probability is a probability sufficient to undermine confidence in the outcome.” Wiggins v. Smith, 589 U.S. 510, 534, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003) (quoting Strickland, at 694, 104 S.Ct. 2052).
Today, the Court replaces this familiar two-part test with a one-part test of its own making. Under this new version of Strickland, an attorney’s conduct is ineffective if the attorney failed to pursue an action that “had a reasonable likelihood of success.” Maj. Op. at 1088. The Court is open about this elision, calling this “the rare case where both Strickland prongs turn on the same question, whether there is a reasonable probability that the outcome of [Abney’s sentencing] would have been different had this issue been raised.” Maj. Op. at 1088.
Examination reveals this novel conception of the Strickland standard has a limited lineage and, in our circuit, a common progenitor. Only three circuit court opinions have ever stated that the two prongs of Strickland operate, in rare instances, as one. See Etherton v. Rivard, 800 F.3d 737 (6th Cir.2015), petition for cert. pending, Docket No. 15-723; Payne v. Stansberry, 760 F.3d 10 (D.C.Cir.2014); and Roe v. Delo, 160 F.3d 416 (8th Cir.1998). The idea originated in dicta in Roe, a case in which counsel failed to raise a plain-error claim on appeal despite what the court found to be a reasonable possibility that such a claim would succeed under the circumstances. See 160 F.3d at 420. Counsel could not recall why the claim had been omitted and stated that, whatever the reason, it was not a strategic decision. Id. at 418-19.
This Court picked up the idea sixteen years later, in an opinion authored by the same judge writing for the majority today. Payne, 760 F.3d 10 (D.C.Cir.2014). Citing only Roe’s diktat for support, this Court professed to have found another “rare case” in which the two prongs of Strickland mean the same thing. Payne, 760 F.3d at 14. The Court recycles the same *1096language today — raising the curious question of how the “rare” cases in which the Strickland standard conveniently reduces itself to a one-part efficacy test has occurred twice in little more than a year. Is there now a reasonable possibility this rarity will soon become routine?
The single-prong Strickland test has a shallow provenance for good reason. It is an elision of an important constitutional test that results in substantively different outcomes. The Court imports the reasonable probability standard of Strickland’s second part into its first part. Under this novel formulation, a defendant who satisfies the second, prejudice prong of Strickland will always satisfy the first, objective-standard prong of Strickland, because whether an attorney’s conduct fell below a professional standard depends on whether the attorney took every step that had some reasonable possibility of success for the client. By collapsing Strickland into this single step, the Court transforms the objective standard of Strickland into a retroactive assessment of the “rightness” of a defendant’s outcome. Defendants who could have had better outcomes if their counsel had made different arguments will, by inference, have received ineffective assistance. Under the Court’s approach, a lawyer is obligated to pursue any action that might' reasonably benefit the client. Failing to do so constitutes unreasonable attorney conduct, and hence, it amounts to constitutionally ineffective assistance.
This is not the approach endorsed by Strickland. Strickland’s second prong is about whether a defendant was prejudiced by his counsel’s ineffective assistance. But a court may reach the second prong only if counsel’s performance is adjudged ineffective under an objective standard. That’s because the second prong of the Strickland test is remedial in nature — it’s not about whether there was a constitutional violation but about what we do after we know there was a constitutional violation. If a defendant’s outcome would have been the same, we say the constitutional violation was harmless. If there is a “reasonable probability” the outcome would have been different, then we take some type of action to remedy the constitutional problem.
Today the Court says the existence of a “reasonable probability” of a different outcome actually answers the first prong of Strickland: whether there was a constitutional violation in the first place. Never mind that prong one of Strickland has its own standards for determining whether a constitutional violation occurs — standards that focus entirely on the realm of reasonable legal strategies, not the range of preferable legal outcomes. Strickland has always been about acknowledging the strategic aspect of lawyering. It is a commonplace observation that a lawyer simply cannot predict what will and will not work. Defendants have no right to counsel with flawless judgment; rather, the purpose of the right to effective counsel “is simply to ensure that criminal defendants receive a fair trial.” Id. at 689, 104 S.Ct. 2052.
The worst part of the Court’s new, amalgamated Strickland standard is that it doesn’t work. Trial counsel often face multiple strategic options, each presenting a reasonable possibility of benefitting the client’s interests, but also presenting a reasonable possibility of harming the client’s interests. For example, in most circumstances, when defense counsel considers whether to present a witness, there is a reasonable possibility the witness will benefit the client’s case, but also some possibility the witness will be a dud or even diminish the client in the jury’s eyes. If we second-guessed all of counsel’s decisions on the basis of whether they might have had a reasonable possibility of chang*1097ing an outcome, then nearly any decision that doesn’t pan out would amount to ineffective assistance. And that is exactly what Strickland says we shouldn’t do: “Judicial scrutiny of counsel’s performance must be highly deferential,” and “every effort [must] be made to eliminate the distorting effects of hindsight, to reconstruct the circumstances of counsel’s challenged conduct, and to evaluate the conduct from counsel’s perspective at the time.” Strickland, 466 U.S. at 689, 104 S.Ct. 2052.
So applying the Strickland standard in the right order, we should ask whether Abney’s counsel rendered ineffective assistance in the first place — whether it was objectively unreasonable for counsel not to seek a continuance of Abney’s sentencing on the basis of the pending Fair Sentencing Act.
To start, we know Abney’s 'counsel was aware of the Fair Sentencing Act, which was then awaiting the President’s signing decision. Abney’s counsel noted at sentencing that “new penalties will soon be in place” that would leave Abney “much more harshly punished than those committing and convicted of the same crime in the near future.” J.A. 18. But Abney’s counsel did not think Abney could benefit from the new law’s more lenient penalties “absent subsequent legislation making [the Fair Sentencing Act] retroactive.” J.A. at 25. In other words, Abney’s counsel was aware of the developing legal changes, and he considered the law’s applicability to Abney. Cf. Torres v. United States, 2013 WL 1349126, at *3 (S.D.W.Va. Apr. 1, 2013) (finding counsel sufficient despite no evidence counsel actually knew of the Fair Sentencing Act’s impending enactment). Counsel concluded, however, that it would take some form of retroactivity to make Abney’s pre-Act crimes eligible for post-Act penalties, a view of the Act that both the government and the district court shared.
Congress enacted the Fair Sentencing Act of 2010, Pub. L. No. 111-220, 124 Stat. 2372, without any explicit provision addressing whether and how pre-Act offenders would be treated under the law. The language of the Act was forward-looking. Even the Court acknowledges today that “[t]he retroactive effect of the [Fair Sentencing Act] was ambiguous” as enacted. Maj. Op. at 1089.
After President Obama signed the Act into law, courts confronted the question of the Act’s possible retroactivity. This Court agreed with every other circuit court to address the issue: “there [was] simply ‘no evidence that Congress intended the [Fair Sentencing Act] to apply to defendants who had been sentenced prior to the August 3, 2010 date of the Act’s enactment.” United States v. Bigesby, 685 F.3d 1060, 1066 (D.C.Cir.2012). For defendants who committed their crimes before that date but were sentenced after enactment, however, the results were more mixed.
Three circuits concluded defendants who committed their crimes before the Act took effect could not take advantage of its lower sentences, even if they were sentenced after the Act was implemented. United States v. Fisher, 635 F.3d 336, 339-340 (7th Cir.2011) (holding Act does not apply to those who committed their crimes before August 3, 2010); United States v. Sidney, 648 F.3d 904, 910 (8th Cir.2011) (same); United States v. Tickles, 661 F.3d 212, 215 (5th Cir.2011) (per curiam) (same). In his statement respecting the denial of en banc rehearing, Judge Easter-brook eloquently explained the reasoning behind this view. United States v. Holcomb, 657 F.3d 445 (7th Cir.2011). Prior to its decision in Dorsey, the Supreme Court “never held any change in a criminal *1098penalty to be partially retroactive.” Id. at 446. Retroactivity had historically been an all-or-nothing proposition. Criminal statutes with retroactive application applied to all individuals regardless of the date on which they committed their crimes, while prospective statutes applied only to individuals who committed their crimes after the effective date of the new legislation. See id. This understanding of retroactivity did not depend on the date of sentencing, “which reflects how long it took to catch a criminal, and the state of the district judge’s calendar, rather than principles of deterrence or desert.” Id. at 447.
Three other circuits concluded that pre-Act offenders could take advantage of the Act’s lower mandatory mínimums. These circuits adopted a “partial retroac-tivity” approach, in which the Act applied retroactively to those who committed their offenses before its enactment but were sentenced afterwards. These “partial retroactivity” circuits, however, could not agree on which date would make the critical difference for sentencing purposes. The First Circuit held the Act’s new mandatory mínimums applied only to defendants sentenced after the new sentencing guidelines went into effect on November, 1, 2010. United States v. Douglas, 644 F.3d 39, 46 (1st Cir.2011). But in the Third and Eleventh Circuits, defendants received the benefit of the Act’s new mínimums if they were sentenced on or after August 3, 2010. United States v. Dixon, 648 F.3d 195, 203 (3d. Cir.2011); United States v. Rojas, 645 F.3d 1234 (11th Cir.2011).
“In light of [this] disagreement” among the circuit courts of appeals, the Supreme Court took up the Act’s retroactivity question, and in a 5-4 decision, held “the Fair Sentencing Act’s new, lower mandatory mínimums apply to the post-Act sentencing of pre-Act offenders.” Dorsey v. United States, — U.S. -, 132 S.Ct. 2321, 2335, 183 L.Ed.2d 250 (2012). The Court itself acknowledged the partial-retroactivity interpretation was not obvious. Rather, the “timing issue ... is difficult in part because relevant language in different statutes argues in opposite directions.” 132 S.Ct. at 2330. Five members of the Court resolved the statutory chaos through a matrix of “six considerations, [which] taken together, convince us that Congress intended the Fair Sentencing Act’s more lenient penalties to apply to those offenders whose crimes preceded August 3, 2010, but who are sentenced after that date.” Id. at 2331.
Given that background, it is impossible to say Abney’s counsel performed incompetently by not asking for a continuance of Abney’s sentencing. At the time Abney’s counsel decided whether to seek a continuance, no court had yet enunciated the theory of partial-retroactivity. Even when that theory emerged, federal courts divided over whether it was a proper interpretation of the Act. Moreover, the Supreme Court was itself divided over the theory, by a 5-4 vote. And even for those justices in the majority, the partial retroactivity interpretation was, at best, a means of sorting out a tangle of statutory language pulling in different directions. In contrast to what the Court says, then, counsel here neither ignored established guidelines, United States v. Soto, 132 F.3d 56, 59 (D.C.Cir.1997) (holding counsel was ineffective for ignoring a relevant Guideline provision), nor failed to demonstrate familiarity with settled legal principles, Thomas v. Varner, 428 F.3d 491, 501 (3rd Cir.2005) (stating courts routinely declare counsel ineffective when counsel fails to present a strong defense .because of unfamiliarity with clearly settled legal principles), nor ignored established case precedent inter*1099preting guidelines, United States v. Gaviria, 116 F.3d 1498, 1512 (D.C.Cir.1997).
Here, counsel did not face “two reasonably likely but uncertain readings of the [Fair Sentencing Act],” as the Court insists. Maj. Op. at 1089. Counsel had every reason to believe retroactivity would continue to be the all-or-nothing concept-it had always been — retroactive or prospective, but nothing in between. See Holcomb, 657 F.3d at 446 (statement of Easterbrook, J.). It is irrelevant, as the Court states, that before Dorsey, some “defense attorneys argued — and some courts agreed” that the Fair Sentencing Act was susceptible to a novel, partial-retroactivity interpretation. Maj. Op. at 1089. Those arguments arose well after the time by which Abney’s counsel would have had to decide whether to seek a continuance. Counsel’s legal judgment cannot be impugned on the basis of creative arguments that were yet to be made.
The Court asserts the outcome of Dorsey was “reasonably likely,” and therefore counsel should have taken steps to ensure that his client’s interests were protected if this “reasonably likely” interpretation ultimately carried the day. Yet the analysis offers no support for its view that the Dorsey result was “reasonably likely” aside from its own say-so. The best the court can muster is the fact that some other defense counsel were seeking continuances at this time. This assertion, however, only demonstrates that some attorneys thought the date of sentencing might be of subsequent importance, not that all competent attorneys would think so. A few motions hardly serve as the kind of professional norm contemplated in Strickland. See 466 U.S. at 688, 104 S.Ct. 2052. Furthermore, we have no comparison of the facts of these other cases. We do not know if these other defendants had also been the source of long sentencing delays, as Abney had been, or whether they had also committed their crimes long before the Act was even out of committee. We cannot measure Abney’s counsel’s conduct against the standard of these other defense attorneys without knowing that those attorneys found themselves in substantively similar positions.
Not only could Abney’s counsel not have reasonably predicted the Court’s partial-retroactivity approach in Dorsey, he was also laboring under the limitations of his client’s case — which gave him sound, strategic reasons for not seeking a continuance. Under ordinary circumstances, a short continuance might be a small ask, but Abney’s was no ordinary case: by the time of his scheduled sentencing hearing, he had already been avoiding sentencing for nearly three years. After initially facing a ten-year mandatory minimum sentence, Abney had entered a plea agreement with the government in which he agreed to provide investigative assistance in exchange for his release with supervision. See Appellee Br. 3-4. Thus, he could, through his own initiative, dramatically reduce his sentence. But Abney could not bring himself to fulfill the conditions of his release: he tested positive for cocaine and marijuana; he “waterloaded” his urine samples in an effort to thwart his mandatory drug testing; he failed to appear for drug tests; and, ultimately, he lost contact altogether with supervising law enforcement. Above all of this, the government never received any cooperation from Abney. Realizing that they had been duped, the government obtained a sentencing date of October 13, 2009. Ab-ney again managed to avoid sentencing, this time because Maryland authorities arrested and detained him on charges of attempted first-degree murder. J.A. 11. Abney’s subsequent incarceration -in Maryland forced the district court to postpone his sentencing twice. J.A. 14.
*1100So it was that Abney arrived at his August 2, 2010, sentencing hearing, having already spent nearly three years — much of it as a free man — avoiding the imposition of his sentence. Indeed, it is a mere accident of history that Abney’s Fair Sentencing Act near-miss is a near-miss at all. Had he been sentenced according to a more conventional schedule, he would have needed a continuance of at least many months, and likely years. A continuance of that kind would certainly not have been granted. See, e.g.,. United States v. Fields, 699 F.3d 518, 523 (D.C.Cir.2012) (trial court did not abuse its discretion in denying postponement while Act was pending because pending legislation was too far off to be compelled consideration at sentencing) (internal citation omitted).
The problems with the Court’s approach do not stop there. Even if Abney’s counsel had suspected partial retroactivity would become the rule, on what date would retroactivity take effect? Would counsel have needed a continuance of ten days, enough time to allow the law to be signed? Or a few months, to allow time for new sentencing guidelines to be written and go into effect? See United States v. Thompson, 721 F.3d 711, 715 (D.C.Cir.2013) (affirming district court’s denial of a continuance and stay of sentencing several months prior to the Act’s passage); Fields, 699 F.3d at 521 (same). Even the Court does not know how long the Constitution would require competent counsel to continue the proceedings. Maj. Op. at 1092 (endorsing Abney’s acknowledgment that whether August 3 or November 1 would be the relevant date was unclear, and yet holding such confusion cannot “validate counsel’s failure” to seek a continuance of indeterminate length).
Even were we to accept that Abney’s counsel performed ineffectively, Abney would still need to satisfy Strickland’s second prong by showing “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. Because the effect of Dorsey is to lessen the sentence of defendants sentenced on or after August 3, 2010, the prejudice question turns on whether Abney would have gotten a continuance had his lawyer sought one.
Continuances fall within the discretion of the trial judge, United States v. Burton, 584 F.2d 485, 490 (D.C.Cir.1978), and are reviewed “only to determine whether the judge clearly abused his discretion.” United States v. Gantt, 140 F.3d 249, 256 (D.C.Cir.1998). Review must assess “the circumstances in every case, particularly in the reasons presented to the trial judge at the time the request is denied.” Ungar v. Sarafite, 376 U.S. 575, 589, 84 S.Ct. 841, 11 L.Ed.2d 921 (1964) (emphasis added). The Court, relying on a hypothetical, dispassionate decision-maker — a “reasonable, conscientious, and impartial district court,” Maj. Op. at 1093 — speculates that Abney’s request would have been granted. In fact, we do not need to speculate about what would have happened if Abney’s counsel had sought a continuance: the sentencing judge answered the question in his opinion below, saying he would have been unlikely to grant a continuance and further “delay reaching finality on a three-year-old plea.” J.A. 87. The district court “considered whether the Fair Sentencing Act could benefit Abney, but came to the reasonable estimation that “it was unclear that [the Fair Sentencing Act] would even apply retroactively to Abney.” Id. The district court concluded it was “completely speculative — not reasonably probable” that -a motion for continuance to some “unspecified future date” would have been granted. Id. Given the broad authority district court judges have to manage their calendars and *1101grant or deny continuances, the district court’s assessment amounts to a reasonable exercise of discretion. The Court disagrees. Even after upending the Strickland standard, the Court can only make a case for prejudice by discrediting the district court as “[unjreasonable, [unjcon-scientious, and [ jpartial.” See Maj. Op. at 1093.
But abusing the district court is still not enough. Rather than defer to the sentencing judge’s reasonable assessment of the likelihood of his granting relief, the Court takes yet another unpredictable step, relying on de novo review. Whether de novo review applies in these circumstances is a question this Circuit has left unanswered for nearly thirty years. See, e.g., United States v. Askew, 88 F.3d 1065, 1071 (D.C.Cir.1996). Today, the Court decides that question, and applying de novo review, it concludes that granting a continuance would have been likely.
At the heart of the Court’s misreading of Strickland is a misapprehension about the realities of our republican form of government. The Court seems baffled that Congress would “insist on the near-immediate reduction of the 100-to-l disparity in the Sentencing Guidelines for defendants sentenced after the [Fair Sentencing Act’s] enactment but leave in place for those same defendants the old, unfair mandatory mínimums.” Maj. Op. at 1091. Congress did not address retroactivity in the Fair Sentencing Act, in all likelihood, because it could not do so; had the Act taken up that issue, it may well have never garnered the votes necessary for passage. In our form of government, legislative compromise often produces imperfect outcomes. Here, those imperfections are most evident in the line-drawing problems caused by the new mínimums. But this should not surprise the Court. Dorsey, in fact, recognizes that there are line-drawing problems with the application of the new mínimums: pre-Act offenders who were also sentenced pre-Act remain in prison, serving lengthy sentences premised on a now-rejected cocaine-to-powder ratio. See- 132 S.Ct. at 2335. The only solution for this disparity is a congressional act making the Fair Sentencing Act retroactive, even for pre-Act offenders; absent this, there is no way to even out the ratio-based sentences being served by all offenders.
Instead, the Court attempts to smooth out some of the disparity on its own, by moving Abney from the old, discredited sentencing regime into the new one. And it’s not easy. First, the Court revises the Strickland test, from two prongs into one. Second, the Court decides to apply de novo review, forcing an answer to a question unsettled in this jurisdiction for more than thirty years. I do not doubt that in its determination to rescue Abney, the Court acts with good intentions. In Dorsey, the high court acted with similar motives when it decided to pick and choose among opposing provisions to make the Fair Sentencing Act’s more lenient provisions available to a larger number of defendants. That result might be what Congress should have adopted or would have enacted had decades of study produced sufficient votes. Instead, the Supreme Court fashioned an ad hoc measure to implement its widely-favored policy and in doing so offered neither rule nor guidance for future cases. If separation of powers is to have any meaning, the result in Dorsey cannot be deemed ordinary — certainly not so ordinary that counsel should be trained to expect it.
Here, trial counsel and the district court discussed, openly and on the record, that in order to benefit from the Fair Sentencing Act, Abney would need a legislative fix. Abney would need Congress to act in order to give him the benefit of the law. Today, the Court turns that healthy, con*1102stitutionally-grounded presumption on its head: not only did Abney not need to wait for Congress to make the Fair Sentencing Act retroactive, but his counsel was incompetent for thinking the legislature’s involvement would be required. It is no longer enough for counsel to ask whether the courts have acted, to fix some perceived problem Congress has caused; instead, counsel must presume both that courts can act and that they will. The failure to anticipate the “reasonable possibility” of judicial policymaking is proof of both counsel’s incompetence and the district court’s lack of conscientiousness — a result that seems both perverse and pernicious.