DISSENT
BOGGS, Circuit Judge,dissenting.
This case presents a tangled web of factual circumstances worthy of a soap opera, AEDPA procedural complications arising from a coincidence of timing, and interesting and important Sixth Amendment constitutional law. Despite the majority’s lucid analysis, I would hold that the Michigan courts’ ultimate conclusion — that the admission of Cassaday’s suicide note to his parents did not violate the Confrontation Clause — is both correct, and a fortiori, not an “unreasonable application” of Supreme Court precedent. I would also hold that the state did not waive harmless-error review, and that even if it did, we should exercise our discretion to review for harmlessness sua sponte. Finally, I would hold that the overwhelming evidence that was properly admitted rendered any hypothetical error harmless. I therefore respectfully dissent.
I. Applicability of AEDPA Deference
The majority holds that AEDPA deference does not apply here, and that we are free to consider Miller’s Confrontation Clause claim de novo using outcome-determinative legal principles that were not available to the Michigan state courts at the time of Miller’s trial and appeal. I believe that the majority’s interpretation of AEDPA skirts the statute’s purposes of promoting deference to reasonably decided state-court decisions and respect for the states’ interest in the finality of their own criminal convictions.
As an initial matter, this case presents a relatively rare situation. As the majority notes, the Michigan trial court correctly ruled the suicide note admissible under then-controlling Confrontation Clause jurisprudence. In June 2003, the Michigan Court of Appeals issued a reasoned decision, also correct under the law as it stood at the time, affirming the trial court’s Confrontation Clause analysis. On March 8, 2004, after briefing had been completed in the Michigan Supreme Court, the United States Supreme Court decided Crawford v. Washington, 541 U.S. 36, 124 S.Ct. 1354, 158 L.Ed.2d 177 (2004), which effected a sea change in Confrontation Clause doctrine. The Michigan Supreme Court denied leave to appeal on April 1, 2004, which was entirely within its discretion, see Mich. Ct. R. 7.302; it also denied a motion to reconsider (in which Crawford was specifically brought to its attention) on June 30, 2004. Miller did not apply to the United States Supreme Court for a writ of certiorari.
As the majority recognizes, we afford “modified AEDPA deference” whenever state courts adjudicate claims on the merits without articulating their reasoning. See, e.g., Harris v. Stovall, 212 F.3d 940, 943 & n. 1 (6th Cir.2000). In such cases, “a habeas court [must] focus on the result of the state court’s decision,” id. at 943 n. 1 (emphasis added), and “uphold the state court’s summary decision unless [the habeas court’s] independent review of the record and pertinent federal law persuades [it] that [the state court’s] result contravenes or unreasonably applies clearly established federal law” as expressed in the *929holdings of the United States Supreme Court, id. at 944 (quoting Aycox v. Lytle, 196 F.3d 1174, 1177-78 (10th Cir.1999)) (emphasis added). The majority does not dispute that the Michigan Court of Appeals’s result — its finding of no Sixth Amendment violation under these circumstances — does not “contravene[ ] or unreasonably appl[y]” any United States Supreme Court holding. Accordingly, in this case, had the Michigan Court of Appeals (or the Michigan Supreme Court in denying leave for appeal) simply said “we have considered Miller’s Sixth Amendment claim” or “we have considered Miller’s Confrontation Clause claim,” and rejected it without further reasoning, we would be obliged to respect that result.
Nonetheless, here, even though the Michigan Court of Appeals took the extra step of explaining its reasoning, the majority holds that we now cannot give its decision any deference, even though that decision in fact correctly applied the law that governed at the time. This result seems anomalous, and is not compelled by any of the cases cited by the majority. See Merced v. McGrath, 426 F.3d 1076, 1081 (9th Cir.2005) (“[I]t is the state court’s decision, as opposed to its reasoning, that is judged under the ‘unreasonable application’ standard.”); Cruz v. Miller, 255 F.3d 77, 86 (2d Cir.2001) (stating that, in applying § 2254(d)(1), “we are determining the reasonableness of the state courts’ ‘decision,’ not grading their papers.”).
Further, it is not even clear that Crawford is relevant to the question of AEDPA deference. Under 28 U.S.C. § 2254(d)(1), state-court judgments are immune from attack on habeas unless they are “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” (emphasis added). In Terry Williams v. Taylor, one of the Supreme Court’s seminal AEDPA cases, the Court made two contradictory statements as to the crucial time frame we must look to for determining what constitutes “clearly established Federal law.” Justice O’Connor, writing for herself and four other justices, said that this phrase refers to “this Court’s decisions as of the time of the relevant state-court decision.” 529 U.S. 362, 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000) (emphasis added); see also Lockyer v. Andrade, 538 U.S. 63, 71-72, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003) (O’Connor, J.) (stating that “ ‘clearly established Federal law’ under § 2254(d)(1) is the governing legal principle or principles set forth by the Supreme Court at the time the state court renders its decision.” (emphasis added)). However, Justice Stevens, writing for a different majority in Terry Williams, stated that AEDPA requires us to look to “law that was clearly established at the time the state-court conviction became final” 529 U.S. at 390, 120 S.Ct. 1495 (emphasis added).
This disjunction has been little noted, perhaps because it is rare to have a situation such as the one we face here, where the relevant law changed between the two alternative times (and where the defendant did not seek certiorari, which would usually have resolved the matter). Our cases have cited the O’Connor formulation far more often than the Stevens formulation, although we appear not to have confronted a case up to this point where the formulation chosen would make a difference. The Eleventh Circuit has noted the contradiction and chosen to follow the O’Connor formulation, noting that the Supreme Court has cited it at least four times after Williams, while not citing the Stevens formulation again. Newland v. Hall, 527 F.3d 1162, 1198 n. 62 (11th Cir.2008).1
*930As the Supreme Court has explained in no uncertain terms,
AEDPA’s purpose [is] to further the principles of comity, finality, and federalism. There is no doubt Congress intended AEDPA to advance these doctrines. Federal habeas corpus principles must inform and shape the historic and still vital relation of mutual respect and common purpose existing between the States and the federal courts. In keeping this delicate balance we have been careful to limit the scope of federal intrusion into state criminal adjudications and to safeguard the States’ interest in the integrity of their criminal and collateral proceedings.
Michael Williams v. Taylor, 529 U.S. 420, 436, 120 S.Ct. 1479, 146 L.Ed.2d 435 (2000). Because our interference in a state criminal proceeding is a matter of considerable gravity, infringing on these “principles of comity, finality, and federalism,” I would at least lean toward the O’Connor formulation, which requires that we defer unless the state court actually committed error. See Herbert v. Billy, 160 F.3d 1131, 1135 (6th Cir.1998) (“[AED-PA] tells federal courts: Hands off, unless the judgment in place is based on an error grave enough to be called ‘unreasonable.’ ” (emphasis added)).2
The majority opts for the Stevens formulation, primarily on the ground that it matches up with the temporal dividing line between “old rules” and “new rules” which the Supreme Court established in Teague v. Lane, 489 U.S. 288, 301, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989) (stating that a rule of law is “old,” and therefore applies on collateral review, if it was established by the time a habeas petitioner’s conviction became final). While this symmetry may be superficially pleasing, I see no a priori reason why the temporal divide between old and new rules for purposes of Teague must align with the relevant temporal divide for purposes of AEDPA. To the contrary, it makes perfect sense that the respective dividing lines should differ, for, as the majority acknowledges, the Supreme Court has observed that “[t]he AEDPA and Teague inquiries are distinct.” Maj. Op. at 920 (quoting Horn v. Banks, 536 U.S. 266, 272, 122 S.Ct. 2147, 153 L.Ed.2d 301 (2002)).3
*931The Teague inquiry is functional; it determines what law governs the merits of a petitioner’s habeas claim. See Maj. Op. at 920. The Teague Court had to select a cutoff point to determine what rules would apply on habeas, and settled on the time “at which a conviction becomes final, primarily because the Court felt that this was the point at which the systemic interest in “repose” and reliance on concluded litigation “outweigh[ed] ... the [petitioner’s] competing interest in readjudicating [his] conviction[ ] according to [newly declared] legal standards.... ” 489 U.S. at 306, 109 S.Ct. 1060 (quoting Mackey v. United States, 401 U.S. 667, 682-83, 91 S.Ct. 1160, 28 L.Ed.2d 404 (1971) (Harlan, J., concurring in judgments in part and dissenting in part)).
AEDPA, by contrast, embodies Congress’s desire to create an initial hurdle to be surmounted before the remedy of habeas becomes available at all. See Uttecht v. Brown, 551 U.S. 1, 10, 127 S.Ct. 2218, 167 L.Ed.2d 1014 (2007) (“The provisions of [AEDPA] create an independent, high standard to be met before a federal court may issue a writ of habeas corpus to set aside state-court rulings.”); Danforth v. Minnesota, 552 U.S. 264, 309, 128 S.Ct. 1029, 169 L.Ed.2d 859 (2008) (Roberts, C.J., dissenting) (noting that unlike the Teague inquiry, which determines “whether new or old law should apply in a particular case,” AEDPA represents Congress’s intent to “control ... federal courts’ ability to grant postconviction remedies”). In other words, by enacting AEDPA, Congress intended to superimpose an added layer of deference to state-court decisions as stick, — i.e., to “tell[ ] federal courts: Hands off, unless the judgment in place is based on [a state-court] error grave enough to be called ‘unreasonable.’ ” Herbert, 160 F.3d at 1135.
The Teague inquiry does not share AEDPA’s primary concern with fostering respect for state-court decisions and protecting “the States’ interest [as against the federal government] in the integrity of their criminal and collateral proceedings.” Michael Williams, 529 U.S. at 436, 120 S.Ct. 1479. Thus, the majority’s desire to match up the two inquiries’ temporal dividing lines is misplaced. Rather, as stated previously, it is eminently reasonable that the AEDPA dividing line should be drawn so as not to penalize state courts for making correct decisions that merely fail to predict the future.
The majority correctly observes that in her opinion in Terry Williams, Justice O’Connor stated that “[w]ith [the] caveat [that the inquiry is restricted to the decisions of the Supreme Court], whatever would qualify as an old rule under our Teague jurisprudence will constitute ‘clearly established Federal law ... ’ under [AEDPA].” 529 U.S. at 412, 120 S.Ct. 1495. However, the majority reads too much into this statement. As one judge on this circuit has observed, Justice O’Con-nor was not discussing Teague’s temporal aspect in this passage, but rather, Teague’s approach to the proper level of specificity with which a “rule” must have been pronounced as of the relevant time to qualify as established:
The Supreme Court has adopted the spectrum of abstraction of Teague v. Lane to determine whether a particular legal principle was clearly established at the relevant time. See [Terry] Williams, 529 U.S. at 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (With the caveat *932that the source of clearly established law is Supreme Court jurisprudence, “whatever would qualify as an old rule under our Teague jurisprudence will constitute ‘clearly established Federal law ... ’ under § 2254(d)(1).”). At one end of the spectrum lie legal principles with such a high level of generality, like the Eighth Amendment principle of reliability in sentencing, whose application does not necessarily lead to a “predictable development” in the relevant law and therefore cannot be considered clearly established. On the other end are narrowly drawn bright-line rules with little application beyond factually indistinguishable situations. In the middle of the spectrum lie those general principles of law crafted by the Supreme Court to constitute clearly established law in a wide range of factual situations.
Davis v. Straub, 430 F.3d 281, 292 (6th Cir.2005) (Merritt, J., dissenting) (internal citations omitted). In other words, it appears that all Justice O’Connor meant to imply with her comment was that a rule that had been announced at a sufficient level of specificity to qualify as “old” under Teague will also have been announced with a sufficient level of specificity to qualify as “clearly established” under AEDPA. She was not, however, addressing the proper time frame for making this assessment.
For these reasons, I have serious doubts about the majority’s adoption of the Stevens standard as the law of this circuit. Even under the majority’s timing rule, however, I would still accord the state court’s result as to Miller’s Confrontation Clause claim — i.e., the Michigan Court of Appeals’s holding of no Sixth Amendment violation — the same degree of deference as we would afford it if that court had simply been silent as to its reasoning.
II. Merits of Confrontation Clause Claim
Having disposed of the issue of AEDPA deference, the majority examines Miller’s Confrontation Clause claim de novo and finds it meritorious. As I have stated, I would review Miller’s claim applying AED-PA deference, and under that standard, I would find her claim without merit. However, even under de novo review, I would find Miller’s Confrontation Clause claim unpersuasive.
A. Applying AEDPA Deference
First, under AEDPA, we are constrained to inquire whether the state court decision is “contrary to” or an “unreasonable application of’ Supreme Court precedent at the relevant juncture. Even assuming arguendo that the relevant time frame is when Miller’s conviction became final (i.e., applying the Stevens rule), Miller cannot meet this standard. In Crawford, the Supreme Court established that “testimonial” hearsay from an unavailable witness violated the Confrontation Clause, but pointedly “le[ft] for another day any effort to spell out a comprehensive definition of ‘testimonial.’ ” 541 U.S. at 68, 124 S.Ct. 1354. All that Crawford clearly established with respect to what constitutes “testimonial” hearsay is that “[wjhatever else the term covers, it applies at a minimum to prior testimony at a preliminary hearing, before a grand jury, or at a former trial; and to police interrogations”— i.e., “the modern practices with closest kinship to the abuses at which the Confrontation Clause was directed.” Ibid. While the Crawford Court listed various generalized definitions of “testimonial” proposed by the petitioner, an amicus, and a concurrence by Justice Thomas, it did not select among them; nor did it state that those were the only possible definitions.
*933The type of hearsay at issue here — a suicide note; in a sealed envelope, addressed to one’s parents — was obviously not directly at issue in Crawford. Nor does it fall into any of the classes of hearsay which the Crawford Court conclusively identified as “testimonial.” As for the Court’s three proposed definitions of “testimonial,” the suicide note qualifies under at most one (and even this is doubtful, as I discuss below) — “statements that were made under circumstances which would lead an objective witness reasonably to believe that the statement would be available for use at a later trial.” Id. at 52,124 S.Ct. 1354 (quoting Brief for Nat’l Ass’n of Criminal Defense Lawyers et al. as Amici Curiae at 3). However, in light of the open-endedness of the Court’s holding, it would not be an “unreasonable” application of Crawford to use one of the other two proposed definitions, or another definition entirely, under which the suicide note would not qualify. Under AEDPA, our inquiry ends here.
B. Under De Novo Review
I believe that Miller’s Confrontation Clause claim fares little better under de novo review. Without the strictures of AEDPA, we are free to consider post-Crawford case law from both the Supreme Court and this court. The only such decision directly on point is United States v. Cromer, 389 F.3d 662 (6th Cir.2004), in which a panel of this circuit adopted something close to one of the Crawford Court’s several proposed definitions of “testimonial.” In Cromer, however, we said two somewhat contradictory things in direct succession. We first said that the “proper inquiry ... is whether the declarant intends to bear testimony against the accused.” 389 F.3d at 675 (emphasis added). This sentence obviously focuses on the declarant’s actual intent. The very next sentence abandons the focus on subjective intent, stating that the intent “may be determined by querying whether a reasonable person in the declarant’s position would anticipate his statement being used against the accused in investigating and prosecuting the crime.” Ibid, (emphasis added).4 This second sentence resembles the Crawford Court’s broadest proposed definition of “testimonial” statements (i.e., those “made under circumstances which would lead an objective witness reasonably to believe that the statement would be available for use at a later trial”),5 with the notable substitution of the still broader phrase “use[ ] ... in investigating and prosecuting the crime” for the Crawford Court’s phrase “use at ... trial.”6
The Cromer panel’s standard leads immediately to a number of unanswered questions. The first involves the multiple meanings of “intend” and “anticipate.” *934Cassaday obviously intended that his parents understand that his committing suicide had to do with the murder. The majority places considerable emphasis on Cassaday’s background as a police officer, and uses that to attempt to divine what he could “anticipate” might happen as a result of his note. However, that very argument demonstrates the slipperiness of the term “anticipate.” Does that term mean that a reasonable declarant would believe it very likely that something will happen as a direct result of his actions? Or only somewhat likely? Or simply barely possible?7
Consider, for example, a series of hypotheticals involving other possible iterations of Cassaday’s suicide note. Suppose that the note had been written, but then thrown in the wastebasket. Certainly, as a police officer, Cassaday could “anticipate” that in the wake of his suicide (and the previous murder, to which he knew that suspicion would attach, at a minimum, as a result of his briefcase evidence), his house and trash might be searched by family or police. Or suppose he had written the note and then torn it into little pieces and left it in his briefcase (as occurred with the Vince Foster suicide note). Or suppose he had written the note on a pad but then completely destroyed the note, leaving, however, the impression of his words on the next page of the pad (which he did not destroy); as an officer, he would know of the means of recovering such “indented writing.” In each of these cases, perhaps in descending order of probability, Cassaday might have “anticipate^” that the actions of police or of his family would eventually lead to the information being made available to authorities.8
When we move to attempting to assess the facts of this case under the relatively vague standard in Cromer, I come to a different conclusion than the majority. In particular, I would note that Cassaday penned three suicide notes, all directed in sealed envelopes to various persons to whom he wished to explain himself. - As near as we can tell from the evidence, only one of those three notes in fact made it into the hands of the police or into court testimony. While I cannot glean definitively from the record, it certainly seems to be the case that the other two notes either were not voluntarily produced to the police; were not subpoenaed; or, in contradistinction to the notes to the parents, *935did not contain incriminating information. In any event, those facts lead me to see each of the notes as having been driven entirely by personal considerations.
Cassaday obviously understood how to provide evidence to authorities, with a good chain of custody to enhance its usability as evidence. The material in the briefcase was clearly designed for the use of the police. There is no evidence of sealing or date-stamping of the briefcase that would have prevented him from placing the suicide note in the briefcase to ensure that it, too, would go to the authorities. Further, the one letter out of the three which is known to us — the letter to Cassaday’s parents — speaks of his taking completely independent steps for the purposes of revenge and law enforcement (“I’m sending [the evidence of Miller’s involvement] to the police. She will get what is coming.”). It says nothing directing or anticipating that the note itself be given to the police, and the phrasing of the note suggests that Cassaday did not view it as part of the evidence against Miller. For all of these reasons, I would not hold under the first, subjective half of the Cromer standard that Cassaday “intended” to bear testimony through the note.
Nor would I hold that Cassaday’s note meets the second, objective half of our Cromer standard. In particular, I note that the cases applying the Cromer standard cited by the majority seem to follow a more narrow interpretation of “anticipate.” See Maj. Op. at 924 (citing United States v. Mooneyham, 473 F.3d 280, 286-87 (6th Cir.2007); United States v. Johnson, 440 F.3d 832, 843 (6th Cir.2006); United States v. Barry-Scott, 251 Fed.Appx. 983, 989-90 (6th Cir.2007)). Only in Barry-Scott (an unpublished ease) was a statement found testimonial, and in that case, it was made directly to police. In the other two cases, statements were held not to be testimonial when made to private citizens, even though a reasonably prudent person should certainly “anticipate” that anyone he talks to about his unlawful activity is a potential witness against him.
III. Harmless Error
A. Waiver
The majority holds that the state waived its harmless-error argument by failing to raise the issue of harmless error as to the suicide note in its initial answer to Miller’s habeas petition. I disagree that an invocation of waiver is appropriate here. The majority relies on United States v. Johnson, 467 F.3d 559 (6th Cir.2006), in which we observed that mere “perfunctory discussion” of harmless error in a government appellate brief “waive[s] that argument on appeal,” id. at 564. Here, by contrast, the state has thoroughly briefed the issue of harmless error before this court; accordingly, Johnson is not on point. In fact, the majority cites no case from the Supreme Court or from this circuit (and I can find none) holding that a harmless-error argument is forever waived if not raised in a habeas respondent’s initial answer before the district court.
Two considerations specific to this case further counsel against a finding of waiver. First, as the majority concedes, the state did argue the harmless-error doctrine in its answer before the district court, albeit in the section of its memorandum discussing the IMs between Cassaday and Miller, rather than the section dealing with the suicide note. Thus, even if it were the case that the state is required to claim “harmless error” in its answer or forever waive it — a proposition not established by any case that binds us — it is another thing entirely to assert that a harmless-error argument may be waived as to particular pieces of improperly admitted evidence if it is not separately asserted in the respondent’s answer with respect to each exhibit. *936I am aware of no authority from any jurisdiction suggesting as much. Nor would such a proposition make sense. The state’s answer gave Miller fair notice that the harmless-error doctrine would be at issue in this proceeding. Moreover, that doctrine involves the consideration of the value of the complained-of evidence relative to the value of all evidence before the jury. See Brecht v. Abrahamson, 507 U.S. 619, 641, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993). The state’s assertion of harmless error with respect to the IMs therefore placed in issue the relative evidentiary value of the suicide note.
Secondly, in Miller’s reply to the state’s answer, she explicitly argued that the admission of the suicide note (not the IMs), if erroneous, “was not harmless error.” Accordingly, it is difficult to claim that the state’s answer did not, in fact, fairly apprise Miller and the court that the harmfulness of the suicide note’s admission was at issue in the proceeding. Nor can it be argued that Miller was actually prejudiced by the state’s purported failure to raise the issue of harmless error as to the suicide note, as she argued that issue in any event. Thus, I would find that the state’s harmless error argument has not been waived.
Finally, even if the technical requirements for waiver are met, many of our sister circuits have concluded that courts have the discretion to conduct harmless-error review sua sponte, even when the government utterly fails to raise the issue at any point. See, e.g., Sanders v. Cotton, 398 F.3d 572, 582 (7th Cir.2005); United States v. Adams, 1 F.3d 1566, 1575-76 (11th Cir.1993); Lufkins v. Leapley, 965 F.2d 1477, 1481 (8th Cir.1992); United States v. Langston, 970 F.2d 692, 704 n. 9 (10th Cir.1992); United States v. Rodriguez Cortes, 949 F.2d 532, 542-43 (1st Cir.1991); United States v. Pryce, 938 F.2d 1343, 1347-48 (D.C.Cir.1991); United States v. Giovannetti, 928 F.2d 225, 227 (7th Cir.1991); cf. Sowell v. Bradshaw, 372 F.3d 821, 830 (6th Cir.2004) (“[T]his court may consider a newly-raised [procedural] default argument, if it so wishes.”).
With no meaningful analysis, the majority elects not to exercise its discretion to overlook waiver; the sole reason which the majority gives for its choice is “the State’s [ostensible] decision not to raise this argument to the district court despite having every reason to do so.... ” Maj. Op. at 927. But this is nothing more than restating the majority’s (questionable) finding that waiver occurred in the first place.
Our sister circuits have recognized that overturning a state-court judgment “may be an excessive sanction for the government’s having failed to argue harmless error,” Giovannetti, 928 F.2d at 227, and that several independent considerations should be analyzed to determine whether it is appropriate to overlook waiver of harmless-error review. Among these are “whether the harmlessness of the error or errors found is certain or debatable,” ibid., whether conducting harmless-error review would “waste[ ] ... judicial resources [by] requirfing] the appellate bench to delve independently into a complex record without the aid of the government’s brief and the defendant’s responses to it,” Rodriguez Cortes, 949 F.2d at 543, and whether “defense counsel’s lack of opportunity to answer potential harmless error arguments may lead the court to miss an angle that would have shown the error to have been prejudicial,” Pryce, 938 F.2d at 1347.
Here, because both parties have briefed the issue fully at the appellate level, overlooking waiver would not waste this court’s judicial resources and would cause no prejudice to Miller. Further, as I discuss below, the harmlessness of any Confrontation Clause error is so obvious that insistence on strict observation of the waiver doctrine would merely “result in protract*937ed, costly, and ultimately futile proceedings in the district court,” thus not helping Miller “beyond such undeserved benefits as [s]he may derive from delay.” Giovannetti, 928 F.2d at 226-27. I see no countervailing considerations in this case that weigh against exercising our prerogative to conduct harmless-error review notwithstanding any supposed waiver.
B. Harmlessness
Because the majority premises its holding that the writ must issue entirely on its finding that harmless-error review was waived, see Maj. Op. at 927-28, it does not reach the issue of harmlessness, other than to observe in a footnote that the prosecution mentioned the suicide note in its closing argument and that the state trial court chose to admit the note under Michigan’s residual hearsay exception, which requires that court to determine that “the statement is more probative on the point for which it is offered than any other evidence” reasonably available. See Mich. R. Evid. 804(b)(7).9
I would find any Confrontation Clause error harmless because the admission of the suicide note did not “ha[ve] a substantial and injurious effect or influence in determining the jury’s verdict.” Fry v. Pliler, 551 U.S. 112, 116, 127 S.Ct. 2321, 168 L.Ed.2d 16 (2007). The most important factor in the harmlessness calculus is the (properly admitted) transcript of the IM conversation from the night before Bruce was killed, in which Cassaday and Miller discussed the impending murder in chilling detail. It is beyond debate that if the jury believed this IM transcript was authentic, the case would be open-and-shut, and the suicide note would have been entirely cumulative.10
And while Miller argues that the transcript could have been forged by Cassaday, I do not believe any reasonable jury could have concluded that the transcript was inauthentic. Among other facts corroborating the IM transcript are the following details:
• Records from America Online introduced at trial established that both Cassaday and Miller were online (and offline and online again, repeatedly) at precisely the time the IM transcript indicates that the exchange took place.
• The IM transcript shows Miller and Cassaday speaking of the necessity of removing Bruce’s wallet, and indeed, his wallet was missing from his body and was never found.
• The IM transcript portrays Miller correctly describing to Cassaday the layout of Bruce’s workplace (e.g., the location of a light switch, the direction in which the door opens), which Cassaday had no apparent reason to know independently.
• According to the IM transcript, Miller planned to call Bruce’s workplace once around 5:00 p.m., then call him back *938and keep him on the phone to insure his solitary vulnerability to the impending murder. Miller would then call Cassaday and “[l]et it ring once” to signal that the coast was clear. Finally, she would call Bruce back and keep him on the phone “just until he says [Cassaday] pulled up” at the door. And indeed, telephone records established that Miller’s land line placed a phone call to Bruce’s office at 4:47 p.m. on the day of the murder; then another at 6:08; then a call at 6:16, which rang but was not answered, to her own cellular phone, presumably then in Cassaday’s possession (and which was never used again thereafter); and then another call to Bruce’s workplace at 6:20 — all at approximately the time when the murder undisputedly occurred.
• According to the IM transcript, during that conversation, Cassaday told Miller to “write ... down” the name and telephone number of an individual (his lawyer, John O’Connor) to contact should anything go awry, and that very telephone number was later found on a pad in Miller’s home, and in her handwriting. Further, according to the IM transcript, Cassaday misspelled O’Connor’s name as “Occonnor” and provided incorrect office and home telephone numbers (the former with one incorrect digit, and the latter with two digits transposed). These very same errors were found on the handwritten note in Miller’s home.
Thus, in order to believe that the IM transcript was inauthentic, the jury would have had to believe that Miller and Cassaday — by remarkable coincidence or by Cassaday’s sinister design — were online simultaneously (perhaps chatting about the weather) at the exact same time Cassaday’s damning IM printout suggests they were online plotting the murder; that Cassaday somehow independently knew that Bruce’s wallet had been taken and independently knew the layout of Bruce’s workplace; that Cassaday somehow had access to Miller’s telephone records, or otherwise was able to guess exactly which phone calls Miller placed surrounding the time of the murder; and that Cassaday somehow contrived to communicate O’Con-nor’s (misspelled) name and (incorrect) telephone numbers to Miller at some other point in time and contrived to have her retain them. No reasonable jury could entertain all of these beliefs.
Furthermore, the record is replete with other legitimately admitted circumstantial evidence of Miller and Cassaday’s murderous conspiracy. First, Miller had very specifically indicated to Cassaday on prior occasions the possibility, desirability, and even necessity, of his killing Bruce. For instance, in the two weeks before the murder, Miller sent Cassaday a thinly disguised “story” she had written about a woman (whose biographical details matched her own) trapped in an abusive marriage, whose husband’s “death ... is [the] answer” to her problems; Miller advised Cassaday to “read[] between the lines” and asked him to “come up with a beautiful ending” to the story. She also sent Cassaday an e-mail in which she “wishfed] that Bruce would not wake up from his nap ... [and that] his heart would stop beating” and stated that “living [again] is what I will do when he dies.” Shortly thereafter, and just before the murder, she wrote to Cassaday that she was “[t]rying to find something or someone to help me end this.” There was also undisputed evidence that Cassaday had warned his brother before the murder that he was going out of town on a mission from which he might not return, and it was undisputed that Cassaday had told Miller in an e-mail that he “[would] be coming back [to Michigan] for the last time” on *939precisely the date that the murder subsequently occurred.
In short, in the words of Charles Dodgson, for the presence of the suicide note to have made any appreciable difference in the jury’s verdict, they would have already had to “believe[ ] ... six impossible things before breakfast.” I would therefore hold that any hypothetical Confrontation Clause error was harmless.
. The majority points out that, while the relevant passage from the Eleventh Circuit's opinion in Newland appears in the leading opinion *930by Judge Tjoflat, Judge Anderson specifically declined to join in the portion of Judge Tjoflat’s opinion in which this passage appears, and Judge Wilson concurred only in the result, such that the choice of Justice O’Connor’s formulation over Justice Stevens’s is not binding law in the Eleventh Circuit. This, however, does not diminish the strength of Judge Tjoflat's analysis (which, in any event, would be at most persuasive authority in this circuit even if it were binding in the Eleventh). I also note that, while Judge Anderson did not join in that portion of Judge Tjoflat’s opinion, he refrained from doing so only because he felt it unnecessary to reach the issue, not because he disagreed with Judge Tjoflat's analysis — in fact, he noted that he ‘’th[ought it] very probably entirely accurate and sound.” Newland, 527 F.3d at 1218 (Anderson, J., concurring).
. That choice is especially appropriate here, as Miller did not take the step of seeking certiorari, which could well have resolved the matter by requiring the state court to confront the new law directly. Many defendants in Miller’s position who did seek certiorari in this time frame had their cases reversed and remanded to state court for consideration in light of Crawford. See, e.g., Goff v. Ohio, 541 U.S. 1083, 124 S.Ct. 2819, 159 L.Ed.2d 245 (2004); Siler v. Ohio, 543 U.S. 1019, 125 S.Ct. 671, 160 L.Ed.2d 494 (2004).
. See also Crater v. Galaza, 508 F.3d 1261, 1263 n. 4 (9th Cir.2007) (”[W]hereas under Teague v. Lane a constitutional principle [is] considered ‘old’ (i.e., clearly established) if it was recognized prior to the petitioner exhausting his direct appeals, under AEDPA a principle is clearly established only if it was recognized by the Supreme Court at the time of the petitioner’s conviction.” (internal citation omitted)); Randy Hertz & James S. Liebman, Federal Habeas Corpus Practice and Procedure § 32.3 (5th ed.2001) ("[AEDPA's] *931choice-of-law rule is stricter than Teague's in [that it] ... limits federal review to legal rules that actually were in effect when the state court decided the case. The statute thus apparently abandons Teague's application on habeas corpus of law announced after the state court decision but before certiorari proceedings in the Supreme Court were completed.”).
. The majority appears to be inconsistent about whether this "may” actually means "must.” On one hand, the majority suggests that in situations where a declarant “manifests actual intent not ‘to bear testimony against the accused,' ” a court "would not need to employ Cromer's objective test,” and could look instead to the manifestations of subjective intent. Maj. Op. at 925-26 n. 9 (quoting Cromer, 389 F.3d at 675). On the other hand, the majority faults the state's compelling arguments that the suicide note was not testimonial for "deflect[ing] the Cromer reasonable-expectation test and [instead] depending] on inferences about Cassaday's actual intent based on the physical evidence.” Maj. Op. at 925. It is entirely unclear when the majority would permit resort to inferences about actual intent and when it would not.
. Notably, this is a definition for whose provenance the Supreme Court cited only an amicus brief by an association of criminal defense lawyers.
. Surely there would be nothing unconstitutional in using the note, or suspicions derived from it, merely to investigate the crime. What, then, does the presence of "investigating” legitimately add?
. Personally, I would interpret "anticipate” as requiring a significant likelihood, though perhaps short of 50%.
. The majority believes these hypothetical are off the mark because "[a] person who writes a suicide note and then throws it into the trash, tears it up, or completely destroys it manifests actual intent not 'to bear testimony against the accused.’ ” Maj. Op. at 925 n. 9 (quoting Cromer, 389 F.3d at 675). Thus, according to the majority, "[o]n those facts, we would not need to employ Cromer’s objective test.” Ibid. This demonstrates the untenable nature of Cromer's bipartite standard (to which I adverted in note 4, supra): on the majority’s interpretation of the Confrontation Clause, a statement might have been made under circumstances in which a reasonable declarant would have expected the statement to be used prosecutorially, yet fail to qualify as testimonial because the declarant manifested subjective intent (however unreasonable) that it not be so used. What result if Cassaday, merely intending to set the record straight before his death, had hand-delivered to the Flint, Michigan chief of police a letter graphically implicating Miller in her husband’s death, with a polite (and genuine) request that it not be used against her? What if Cassaday had delivered the same letter to the chief of police with no special requests, but mentioned to his brother (or some other third party) that he hoped the authorities would elect not to use it against Miller? In either situation, Cassaday would have "manifest[ed] actual intent not 'to bear testimony against [Miller].' ” Would the majority therefore find these hypothetical letters non-testimonial, too? Or would the objective half of the Cromer standard now trump the subjective one— and, if so, why?
. Even if we grant that the state trial judge was correct that the suicide note was "more probative” than any other single piece of evidence in this case, that does not imply that the remainder of the voluminous evidence against Miller would not have led any reasonable jury to convict, even if no one piece of it was as probative as the suicide note.
. This is so notwithstanding the vague and contradicted testimony that Brace’s co-worker, John Hutchinson, had threatened to "dispose of” him. See Maj. Op. at 917. For the jury to have found that Hutchinson had committed the murder independently and that Miller was blameless, it would have had to believe that Hutchinson, by the most galactically improbable string of coincidences, killed Bruce at the same time Cassaday was in Flint on the fateful mission about which he had spoken to his brother and Miller in advance, and in precisely the manner that Miller and Cassaday had discussed the night before. See infra at 938-39.