In our constitutional order legislative enactments normally apply only prospectively while judicial decisions also bear retroactive application. But what’s the rule when an executive agency exercises delegated legislative policymaking authority in what looks like a judicial proceeding? That’s the peculiar question posed by this case. It comes to us this way. Buried deep in our immigration laws lie these two provisions: 8 U.S.C. §§ 1255(i)(2)(A) and 1182(a)(9)(C)(i)(I). Enacted first, § 1255(i) grants the Attorney General discretion to “adjust the status” of those who have entered the country illegally and afford them lawful residency. But growing concerns about illegal immigration eventually induced Congress to enact § 1182(a), which appears to take away at least part of the discretion § 1255(i) gives. Among other things, § 1182(a)(9)(C) provides that certain persons who have entered this country illegally more than once are categorically prohibited from winning lawful residency here — that is, unless they first serve a ten-year waiting period outside our borders. Taken together, these provisions seem to render Alfonzo De Niz Robles and others like him simultaneously eligible and ineligible for relief.
How to make sense of this statutory tension? When confronted with the question in 2005, this court held that § 1255(i) trumped and that the Attorney General’s discretion to afford relief remained intact. Meaning that Mr. De Niz Robles and those similarly situated could file petitions for adjustment of status. Which is exactly what Mr. De Niz Robles did, filing a petition in reliance on our decision not long after it issued. See Padilla-Caldera v. Gonzales (Padilla-Caldera I), 426 F.3d 1294, 1300-01 (10th Cir.2005), amended and superseded on reh’g, 453 F.3d 1237, 1244 (10th Cir.2006).
But ours turned out to be hardly the last word. In 2007, the Board of Immigration Appeals issued In re Briones, 24 I. & N. Dec. 355 (BIA 2007). There the agency held that it’s § 1182(a)(9)(C)(i)(I) that does the trumping and, as a result, the Attorney General lacks any discretion to adjust the status of illegal reentrants like Mr. De Niz Robles. Neither is there any way to reconcile Padilla-Caldera I and Briones. By everyone’s admission, they are just implacably opposed. So whose opinion governs?
Usually, executive agencies can’t overrule courts when it comes to interpreting the law. Chi. & S. Air Lines, Inc. v. Waterman S.S. Corp., 333 U.S. 103, 113, 68 S.Ct. 431, 92 L.Ed. 568 (1948). But like most rules this one bears its exceptions and here we confront a curious one. If a statutory scheme administered by an executive agency is “ambiguous,” then “step two” of Chevron, U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984), requires this court both to assume that Congress has “delegated policy-making responsibilities” to the agency and to defer to the agency’s “policy choice[ ]” so long as that choice is “reasonably]” consistent with the legislative scheme. Id. at 865-66, 104 S.Ct. 2778. Taking the point further still, National Cable & Telecommunications Ass’n v. Brand X Internet Services (Brand X), 545 U.S. 967, 125 S.Ct. 2688, 162 L.Ed.2d 820 (2005), requires this court to defer to the agency’s policy choice even when doing so means we must overrule our own preexisting and governing statutory interpretation. Id. at 981-83, 125 S.Ct. 2688. Together, then, these decisions mean that there are indeed some occasions when a federal bureaucracy can effectively overrule a judicial decision.
That’s exactly what happened here. After the BIA handed down Briones this court acknowledged that the statutory di*1168rectives found in §§ 1255(i)(2)(A) and 1182(a)(9)(C)(i)(I) are ambiguous; that under Chevron step two this ambiguity entitled the BIA to make a reasonable quasi-legislative policy choice of its own; and that Briones qualified as just such a judgment. Accordingly and pursuant to Brand X, we held that we were obliged to discard Padilla-Caldera I and defer to the BIA’s pronouncement in Briones. Padilla-Caldera v. Holder (Padilla-Caldera II), 637 F.3d 1140, 1153 (10th Cir.2011).1
What do all these twists and turns mean for Mr. De Niz Robles? Though he filed an administrative application for an adjustment of status after and in reliance on our decision in Padilla-Caldera I, his petition languished for years. It languished so long that it still stood waiting for a decision four years later, after the BIA handed down Briones and this court decided Padilla-Caldera II. And when the BIA finally did take up Mr. De Niz Robles’s petition in 2013, it decided to apply its decision in Briones retroactively to his case, using it to hold him categorically ineligible for an adjustment of status and subject to removal.
It is this ruling Mr. De Niz Robles now asks us to overturn. In his view, Chevron step two and Brand X may mean that the BIA can apply Briones prospectively to administrative petitions filed after the date of that decision. But nothing in those decisions permits the agency to apply Briones retroactively to petitions filed before its date of decision. Indeed, Mr. De Niz Robles suggests that settled principles of due process and equal protection and the precedents that embody them preclude agencies from retroactively enforcing the new policies they announce under the authority granted to them by Chevron step two and Brand X. All of which means that, for petitions like his, filed in reliance on Padilla-Caldera I and before Briones, the BIA may not automatically deny adjustment of status but must instead afford the discretionary administrative review Padilla-Caldera I promised. Both sides agree Mr. De Niz Robles’s appeal — his challenge to the BIA’s authority to enforce Briones retroactively — raises a pure question of law we may assess de novo. See, e.g., Barrera-Quintero v. Holder, 699 F.3d 1239, 1243 (10th Cir.2012).
Having said that, the BIA denies there’s anything retroactive about its decision in this case and suggests we might simply dismiss this appeal on that basis alone. But this case is hardly so simple. A statute, order, or edict “operates retroactively” when it seeks to impose “new legal consequences to events completed before its” announcement. INS v. St. Cyr, 533 U.S. 289, 321, 121 S.Ct. 2271, 150 L.Ed.2d 347 (2001). And there’s no question the BIA order before us seeks to do just that. When he filed his administrative petition in 2007, Mr. De Niz Robles possessed two lawful options for trying to win permanent residency: he was free to seek an adjustment of status pursuant to Padilla-Caldera I or he could áccept a ten-year waiting period outside the country. Relying on circuit precedent, he chose the former option. The BIA order now before us purports to treat that option as if it never existed. In this way, it surely attaches new consequences to past conduct Mr. De Niz Robles cannot now alter. After all, had he been armed in *11692007 with the knowledge that leaving the country was his only path to lawful residency he could’ve departed then and today stand just two years away from eligibility. But that opportunity is long forgone and cannot be recaptured now: for Mr. De Niz Robles the BIA’s decision means it’s eight years lost and ten still to wait. The real question in this case, then, isn’t whether the BIA’s decision seeks to impose new legal consequences on Mr. De Niz Robles’s past conduct, but whether lawfully it may. See Garfias-Rodriguez v. Holder, 702 F.3d 504, 515-16 (9th Cir.2012) (en banc); Acosta-Olivarria v. Lynch, 799 F.3d 1271, 1274-75 (9th Cir.2015).2
When it comes to retroactivity and the law we can say a couple things with certainty. First and foremost, we know that legislation is rarely afforded retroactive effect. “It is a principle which has always been held sacred in the United States, that laws by which human action is to be regulated, look forwards, not backwards; and are never to be construed retrospectively unless the language of the act shall render such construction indispensable.” Reynolds v. McArthur, 27 U.S. (2 Pet.) 417, 434, 7 L.Ed. 470 (1829) (Marshall, C.J.); see also Landgraf v. USI Film Prods., 511 U.S. 244, 265, 114 S.Ct. 1483, 128 L.Ed.2d 229 (1994) (“[T]he presumption against retroactive legislation is deeply rooted in our jurisprudence.”). After all, legislation is powerful medicine: it usually announces a rule of general applicability and regulates otherwise private conduct. By design it’s meant to alter existing expectations and further policy and political ends, for legislators explicitly serve as policymakers and partisans. Recognizing this, and in service of the due process interests of “fair notice, reasonable reliance, and settled expectations,” the law applies a presumption that new legislation governs only prospectively. Id. at 270,114 S.Ct. 1483. This presumption serves an important equal protection interest too, preventing the state from singling out disfavored individuals or groups and condemning them for past conduct they are now powerless to change. See id. at 266, 114 S.Ct. 1483; Adrian Vermeule, Essay, Veil of Ignorance Rules in Constitutional Law, 111 Yale L.J. 399, 408 (2001).
For all these reasons and no doubt more besides, the presumption that legislation operates only prospectively is nearly as old as the common law. See, e.g., 3 Henry de Bracton, De Legibus et Consuetudinibus Angliae 530-31 (Travers Twiss ed. & trans., 1880) (1257); 1 William Blackstone, Commentaries on the Laws of England *46 (1765) (“All laws should be therefore made to commence in futuro, and be notified before their commencement.”). Indeed, the presumption is sometimes said to inhere in the very meaning of the “legisla*1170tive Powers” the framers assigned to Congress in Article I of our Constitution. See, e.g., 2 Joseph Story, Commentaries on the Constitution of the United States § 1398 (Melville M. Bigelow ed., 1994) (1833) (“[RJetrospective laws ... neither accord with sound legislation nor with the fundamental principles of the social compact.”). To overcome the presumption of pros-pectivity, the Supreme Court has held that Congress must declare unequivocally its intention to regulate past conduct — and even then due process and equal protection demands may sometimes bar its way. Landgraf, 511 U.S. at 268, 270-72, 114 S.Ct. 1483; Pension Benefit Guar. Corp. v. R.A. Gray & Co., 467 U.S. 717, 730, 104 S.Ct. 2709, 81 L.Ed.2d 601 (1984). Indeed, in at least two arenas (contract and crime) the Constitution transforms the presumption of prospectivity into a mandate, precluding any possibility of retroactive legislation. U.S. Const. art. I, §§ 9-10 (Ex Post Facto Clauses); id. § 10, cl. 1 (Contracts Clause); see also Calder v. Bull, 3 U.S. (3 Dall.) 386, 1 L.Ed. 648 (1798) (interpreting the Ex Post Facto Clauses to prohibit only retroactive criminal legislation).3
Quite the opposite from legislation (and with equal certainty) we can say that judicial decisions “have had retrospective operation for near a thousand years.” Kuhn v. Fairmont Coal Co., 215 U.S. 349, 372, 30 S.Ct. 140, 54 L.Ed. 228 (1910) (Holmes, J., dissenting). In fact, “[a]t common law there was no authority for the proposition that judicial decisions made law only for the future.” Linkletter v. Walker, 381 U.S. 618, 622, 85 S.Ct. 1731, 14 L.Ed.2d 601 (1965). Following this ancient tradition, the Supreme Court has forbidden federal courts from rendering purely prospective judicial decisions in the criminal arena, Griffith v. Kentucky, 479 U.S. 314, 328, 107 S.Ct. 708, 93 L.Ed.2d 649 (1987), and the Court will barely tolerate the practice in the civil arena, Harper v. Va. Dep’t of Taxation, 509 U.S. 86, 97, 113 S.Ct. 2510, 125 L.Ed.2d 74 (1993).4 The Court has suggested, too, that the presumption of retroactivity attaching to judicial decisions was anticipated by the Constitution and inheres in its separation of powers. See, e.g., id. at 94, 113 S.Ct. 2510.
You might wonder why the due process and equal protection concerns that counsel in favor of prospectivity in legislation don’t operate similarly when it comes to judicial decisions. The answer, we think, lies in the fact that for civil society to function the people need courts to provide backward-looking resolutions for their disputes. And accepting this premise, the Constitution has sought to mitigate the due process and equal protection concerns associated with retroactive decisionmaking in other ways, by rules circumscribing the nature *1171of the judicial function and the judicial actor. The Constitution assigns to judges the limited “judicial Power” to decide “Cases” and “Controversies.” U.S. Const, art. Ill, § 2. It invests judges with none of the “legislative Power[]” to devise new rules of general applicability, a power Article I reserves to Congress and its elected officials alone. The Constitution assigns the “judicial Power,” as well, not to avowed policymakers and politicians answerable to the people but to judges insulated from partisan influence and retribution and appointed without term. See id. § 1. It is in these ways that the Constitution and its separation of powers seek to meet a necessity of civil society while mitigating the due process and equal protection concerns sometimes associated with retroactive decisionmaking. See Harper, 509 U.S. at 94-95, 113 S.Ct. 2510.
So when it comes to Congress we know its handiwork is presumptively prospective. And when it comes to the judiciary we know its decisions are presumptively retroactive. But what does the law have to say when the decision at issue comes from an executive agency? When that agency exercises “delegated” power from Congress to make a “policy” decision, an authority granted to it by Chevron step two? In, no less, an “adjudication” that effectively overrules a judicial decision, a power the agency enjoys thanks to Brand X?
The Constitution speaks far less directly to that peculiar question. Perhaps because the framers anticipated an Executive charged with enforcing the decisions of the other branches — not with exercising delegated legislative authority, let alone exercising that authority in a quasi-judicial tribunal empowered to overrule judicial decisions. Indeed, one might question whether Chevron step two muddles the separation of powers by delegating to the Executive the power to legislate generally applicable rules of private conduct. See, e.g., Dep’t of Transp. v. Ass’n of Am. R.Rs., — U.S. -, 135 S.Ct. 1225, 1240-42, 191 L.Ed.2d 153 (2015) (Thomas, J., concurring in the judgment); id. at 1237 (Alito, J., concurring). And whether the combination of Chevron and Brand X further muddles the muddle by intruding on the judicial function too. See, e.g., Brand X, 545 U.S. at 1017, 125 S.Ct. 2688 (Scalia, J., dissenting) (“Article III courts do not sit to render decisions that can be reversed or ignored by executive officers.”); Perez v. Mortg. Bankers Ass’n, — U.S. -, 135 S.Ct. 1199, 1217-19, 191 L.Ed.2d 186 (2015) (Thomas, J., concurring in the judgment); Michigan v. EPA, — U.S. -, 135 S.Ct. 2699, 2712-13, 192 L.Ed.2d 674 (2015) (Thomas, J., concurring).5
Still, as but a court of appeals Chevron and Brand X bind us and the question is what to do in light of them — what the law might have to say about the retroactive application of agency adjudications making delegated legislative policy decisions — accepting -that such agency actions are themselves legally permissible. Coming at it from another angle, if the separation of powers doesn’t forbid this form of decision-*1172making outright, might second-order constitutional protections sounding in due process and equal protection, as embodied in our longstanding traditions and precedents addressing retroactivity in the law, sometimes constrain the retroactive application of its results?
We think the answer yes. In light of the principles and precedents we’ve outlined, it seems to us that the more an agency acts like a judge — applying preexisting rules of general applicability to discrete cases and controversies — the closer it comes to the norm of adjudication and the stronger the case may be for retroactive application of the agency’s decision. But the more an agency acts like a legislator — announcing new rules of general applicability — the closer it comes to the norm of legislation and the stronger the case becomes for limiting application of the agency’s decision to future conduct. The presumption of prospectivity attaches to Congress’s own work unless it plainly indicates an intention to act retroactively. That same presumption, we think, should attach when Congress’s delegates seek to exercise delegated legislative policymaking authority: their rules too should be presumed prospective in operation unless Congress has clearly authorized retroactive application. And this logic, we believe, suffices to resolve our case. For an agency operating under the aegis of Chevron step two and Brand X comes perhaps as close to exercising legislative power as it might ever get. And no one before us contends that Congress has clearly authorized the BIA to apply its decision in Briones retroactively.
If the presumption we’ve described and our application of it sound familiar it’s because they are. In Bowen v. Georgetown University Hospital, 488 U.S. 204, 109 S.Ct. 468, 102 L.Ed.2d 493 (1988), the Court held that, absent express congressional approval, newly promulgated agency rules should apply only prospectively because of their affinity to legislation. “[Congressional enactments ... will not be construed to have retroactive effect unless their language requires this result. By the same principle, a statutory grant of legislative rulemaking authority will not, as a general matter, be understood to encompass the power to promulgate retroactive rules unless that power is conveyed by Congress in express terms.” Id. at 208, 109 S.Ct. 468 (citations omitted). The Court rested its retroactivity analysis, then, on an analogy between legislation and agency rulemaking and an appreciation of the fact that rulemaking is a form of delegated policymaking authority. And we can think of no reason why Bowen’s analogy between legislation and rulemak-ing would prove any less compelling in the Chevron step two/ Brand X scenario we confront today.6
Of course, as a formal matter Bowen addressed agency rulemaking and an agency acting under the authority of Chevron step two and Brand X will usually proceed through adjudication. So, for example, the BIA in this case announced its new rule in Briones in a quasi-judicial proceeding with lawyers and administrative law judges and briefs and arguments and many of the other usual trappings of a judicial proceeding. And for that very reason the BIA now argues before us that it “d[id] not change the law, but rather explain[ed] what the law has always meant.” In re De Niz Robles, No. A074 *1173577 772, 2014 WL 3889484, at *3 (BIA July 11, 2014) (unpublished) (internal quotation mark omitted).
But substance doesn’t always, follow form. In the Chevron step two/ Brand X scenario an agency may conduct what looks like an adjudicatory proceeding. But in that proceeding the agency hardly interprets or applies a preexisting legal rule to the specifics of a case or controversy. Much to the contrary, to reach Chevron step two the agency must first establish that traditional tools of statutory interpretation fail to reveal “what the law has always meant.” At that point the agency — avowedly and self-consciously— exploits the law’s ambiguity and exercises its “delegated” “policy-making” authority to write a new rule of general applicability according to its vision of the law as it should be. And finally, and again very much like a legislature (one dissatisfied with judicial applications of existing law), the agency may go so far as to overthrow judicial interpretations to effect its new vision. Coming at the point from another angle, courts don’t defer to the agency’s new view under Chevron step two and Brand X because it represents a superior interpretation of existing law — indeed, courts may not even ask that question. Instead, courts defer to the agency’s new view because the agency has been authorized to fill gaps in statutory law with its own policy judgments. Form, then, can’t obscure the fact that an agency exercising its Chevron step two/ Brand X powers acts in substance a lot less like a judicial actor interpreting existing law and a good deal more like a legislative actor making new policy — certainly as much, like a legislator as the rulemaking agency in Bowen — and thus fairly subject to the same presumption of prospectivity that attaches there.
Any other conclusion would, as well, leave the Supreme Court’s teaching in Bowen on doubtful footing. While the Court has granted agencies a fair amount of flexibility in choosing between rulemak-ing and adjudication, it has long encouraged the former route because rulemaking offers more notice (due process) and better protects against invidious discrimination (equal protection). See SEC v. Chenery Corp. (Chenery II), 332 U.S. 194, 202-03, 67 S.Ct. 1575, 91 L.Ed. 1995 (1947); NLRB v. Bell Aerospace Co., 416 U.S. 267, 294, 94 S.Ct. 1757, 40 L.Ed.2d 134 (1974). Allowing agencies the benefit of retroactivity always and automatically whenever they choose adjudication over rulemaking would create a strange incentive for them to eschew the Court’s stated preference for rulemaking — and render Bowen easily evaded.
For all these reasons, we join those who have suggested that a new agency rule announced in a Chevron step two/ Brand X adjudication should be treated “no differently] from a new agency rule announced by notice-and-comment rule-making ... for purposes of retroactivity analysis.” Velásquez-García v. Holder, 760 F.3d 571, 581 (7th Cir.2014); see also Abner S. Greene, Adjudicative Retroactivity in Administrative Law, 1991 Sup.Ct. Rev. 261, 274 (“[I]f the rule against retro-activity properly applies to legislation, it ought to apply to administrative adjudication that is functionally equivalent to legislation as well.”); William V. Luneburg, Retroactivity and Administrative Rule-making, 1991 Duke L.J. 106, 152-53 (“Because such a thin line separates rulemak-ing from adjudication, the effect of Bowen may stretch beyond the legislative rule-making arena.”); cf. United States v. Saucedo, 950 F.2d 1508, 1514 (10th Cir.1991) (“[W]hen we are required to overrule precedent in this circuit in order to interpret the guideline consistent with the amended commentary, we cannot agree with the *1174Sentencing Commission that the amendment merely clarified the pre-existing guideline.”), abrogated on other grounds, Stinson v. United States, 508 U.S. 36, 113 S.Ct. 1913, 123 L.Ed.2d 598 (1993).7
Of course, Bowen’s — and our — analogy to legislative activity is just that. And some might well question the attempt to draw any analogies between administrative action and judicial or legislative decision-making. See generally John F. Manning, Nonlegislative Rules, 72 Geo. Wash. L.Rev. 893 (2004) (discussing challenges associated with drawing lines between interpretive and legislative rules). But drawing analogies and lines is no small part of what judges do. And the difficulty of a task is not reason enough to abandon it, especially if it illuminates and aids in the enforcement of underlying constitutional demands. Cf. Clinton v. City of New York, 524 U.S. 417, 452, 118 S.Ct. 2091, 141 L.Ed.2d 393 (1998) (Kennedy, J., concurring) (“Abdication of responsibility is not part of the constitutional design”). Neither does the fact that some agency actions will prove hard to classify mean all will be. And an agency decision under Chevron step two and Brand X strikes us as a particularly easy case, a form of administrative activity about as legislative as they come.
Besides, even if we were inclined to embrace entirely the indeterminacy of analogies to judicial and legislative action and eschew any effort to enforce the interpretive-legislative divide, the question would still remain (to be decided in each case unaided by analogical reasoning) whether and when an agency’s decision should be afforded retroactive effect. See Manning, supra, at 928. And it seems to us that looking directly to the underlying due process and equal protection concerns that traditionally attend retroactive lawmaking would lead us to the same conclusion in any event. For the concerns that attend retroactive legislation equally attend retroactive agency action in the Chevron step two/ Brand X context. In both cases permitting retroactivity would undo settled expectations in favor of a new rule of general applicability rendered by a deci-sionmaker expressly influenced by policy and politics. In both cases the decision-maker would be able to punish those who *1175have done no more than order their affairs around existing law — and could accomplish all this with full view of who will stand to flourish or flounder, left not merely to predict who the winners and losers might be, but able to single out disfavored persons and groups and punish them for past conduct they cannot now alter. To avoid problems like these, the work of the primary legislative actor in our legal order (Congress) has always been presumptively prospective — and the saliency of these same worries in the Chevron step two/ Brand X context suggests to us that the same rule should apply here too. See Greene, supra, at 282 (“Whenever an agency acts as an ‘interpretive lawmaker,’ either in its rulemaking or adjudicative capacity, it should be subject to the same rule disfavoring retroactivity as the original, ‘noninterpretive lawmaker,’ Congress.”).
Our assessment of the underlying due process and equal protection principles at play in this context is consistent with and confirmed by the Supreme Court’s leading discussion of agency adjudicatory powers in Chenery II. At one point in our history even leading exponents of the administrative state (and some very wise judges) suggested that most every administrative decision should be presumptively prospective. After all, they noted, agency deci-sionmakers aren’t insulated from politics and policymaking in the way Article III judges are and as such may be expected to exploit the power of retroactivity in ways worrisome to due process and equal protection even when they’re acting in proceedings that look and feel like “adjudications.” See, e.g., Chenery II, 332 U.S. at 209-18, 67 S.Ct. 1575 (Jackson, J., dissenting); Bell Aerospace Co. v. NLRB, 475 F.2d 485, 495-97 (2d Cir.1973) (Friendly, C.J.), rev’d in part, 416 U.S. 267, 94 S.Ct. 1757, 40 L.Ed.2d 134 (1974). Of course, the Supreme Court in Chenery II opted to afford agencies a good deal more leeway than that. See 332 U.S. at 203, 67 S.Ct. 1575. It allowed agencies to give retroactive effect to their adjudicatory decisions applying preexisting rules to new factual circumstances. It even authorized the creation of some new and retroactively applicable general rules through administrative adjudicatory proceedings. Id. But the authority the Court extended to agencies to craft new rules retroactively through adjudication was not boundless. Indeed, Chenery II itself emphasized that “[t]he function of filling in the interstices of [legislation] should be performed, as much as possible, through [the] quasi-legislative promulgation of rules to be applied in the future.” Id. at 202, 67 S.Ct. 1575. Chen-ery II also expressly instructed that agency adjudications sometimes should be given prospective effect only. Id. at 203, 67 S.Ct. 1575. In particular, Chenery II indicated that it’s necessary for agencies and courts to “balance[ ]” the “ill effect[s]” associated with retroactive application of a new rule against “the mischief of producing a result which is contrary to a statutory design or to legal and equitable principles.” Id.
It’s this instruction that we think instructive here. For it seems to us that if there is any area where the “ill effect[s]” of retroactivity in an agency adjudication outweigh the “mischief’ associated with pure prospectivity, the Chevron step two/ Brand X scenario fits the bill. Of course, Chenery 77’s suggestion that we should “balance[ ]” the costs and benefits associated with prospectivity and retroactivity can be little more than a metaphor, for there is no easy way to compare incommensurate goods like, for example, the protection of settled expectations against the government’s policy interests in retroactive application of a new rule. The job is like asking us to compare the weight of a stone to the length of a line. Eugene Volokh, Crime-Facilitating Speech, 57 Stan. *1176L.Rev. 1095, 1137 n.172 (2005) (citing Bendix Autolite Corp. v. Midwesco Enters., Inc., 486 U.S. 888, 897, 108 S.Ct. 2218, 100 L.Ed.2d 896 (1988) (Scalia, J., concurring in the judgment)). But when the various factors we are charged with examining all point in the same direction — when the stone is very heavy and the line very short — then at least we can be relatively sure of the right answer. And this, we think, is one of those cases. See William J. Stuntz, Commentary, O.J. Simpson, Bill Clinton, and the Transsubstantive Fourth Amendment, 114 Harv. L.Rev. 842, 869 n.91 (2001) (“In fact, courts make such judgments regularly, and at least in some cases they do not seem particularly hard to make. Some lines are very short, and some rocks are very heavy.”).
The reasons we’ve alluded to already. In the Chevron step two/ Brand X context, it’s easy to see the “ill effectfs]” of retroac-tivity: upsetting settled expectations with a new rule of general applicability, penalizing persons for past conduct, doing so with a full view of the winners and losers — all with a decisionmaker driven by partisan politics. It’s pretty hard, too, to see how requiring prospectivity in these circumstances would be “contrary to [any] statutory design or to legal and equitable principles” — the sort of “mischief’ Chenery II feared might accompany purely forward-looking agency decisionmaking. By definition, the agency in the Chevron step two/ Brand X scenario isn’t seeking to enforce the law as it is but instead seeks to exploit a gap in the law to implement its own (current but revisable) vision of what the law should be. If it seems like we’re repeating ourselves somewhat, it’s because we are, for the factors Chenery II instructs us to consider are really just another way of ensuring that we consult the underlying and traditional due process and equal protection principles that have always informed retroactivity’s place in the law. So it should be no surprise that they lead to much the same analysis and conclusion.
Supporting our understanding of how Chenery II’s balancing test plays out are the reasons the Court gave for that test in the first place. In Chenery II, the Court declined to ban agencies outright from retroactively applying the results of their adjudicatory proceedings because it foresaw three exigent situations in which it thought that sort of decisionmaking might (sometimes) be appropriate. Not one of those exigencies applies here, in the Chevron step two/ Brand X context. First, the Court said, problems may arise that the agency could not have “reasonably foresee[n]” and that need to be solved retroactively “despite the absence of a [previously announced] general rule.” Chenery II, 332 U.S. at 202, 67 S.Ct. 1575. Yet it’s hard to see how a problem already addressed by a circuit court opinion can be fairly called “unforeseeable.” Second, the Court said, an agency may not have “sufficient experience” with the problem “to warrant rigidi-fying its tentative judgment into a hard and fast rule.” Id. But an agency choosing to overrule settled judicial interpretations to announce a new rule of general applicability can’t easily be said to be burdened by such reticence. Finally, the Court reasoned, the problem at hand may be “so specialized and varying in nature” that it’s “impossible” to capture with a general rule. Id. at 203, 67 S.Ct. 1575. But announcing a new policy of general applicability is precisely what the agency does in the Chevron step two/ Brand X scenario. In these three situations, the Supreme Court said, agencies might need the flexibility to proceed retroactively and on a case-by-case basis — in these situations the “mischief’ of prospectivity may at times outweigh the “ill effect[s]” of retroactivity. But these situations are not every sitúa*1177tion, and we think they are quite incomparable to the one we face today.8
Finally, our decision finds support in circuit precedent as well. Seeking to give further content to Chenery II’s directive that courts should “balance[]” the costs and benefits associated with giving retroactive effect to agency adjudications, this court in Stewart Capital Corp. v. Andrus, 701 F.2d 846 (10th Cir.1983), developed an elaborate five-factor “balancing” test of its own for application on a case-by-case basis. Even then, the court did not suggest its five factors were exclusive or even always the most pertinent. Indeed, by everyone’s admission Stewart Capital’s first factor— “[wjhether the particular case is one of first impression,” id. at 848 — is irrelevant here.9 Still, the remaining factors capture (again) many of the due process and equal protection concerns associated with the retroactivity analysis we’ve already explored. And their application to this ease proves consistent with and supportive of our conclusion that Chevron step two/ Brand X decisions normally merit prospective application only.
What are the remaining Stewart Capital factors? The second asks whether the agency’s action “represents an abrupt departure from well established practice or merely attempts to fill a void in an unsettled area of the law.” Id. The third asks us to explore whether and to what extent the “party, against whom the new rule is applied, relied on the former rule.” Id. Really, then, both of these factors direct our attention to the question whether the petitioner can claim reasonable rebanee on some past rule or decision, a due process concern always at the heart of retroactivity analysis. See Landgraf, 511 U.S. at 266, 114 S.Ct. 1483. And not unlike the two that precede them, the fourth and fifth factors might be described as flip sides of the same coin, the one asking how weighty is the burden the petitioner would feel from retroactive application of the agency decision and the other how weighty is the agency’s interest in retroactivity. Stewart Capital, 701 F.2d at 848.
How do these various factors play out in our case? Looking to the second factor, the BIA contends that its decision in Briones was designed to “fill a void,” not to effect “an abrupt departure” from existing practice. And in one sense it has a point: Chevron step two permits an agency bke the BIA to enforce its new policy decision only because of a gap in existing *1178statutory law. But this observation tells at most only half the story, failing to capture the Brand X twist present here. After all, at the time Mr. De Niz Robles filed his petition, Padilla-Caldera I was on the books and expressly permitted his application. And it was only thanks to Brand X that the agency was able to have this precedent overruled and replaced with its 180-degree-opposed judgment. Indeed, it seems fair to say that the whole point of Brand X is to permit the agency to effect, as it did here, “an abrupt departure” from existing judicial precedent.
Turning to the third factor, there’s no dispute that Mr. De Niz Robles relied on Padilla-Caldera I when he filed his petition. Everyone acknowledges that at the time he began his administrative odyssey Mr. De Niz Robles possessed two lawful options to secure permanent residency in this country — file a petition for adjustment of status under Padillar-Caldera I or leave the country and begin the waiting period. And everyone, the BIA included, accepts that Mr. De Niz Robles relied on Padilla-Caldera I in choosing the first course.
But while accepting that Mr. De Niz Robles did rely on our precedent, the agency disputes whether he (or anyone else) could have reasonably done so. The BIA offers three reasons why no reasonable person could’ve relied on Padillar-Caldera I. The agency argues that it was “obvious” “[fjrom the outset” that a statutory tension existed between §§ 1255(i)(2)(A) and 1182(a)(9)(C)(i)(I); that any reliance on Padilla-Caldera I necessarily carried some risk because it was “subject to revision” by the BIA under Chevron step two and Brand X; and that the “risk” of an adverse agency “revision” was apparent from the start given that a Fifth Circuit decision had previously reached a result contrary to the one this court reached in Padilla-Caldera I. For all these reasons, the BIA says, Mr. De Niz Robles had no business relying on this court’s precedent.
We do not agree. Undergirding all of the agency’s arguments lies the implicit premise that a party cannot reasonably rely on a judicial decision whenever its interpretation of the law is foreseeably subject to “revision” by an executive agency. But judicial decisions interpreting the law are not usually thought unworthy of reliance just because the legislature may foreseeably reenter the field and revise the statute. Indeed, legislation bears presumptively prospective effect in our legal order precisely to ensure that the people may rely on the law as it is, so long as it is, including any of its associated judicial interpretations. See U.S. Bancorp Mortg. Co. v. Bonner Mall P’ship, 513 U.S. 18, 26, 115 S.Ct. 386, 130 L.Ed.2d 233 (1994) (“Judicial precedents are presumptively correct and valuable to the legal community as a whole.” (internal quotation mark omitted)); AT & T Corp. v. Hulteen, 556 U.S. 701, 711-13, 129 S.Ct. 1962, 173 L.Ed.2d 898 (2009) (declining to apply a statutory change retroactively to a party who relied on a prior Supreme Court decision). We can see no good reason to reach the (highly anomalous) conclusion that it’s reasonable for a party to rely on prior circuit (maybe even Supreme Court) precedent in the face of foreseeable revision by our policymaking principal (Congress), but not in the face of foreseeable revision by its agent (the BIA). In an age where there is so much law and so many lawmaking authorities, trying to figure out what the controlling rule is and how to order your affairs accordingly can be tough enough. To suggest that even when you find a controlling judicial decision on point you can’t rely on it because an agency (mind you, not Congress) could someday act to revise it would be to create a trap for the unwary and paradoxically encourage those *1179who bother to consult the law to disregard what they find.
Besides this categorical error, the BIA’s reasonable reliance arguments bear their warts even when viewed on their own terms. Even accepting that any tension between §§ 1255(i)(2)(A) and 1182(a)(9)(C)(i)(I) was “obvious” “[f]rom the outset” speaks not at all to the question whether it was reasonable for Mr. De Niz Robles to rely on this court’s resolution of that tension in Padillar-Caldera I. Neither was it clear at the time of Mr. De Niz Robles’s application for adjustment of status that Padilla-Caldera I was “subject to revision” by the BIA under Chevron step two and Brand X. Only in Padillar-Caldera II did this court make clear that it considered the statutory scheme ambiguous and the agency entitled to override Padillar-Caldera I.10 And it was hardly clear that the BIA would reject Padillar-Caldera I even if given the chance: INS officials themselves offered conflicting informal guidance before the BIA issued Briones.11
Turning to the fourth and fifth Stewart Capital factors — the balance of comparative harms — the story here proves much the same. Retroactive application of Briones would mean that Mr. De Niz Robles loses the opportunity promised by Padilla-Caldera I to have his petition for adjustment of status considered by the Attorney General, a petition that could secure his permanent lawful residency in this country. Of course, Padillar-Caldera I by no means promised anyone a positive result. But it did guarantee petitioners like Mr. De Niz Robles the chance to apply, a chance the retroactive application of Briones would take away. And the loss of that chance to remain in the country lawfully cannot be dismissed as nothing. Even the BIA acknowledges that the burden on Mr. De Niz Robles from the retroactive application of Briones would be “significant.” See St. Cyr, 533 U.S. at 325,121 S.Ct. 2271 (“There is a clear difference, for the purposes of retroactivity analysis, between facing possible deportation and facing certain deportation.”).
Meanwhile, it’s hard to see any significant — or really any — harm befalling the BIA if it must abide by Padilla-Caldera I when it comes to Mr. De Niz Robles’s petition. After all and again, that decision would only require the Attorney General to consider, not automatically grant, his petition. The BIA doesn’t dispute (or even address) this point but replies that it has an interest in seeing the law applied “uniformly.” But that is the very question before us — whether the government can identify sufficient reason to apply a (newly announced) agency rule uniformly to past cases — and the BIA’s argument seems simply to assume the answer. The gov*1180ernment’s claimed interest in uniformity-faces another problem, too, in its level of generality. As against Mr. De Niz Robles’s personal interest in seeing his individual petition adjudicated the government asks us to weigh a systemic one. But abstracted to a sufficiently great height almost any governmental interest might be said to touch on some grave and fundamental concern. See Yellowbear v. Lampert, 741 F.3d 48, 57 (10th Cir.2014). And it would be comparing lines to rocks in a particularly strange way if we were to weigh an individual person’s interests in “balance” against a governmental objective abstracted to a high level of generality. See id. Looking to the particulars of this case in the same manner on both sides of the ledger, the BIA identifies no harm it might face from considering Mr. De Niz Robles’s individual petition. And even if we were inclined to examine the BIA’s interest as abstractly as it might wish, it would still be hard to see it amounting to much, for the agency nowhere seeks to explain why it thinks its claimed interest in “uniformity” is compelling. Particularly when applying Padilla-Caldera /’s judicial interpretation would itself advance uniformity, if of a different kind — uniformity in the sense that all petitioners. should receive the benefit of the law as it existed at the time they made their administrative applications. A kind of uniformity interest the law, as we’ve seen, often recognizes.
Trying to commensurate incommensurable legal factors is never an easy judicial chore — and the job isn’t made any easier when the number of factors we’re asked to juggle proliferates. But whether under Stewart Capital or Chenery II the stone on Mr. De Niz Robles’s side of the ledger feels pretty heavy and the line on the BIA’s side turns up pretty short. And Bowen’s analogy points in the same direction too. Neither should it come as a surprise that all these tests yield the same essential conclusion, for as we’ve seen the very same due process and equal protection concerns animate them all. And while we may never substitute another court’s judgment for our own, we take heart from the fact that the only other circuit to have considered the competing factors at play in a case like our own has reached the same judgment we do. In Acostar-Olivarria, the Ninth Circuit employed its version of the Stewart Capital factors and refused to allow the BIA to apply Briones retroactively to a petitioner who sought adjustment of status prior to that decision and after (and in reliance on) that circuit’s counterpart to Padilla-Caldera I. See Acosta-Olivarria, 799 F.3d at 1272. So it is that, in the end, the BIA’s decision to apply Briones retroactively to Mr. De Niz Robles finds no support in our examination of the principles underlying the law of retroactivity, in Supreme Court or circuit precedent, or in relevant authority from other jurisdictions.12
The petition for review is granted and the case remanded to the BIA for further proceedings consistent with this opinion.
. It's a nice question whether Brand X permits executive agencies to override not just circuit precedent but Supreme Court decisions too. Compare 545 U.S. at 1003, 125 S.Ct. 2688 (Stevens, J., concurring) (suggesting that agency override "would not necessarily be applicable to a decision by this Court”), with id. at 1016-17, 125 S.Ct. 2688 (Scalia, J., dissenting) (arguing that the decision frees agencies even to take action "that the Supreme Court [had] found unlawful”).
. A separate writing in Garfias-Rodríguez suggested that no retroactivity problem arose in that case — one somewhat similar to our own — because a person illegally present in this country has "no 'right to continue illegal conduct indefinitely under the terms on which it began.’ ” 702 F.3d at 529 (Kozinski, C. J., disagreeing with everyone) (quoting Fernandez-Vargas v. Gonzales, 548 U.S. 30, 46, 126 S.Ct. 2422, 165 L.Ed.2d 323 (2006)). But the BIA order before us didn’t invoke this rationale in defense of its decision and so the agency may not now employ it on appeal. See SEC v. Chenery Corp. (Chenery I), 318 U.S. 80, 92-95, 63 S.Ct. 454, 87 L.Ed. 626 (1943). We can, as well, see why the BIA order didn’t attempt this argument. For no one suggests that Padilla-Caldera I invested Mr. De Niz Robles (or anyone else) with a right to remain in this country illegally. Instead and much more modestly, that decision held that he and others like him possessed the statutory right to file a petition for adjustment of status, a petition that could be granted or refused at the discretion of the Attorney General. See 453 F.3d at 1240. It is this much more limited státutory right — "the possibility of relief,” id. at 1244 (emphasis added)— that’s at issue before us today and that the BIA seeks to withdraw retroactively.
. Many thoughtful judges and commentators have contended that the Constitution’s straightforward prohibition on "ex post facto Law[s],” as well as the original public understanding of that term, suggest a wider prohibition on retroactive legislation, one that extends as fully to the civil context as it does to the criminal. See, e.g., E. Enters. v. Apfel, 524 U.S. 498, 538-39, 118 S.Ct. 2131, 141 L.Ed.2d 451 (1998) (Thomas, J., concurring); Satterlee v. Matthewson, 27 U.S. (2 Pet.) 380, 416, 7 L.Ed. 458 (1829) (Johnson, J., concurring); Laura Ricciardi & Michael B.W. Sinclair, Retroactive Civil Legislation, 27 U. Tol. L.Rev. 301, 302-12 (1996); Oliver P. Field, Ex Post Facto in the Constitution, 20 Mich. L.Rev. 315, 321-31 (1922); Jane Harris Aiken, Ex Post Facto in the Civil Context: Unbridled Punishment, 81 Ky. L.J. 323, 324-36 (1992); William Winslow Crosskey, 1 Politics and the Constitution in the History of the United States 325-29 (1953).
. We acknowledge this (slight) asymmetry is perhaps an oddity in the law, see Bradley Scott Shannon, The Retroactive and Prospective Application of Judicial Decisions, 26 Harv. J.L. & Pub. Pol’y 811, 836-37 (2003), but happily it’s an oddity that doesn’t play a role in our work today.
. Thoughtful scholars have presented arguments along similar lines, recently and notably including Philip Hamburger, Is Administrative Law Unlawful? 285-321 (2014). See also Bogdan Iancu, Legislative Delegation: The Erosion of Normative Limits in Modem Constitutionalism 230-51 (2012); David Schoenbrod, Power Without Responsibility: How Congress Abuses the People Through Delegation 99-106 (1993); Cynthia R. Farina, Statutory Interpretation and the Balance of Power in the Administrative State, 89 Colum. L.Rev. 452, 478-81 (1989); Linda D. Jellum, “Which Is to Be Master," the Judiciary or the Legislature? When Statutory Directives Violate Separation of Powers, 56 UCLA L.Rev. 837, 897-98 (2009); Nadine Strossen, Delegation as a Danger to Liberty, 20 Cardozo L.Rev. 861, 861-70 (1999).
. Neither in reaching its result did the Court in Bowen depend on the APA alone, for the Court expressly referenced and relied on pre-APA cases. See Bowen, 488 U.S. at 208, 109 S.Ct. 468 (citing, for example, Claridge Apartments Co. v. Comm’r, 323 U.S. 141, 164, 65 S.Ct. 172, 89 L.Ed. 139 (1944), and Miller v. United States, 294 U.S. 435, 439, 55 S.Ct. 440, 79 L.Ed. 977 (1935)).
. There is one wrinkle we must admit. An agency in the Chevron step two/Brand X scenario may enforce its new policy judgment only with judicial approval. So, for example, the BIA depended on Padilla-Caldera II to render Briones effective. Given this arrangement you might ask: are we right to analyze the problem at hand as one of agency action, rather than one of judicial action subject to the presumption of retroactivity that attends the operation of the "judicial Power” under our Constitution? To our minds, there are (at least) two reasons why the answer is yes. First and more formally, what's at issue in these cases is an agency decision seeking to apply an agency policy pronouncement retroactively. So in our case, we have a BIA order purporting to apply a BIA rule retroactively to Mr. De Niz Robles’s past conduct. Second and more substantively, a court doesn’t defer to an agency’s interpretation under Chevron step two and Brand X because the court’s prior reading was wrong or because the agency is more adept at construing statutory text. Rather and as we’ve seen, it does so because Congress delegated to the agency the power to make policy judgments to fill out a statutory scheme. And not insignificantly, the agency judgment to which the court defers is not "a once-and-for-always definition of what the statute means,” but a policy choice subject to revision by the agency "on a continuing basis.” Garfias-Rodriguez, 702 F.3d at 515-16 (quoting Chevron, 467 U.S. at 864, 104 S.Ct. 2778) (internal quotation marks omitted). Under these circumstances, there can be little doubt that the law whose retroactivity we consider here emanates in substance as well as form from the agency, not the court. Id. at 516 ("To [conclude] otherwise would ignore the effect of Chevron and treat the agency decision as though it had issued from the court itself.”).
. We pause to note that our decision is also consistent with the Supreme Court’s decision in NLRB v. Wyman-Gordon Co., 394 U.S. 759, 89 S.Ct. 1426, 22 L.Ed.2d 709 (1969). We read that decision as rejecting agency efforts to avoid the rulemaking prescribed by the APA through the use of prospective adjudication and we fully subscribe to that result and the law’s general preference for rulemaking. In this case, however, there is no allegation that the BIA is bound by the APA or that it is seeking to evade its dictates. See, e.g., Greene, supra, at 296 ("Thus, although the Wyman-Gordon plurality disapproved of a particular purely prospective adjudicatory rule, it-did so because of the manner in which the NLRB had sought to promulgate that rule; it did not issue a stark holding that purely prospective adjudicative rules are always improper.”). Indeed, many courts in similar circumstances have applied adjudicative rules purely prospectively when no allegation of APA subterfuge is involved. See, e.g., Petrolite Corp. v. FERC, 667 F.2d 664 (8th Cir.1981); McDonald v. Watt, 653 F.2d 1035 (5th Cir. Unit A 1981); Transwestern Pipeline Co. v. FERC, 626 F.2d 1266 (5th Cir.1980).
. The first factor aims to ensure that a party seeking to overturn an administrative rule gets the benefit of its efforts, a factor of some significance in the context of disputes between private parties where a private litigant must bring suit. But by everyone's admission this factor isn’t relevant where, as here, the government initiates removal proceedings. See, e.g., Garfias-Rodriguez, 702 F.3d at 521; Retail, Wholesale & Dep’t Store Union v. NLRB, 466 F.2d 380, 390 (D.C.Cir.1972).
. Some language in Padilla-Caldera I seems tantamount to declaring the statutory scheme ambiguous, describing it as involving "two conflicting provisions.” 453 F.3d at 1241. Elsewhere, though, the opinion seems to suggest that the statutory scheme unambiguously leans in Mr. Padilla-Caldera's favor, stating, for example, that "[w]e see no basis upon which we may conclude that” the Attorney General lacked discretion to adjust the status of illegal reentrants, and later adding that "Congress intended” the opposite result. Id. at 1244.
. In one memorandum, the INS took the view that the Attorney General did have the authority to consider adjustment of status for a large class of immigrants who were otherwise inadmissible under the INA. See Memorandum from David Martin, INS Gen. Counsel, to Michael L. Aytes, Assistant Comm'r, Office of Benefits (Feb. 19, 1997), reprinted in 74 No. 11 Interpreter Releases 499 (March 24, 1997). In another, the INS took the opposite view. See Memorandum from Louis D. Crocetti, Jr., INS Assoc. Comm’r, Office of Examinations, to INS Officials (May 1, 1997), reprinted in 74 No. 18 Interpreter Releases 781 (May 12, 1997).
. Though in Garfias-Rodriguez the Ninth Circuit concluded that the same factors favored retroactive application of Briones in the case before it, there’s nothing inconsistent about that decision, Acosta-Olivarria, and ours. There isn’t because the petitioner in Garfias-Rodriguez filed his petition before that circuit’s equivalent of Padilla-Caldera I and so could not claim any reliance on it. 702 F.3d at 522. Indeed, the Garfias-Rodriguez court noted that ”[t]he only window in which [the petitioner’s] reliance interest based on our previous rule might have been reasonable is the” period between the circuit court decision and Briones. Id. Exactly the situation presented in Acosta-Olivarria and in our case.