Derrick Anthony DeBruce v. Commissioner, Alabama Department of Corrections , 758 F.3d 1263 ( 2014 )

TJOFLAT, Circuit Judge,

concurring, in part, and dissenting, in part.

Derrick DeBruce’s lawyers may or may not have been ineffective. I don’t know, and my colleagues don’t either. We can’t know— despite two decades of litigation in state and federal court—because DeBruce failed to develop a record of what his attorneys did (or did not do) in preparation for the penalty phase¹ of his capital murder trial. Without a thorough record, it is impossible for DeBruce to overcome the “strong presumption that counsel’s conduct [fell] within the wide range of reasonable professional assistance,” Strickland v. Washington, 466 U.S. 668, 689, 104 S.Ct. 2052, 2065, 80 L.Ed.2d 674 (1984), and impossible for this court to say—as it must to grant relief under § 2254 — that the State court’s application of Strickland was “objectively unreasonable.” Williams v. Taylor, 529 U.S. 362, 409, 120 S.Ct. 1495, 1521, 146 L.Ed.2d 389 (2000).

The State court denied relief because “DeBruce failed to meet his burden on the majority of his claims,” most of which “were merely statements that were not supported by any specific facts or evidence.” DeBruce v. State, 890 So.2d 1068, 1083 (Ala.Crim.App.2003), cert. denied Ex parte DeBruce, No. 1030617 (Ala. Apr. 30, 2004). DeBruce claims his defense team performed an inadequate mitigation investigation, but the record on that point is almost laughably thin, taking up at most a few pages of the collateral hearing transcript. One of DeBruce’s lawyers is totally unaccounted for, id. at 1085 (“We do not know the extent of his involvement in the case or what investigation he conducted in preparation for trial.”), the other lawyer was not questioned “about what investigations he and [his colleague] conducted,” id., and no one at any point asked the lawyers about their overall strategy. The District Court reviewed the same record and reached the same conclusion, reasoning that DeBruce’s “cryptic,” “stream of consciousness,” “incomplete,” and “factually and legally bare” petition was “too eonclu-sory to state an ineffectiveness claim.” DeBruce v. Campbell, No. 04-2669, ECF Doc. 31, at 22 (N.D.Ala. Aug. 4, 2010).

But now our court, evidently possessed of special insight, reverses and grants ha-beas relief. My colleagues do so by inventing a theory of the case that is both factually unsupported and facially implausible: that DeBruce’s seasoned capital-defense lawyers walked into the penalty phase of trial without knowing anything about the man they were defending. That claim would be dubious standing alone, but here it must overcome the “doubly deferential” standard of review federal courts apply when Strickland and AEDPA operate in tandem. Yarborough v. Gentry, 540 U.S. 1, 6, 124 S.Ct. 1, 4, 157 L.Ed.2d 1 (2003). To grant relief under this demanding standard, the majority needs unambiguous evidence that DeBruce’s attorneys were incompetent. There is no such evidence, so instead the majority embellishes the record, disregards AEDPA, and succumbs to the “all too tempting” impulse “for a court, examining counsel’s defense after it has proved unsuccessful, to conclude that a particular act or omission ... was unreasonable.” Strickland, 466 U.S. at 689, 104 S.Ct. at 2065.

The majority’s approach marks a tiresome revival of errors for which our sister circuits have been repeatedly and pointedly reversed. First, instead of holding De-Bruce to the burden of proving that his attorneys acted unreasonably, the majority “assum[es] that counsel was ineffective where the record [is] silent,” contra Burt v. Titlow, — U.S. -, 134 S.Ct. 10, 13, 187 L.Ed.2d 348 (2013), drawing every available inference on DeBruce’s behalf, and even inventing some “facts” of its own, as part of what it calls “our review of the entire state court record” — effectively de novo review. Second, the majority affords no deference to the State court; the majority merely substitutes its own Strickland analysis, thereby “treating] the unreasonableness question as a test of its confidence in the result it would reach under de novo review.” Contra Harrington v. Richter, 562 U.S. 86, 131 S.Ct. 770, 785, 178 L.Ed.2d 624 (2011). And third, in actually evaluating the attorneys under Strickland, the majority relies not on any particular Supreme Court case but merely on its “own sense of ‘prudence,’ and what appears to be [its] belief that the only reasonable mitigation strategy in capital cases is to ‘help’ the jury ‘understand’ the defendant,” an approach the Supreme Court rejected as “flatly inconsistent with Strickland’s recognition that ‘[t]here are countless ways to provide effective assistance in any given case.’ ” Cullen v. Pinholster, — U.S. -, 131 S.Ct. 1388, 1407, 179 L.Ed.2d 557 (2011) (citations omitted, quoting Strickland, 466 U.S. at 689, 104 S.Ct. at 2065).

These errors will do damage beyond this particular case. Because the majority rules on an incomplete record, it is unable to say what the attorneys here did “wrong” with any precision. Today’s decision thus offers no guidance to lawyers in our circuit, who read our opinions to learn what is expected of them, or to our colleagues on the District Court, who must wrangle “the proliferation of ineffectiveness challenges” invited by the majority’s lackadaisical relationship with the evidence. Strickland, 466 U.S. at 690, 104 S.Ct. at 2066. Even more troubling is the majority’s casual second-guessing of a reasoned State court decision, which the majority overturns on little more than a hunch. Federal habeas relief requires that the State court make “an error ... beyond any possibility for fairminded disagreement,” Harrington, 562 U.S. at -, 131 S.Ct. at 787, but here it seems the State court’s only “error” is disagreeing with the majority’s debatable reading of an ambiguous record. Because “AEDPA prevents defendants—and federal courts—from using federal habeas corpus review as a vehicle to second-guess the reasonable decisions of state courts,” I respectfully dissent. Renico v. Lett, 559 U.S. 766, 779, 130 S.Ct. 1855, 1866, 176 L.Ed.2d 678 (2010).

Part I reviews the factual background of the case. Part II highlights deficiencies in the record that should, under AEDPA, preclude habeas relief. Parts III and IY discuss the majority’s application of Strickland’s performance and prejudice elements, respectively.

I.

Because the majority omits much of the context surrounding DeBruce’s trial, I will provide the details of DeBruce’s crime before describing the actions of his attorneys. Reading the majority opinion, you might think DeBruce simply woke up one morning to find himself on trial for capital murder. That isn’t so: the AutoZone robbery — during which DeBruce fatally shot a man named Douglas Battle, in the back, as Battle lay face-down on the floor — was the culmination of a months-long spree of armed robberies committed or attempted by DeBruce and his associates. DeBruce’s lawyers were deeply (and reasonably) concerned that their client’s criminal history would prejudice DeBruce in the eyes of the jury, and the lawyers labored to deprive the prosecutor of opportunities to put on evidence of DeBruce’s other crimes.

This dynamic, present at virtually every stage of trial, is an essential part of any effort “to reconstruct the circumstances of counsel’s challenged conduct, and to evaluate the conduct from counsel’s perspective at the time,” Strickland, 466 U.S. at 689, 104 S.Ct. at 2065, and yet it goes unmentioned in today’s decision. The majority, sitting at a remove of two decades and hundreds of miles, has the luxury of downplaying the less savory aspects of De-Bruce’s background, but his defense lawyers did not have that option, and this disconnect blinds the majority to things that would have been obvious to anyone in the courtroom.

To take just one example, today’s opinion scolds the lawyers for not presenting mitigation evidence about DeBruce’s supposed “resistance to joining gangs,” Maj. Op. at 1276, which sounds reasonable until one recalls that DeBruce was only on trial in the first place because of his association with a prolific armed-robbery gang. A lawyer on the ground would see immediately that the majority’s preferred strategy would be an outright gift to the prosecutor-and would be, no doubt, the basis of yet another ineffectiveness challenge, this time concerning what an obviously terrible idea it is to “invite[ ] the strongest possible evidence in rebuttal” with “[a] heavyhand-ed case to portray [petitioner] in a positive light.” See Wong v. Belmontes, 558 U.S. 15, 25, 130 S.Ct. 383, 389, 175 L.Ed.2d 328 (2009). This kind of backseat defense law-yering pervades the majority opinion.

The broader point is that Strickland claims are meaningless without context. “[T]rial lawyers, in every case, could have done something more or something different,” and “omissions are inevitable.” Chandler v. United States, 218 F.3d 1305, 1313 (11th Cir.2000) (en banc). Whether those omissions are reasonable is a contextual inquiry, turning on the options realistically available to the attorneys, and their reasons for picking some options over others. Accordingly, the “longstanding case-by-case approach to determining whether an attorney’s performance” was deficient requires that we start with a full understanding of the case’s background. Rompilla v. Beard, 545 U.S. 374, 394, 125 S.Ct. 2456, 2469, 162 L.Ed.2d 360 (2005) (O’Connor, J., concurring).

A.

In January 1991, DeBruce and an associate robbed a man named Bryan White on the street. In addition to stealing White’s paycheck, the robbers took a necklace with a distinctive gold pendant in the shape of a pistol. DeBruce evidently liked the necklace, as he continued to wear it while he committed the other crimes that would lead to his arrest for the AutoZone robbery. In August of that year, the owner of a jewelry store in Sylacauga, Alabama, reported being robbed by, among others, a man wearing a necklace with a gold pistol-shaped pendant. A few days later, De-Bruce and five other men attempted to rob the City National Bank in Sylacauga. Armed with handguns, they entered the bank only to abandon the plan and walk out when one of the robbers lost his nerve. Surveillance cameras recorded their arrival and departure, and captured images of DeBruce, still wearing the gold pistol-shaped pendant.

Unsatisfied, the men drove to Talladega and brainstormed places to rob. They settled on a nearby AutoZone, and after driving past the store for reconnaissance, the men devised a plan: still armed, they would enter the store and pretend to browse the merchandise. Eventually their ringleader, a man named Charles Burton, would initiate the robbery, at which point the robbers would draw their handguns and force all the customers and employees to the floor. If anything went awry and an employee or customer became uncooperative and “needed killing or anything,” Burton said he “would take care of it.” State Court — Trial Transcript, Vol. 6, at 911. The men went through with the robbery, and as they made their getaway DeBruce shot and killed one of the customers, Douglas Battle, with whom DeBruce had some kind of verbal altercation.

The police quickly identified DeBruce and his associates as the robbers, in part because the men had already been targets of an investigation (an effort the police, with characteristic portentousness, called the “Trojan Horse Operation”). The police compared their existing surveillance photos with video taken by the security cameras at the City National Bank. One detective noticed that the necklace De-Bruce wore in the video matched the necklace stolen from Bryan White, which had in turn been worn by one of the people who robbed the jewelry store in Sylacauga. Witnesses to the AutoZone robbery identified several of the men, including DeBruce, in police photo arrays. And a fingerprint recovered from the AutoZone matched a print already on file for Charles Burton. Convinced they had their men, the police arrested DeBruce and his associates on August 23, 1991. During a police interrogation, one of the robbers, LuJuan McCants, identified DeBruce as the man who shot Battle. McCants agreed to cooperate with the State and testify against DeBruce.

B.

Charged with capital murder, DeBruce was initially represented by a court-appointed attorney. DeBruce’s mother, Jessie DeBruce, had little faith in the court-appointed lawyer and approached a local criminal-defense attorney, Erskine Mathis. “I didn’t know any of these people,” Mathis later testified at the Rule 32 evidentiary hearing. “[Jessie DeBruce] came in my office on this particular day, and she was scared to death that her boy was going to end up in the electric chair, and she was afraid that the young lawyer who had been appointed to represent him would not do as good a job as somebody that she hired. I felt sorry for her.” State Court — Collateral Transcript, Vol. 15, at 59. Mathis agreed to take over the case.²

Mathis was at that time an established capital-defense attorney. Mathis testified that “90 to 95” percent of his caseload was criminal. Id. at 170. He had earned CLE credits dealing solely with capital litigation, and had handled “probably 10 or 12” capital cases before representing DeBruce. Id. at 171. Up to that point, no client of his had been sentenced to death. “De-Bruce was the very first person I lost to the death penalty,” Mathis testified later. “Most of the other cases were either capital murder verdicts with life without parole sentences or felony murder verdicts. I won one or two during that period of time outright.... I had been, I believe, very successful in capital litigation up to that point.” Id.

Jessie DeBruee’s fears about her son’s court-appointed attorney turned out to be well-founded: when Mathis took over the case, he saw that almost nothing had been done save for the filing of a discovery motion. Id. at 59-60. Mathis “filed a fistful of motions,” id. at 65, and associated a second attorney, William DelGrosso, as co-counsel once he “realized how big a task I had,” id. at 185. The exact nature of DelGrosso’s involvement is not clear from the record because “DelGrosso did not testify or execute an affidavit for purposes of the postconviction proceedings.” DeBruce, 890 So.2d at 1085. Mathis testified that DelGrosso was “a tremendous help,” that he could not recall “any particular division of duties” between them, that he was “sure” DelGrosso looked through the discovery material before trial, that the two attorneys had worked together before, and that DelGrosso had “probably tried as many capital murder cases as I have.” State Court—Collateral Transcript, Vol. 15, at 179-80.

The record is largely silent about what activities Mathis and DelGrosso undertook in the time leading up to trial. We can infer from the record that the lawyers had regular meetings with their client and their client’s mother: in a pretrial meeting with the court, Mathis made passing reference to “some information I was given by this defendant’s mother prior to having the opportunity to speak at length with him,” State Court — Trial Transcript, Vol. 1, at 12, though it is not clear what “some information” refers to. Subsequently, Mathis said, “I have spoke [sic] with him [i.e., DeBruce] many times.” Id.

It appears that whatever “information” Mathis received related to potential mitigation evidence, because Mathis made those comments as part of a request that the court arrange a pre-trial mental evaluation of his client, which Mathis said “would be for the purposes of the punishment phase of the trial ... if it comes down to it.” Id. at 12-13 (emphasis added). We do not know—because Mathis was never asked at the Rule 32 hearing— about the substance of his conversations with his client or his client’s mother, or about his subsequent use of the pre-trial mental evaluation, or about his overall strategy for the penalty phase. All we know is that the court, in response to Mathis’s request, assigned a psychologist, Gary Garner, to conduct a pre-trial mental evaluation of DeBruee, and that his report (hereinafter the “Garner Report”) now plays an outsize role in DeBruce’s postcon-viction litigation.³

The Garner Report is a short document labeled “Mental Status Exam.” It conveys, as relevant to this appeal, that:

• “He [i.e., DeBruee] reports having to be hospitalized on two occasions during his jail stay for what he describes as ‘throwing up.’ ”

• “He reports a history of problems since childhood including alcohol and marijuana abuse.”

• “He has attempted suicide four times.”

• “He denies any prior mental health treatment or psychiatric hospitalization.”

• “Mr. DeBruce admitted to on going [sic] problems in school.”

• “He refused special education and dropped out of school in the seventh grade at age 16.”

• “He admits that his father is probably an alcoholic.”

• “Mr. DeBruce was oriented and aware in all spheres. His memory appeared fair for past and recent experiences and good for immediate impressions. General grasp and recall appeared fair. Mr. DeBruce’s intellectual ability appeared to be in the low average range.”

Garner Report at 1. At the Rule 32 hearing, no one asked Mathis about the Garner Report.

As for the rest of the attorneys’ preparatory activities, we know very little. Mathis was asked at the Rule 32 hearin—a few brief, narrow questions about his investigation of DeBruce’s background, and he gave brief, narrow answers on which DeBruce (by then represented by new lawyers) never followed up. Mathis testified that he had received a banker’s box full of discovery materials and that he had reviewed the contents. Mathis also testified that he had no recollection of seeing a videotape of LuJuan McCants’s police interrogation, but the Rule 32 judge did not find Mathis credible on this point: “The Court finds that Erskine Mathis, contrary to his recollection, viewed the videotape in advance of trial.” See DeBruce, 890 So.2d at 1087 (quoting the circuit court ruling). Finally, Mathis testified about three specific investigative avenues he did not pursue: he did not hire an outside investigator, he did not gather medical records, and he did not gather academic records.

C.

The guilt phase of DeBruce’s trial began on February 11,1992. The defense team’s objective was to avoid the capital murder conviction: they recognized that the State’s case for death was built entirely around LuJuan McCants’s identification of DeBruce as the shooter inside the Auto-Zone. If they could cast doubt on the credibility of that identification — by suggesting that it was in fact McCants who had shot Battle, taken a plea, and then fingered DeBruce to save himself — then DeBruce might be convicted of felony murder, rather than capital murder, and thereby evade a death sentence. But the defense lawyers also understood that impeaching McCants risked opening the door to testimony about all of DeBruce’s criminal activity preceding the AutoZone robbery, which the attorneys recognized to be prejudicial.

Mathis tried to thread the needle. Anticipating that the prosecutor would ask McCants about the earlier robberies, Mathis moved the court to instruct the prosecutor, on direct examination, not to elicit from McCants anything about “any other crimes for which my client was not on trial here today” because “to get involved in these various robberies that they could very well talk about ... would be highly prejudicial to my client in this particular situation and would far outweigh any probative value it may have regarding this crime for which he is on trial.” State Court — Trial Transcript, Vol. 6, at 892.

Mathis also sought to prevent the prosecutor, on redirect examination, “from going into these various [other] crimes ... if and when I attempt to impeach [McCants] based on a[ ] [plea] agreement, which [McCants] ha[s] reached with the State of Alabama.” Id. at 895. Mathis was concerned that the State’s cooperation agreement with McCants encompassed both the AutoZone and City National Bank crimes and even suggested that the “agreement was designed ... so that it would give the prosecuting attorney ... an opportunity to stand in and argue that if I ask about that agreement, I’m opening the door for him to get into the [attempted] robbery in Sylacauga.” Id. at 895-96.

The prosecutor denied that the agreement was so designed and said it was “fine with me to leave out the ... the attempted robbery [of the City National Bank] in the agreement.” Id. at 896. The prosecutor went on to warn, however, that there was a limit. When the police first questioned McCants, he had initially denied knowing anything about the AutoZone or City National Bank crimes. He had only identified DeBruce as the AutoZone shooter after the police said they might charge McCants with the shooting. At trial, the prosecutor warned that if Mathis tried in cross-examination to bring out that McCants had initially denied knowing about the robbery, then the prosecutor would consider the door opened to evidence of other crimes: “if on cross examination on McCants you start talking about his statement of last week, or the week before ... I’m entitled to show that he gave a statement back on August 24th,” id. at 897-98, referring to a videotape of the police interrogation of McCants.

Mathis replied that he had no intention of bringing up McCants’s earlier denials:

MR. MATHIS: Oh, okay. I know what you are talking about. You just don’t want me trying to allege that [McCants] just come up with this story last week.

[PROSECUTOR]: That’s right.

MR. MATHIS: Okay, don’t worry about that.

[PROSECUTOR]: If you do that, then I’m going to try to get into the August statement.

MR. MATHIS: I’m not doing it.

Id. at 898. At the end of this exchange about the scope of McCants’s testimony, the court instructed McCants “not to give any testimony relative to any other crimes that might have occurred except for the AutoZone charge here” or testify about his plea agreement with the State “except the agreement relative to the AutoZone case here in Talladega.” Id. at 899-900.

Having secured this instruction, Mathis attempted to impeach McCants’s testimony without opening the door to testimony about the other crimes. He asked McCants about the existence of the plea agreement, and then asked, point blank, if McCants had shot Douglas Battle. McCants said he had not. Mathis asked if McCants had bragged to DeBruce’s niece, Kiawatta, about shooting Battle. McCants again said he had not. Mathis subsequently called Kiawatta DeBruce, who testified that McCants had confessed to shooting Battle in a conversation after the robbery.

Mathis focused on this alternate-shooter theory in his closing argument, suggesting that McCants had taken a plea and testified against DeBruce in order to save himself, and referenced Kiawatta DeBruce’s testimony to further undermine McCants’s credibility. Later, at the Rule 32 hearing, the prosecutor testified that he thought Mathis had done such a good job undermining McCants that he had been afraid Mathis’s strategy might work. But it didn’t work: on February 13, 1992, the jury convicted DeBruce of capital murder.

D.

The penalty phase of the trial began and ended the next day. The State, in seeking the death penalty, relied on evidence from the guilt phase of trial and on two aggravating factors: that DeBruce had pled guilty on September 13, 1991, to second-degree robbery for the White theft, and that DeBruce had been convicted of capital murder for the killing of Douglas Battle.

The State appeals court found that the defense lawyers’ “strategy in the penalty phase was to beg for mercy.” DeBruce, 890 So.2d at 1093. The defense called DeBruce’s mother, Jessie DeBruce. She testified that Derrick was the youngest of her eleven children; all six of his brothers were either in or had been in prison. When DeBruce was a child, Jessie DeBruce said she worked cleaning houses while DeBruce’s father worked construction, making “a very small amount” of money. At one point she described the family’s rented house as “[u]nfit to live in.” She said that DeBruce graduated from high school and attended the University of Alabama-Birmingham. She said that DeBruce had a son. Jessie DeBruce expressed love for her son, said that there was “good” in him, and begged the jury to spare his life.

Jessie DeBruce had already been excused when she volunteered that she had “something else” to say. Mathis began redirect examination, and she said that DeBruce’s mind “is not good.” State Court—Trial Transcript, Vol. 7, at 1138. She claimed DeBruce had previously been treated for a mental disorder “[bjecause we had a lot of problems with him when he was in school.” Id. at 1139. She went on to describe various incidents during which DeBruce “would get nervous and get to shaking,” along with an incident during which the family called the police “to talk to him.” Id. She said, finally, that her son was “easily led.” Id.

This testimony appears to have caught the attorneys off-guard. The court went into recess and the prosecutor demanded that he be allowed to call Gary Garner, the psychologist who had evaluated DeBruce before trial. The State believed that Garner’s report contradicted what Jessie De-Bruce had said: for example, according to the report, DeBruce had never been treated for a mental disorder. The prosecutor also objected to the statement that De-Bruce was “easily led,” and asked to call one of the cooperating witnesses to testify that after the murder DeBruce had tried to arrange for LuJuan MeCants to “take the murder rap since [MeCants] is a juvenile” — apparently on the theory that this would show DeBruce to be an active, knowledgeable participant in the group’s criminal activities. See id. at 1146-48. DelGrosso opposed both proposals, arguing that the Garner Report was only hearsay and could not be used to impeach Jessie DeBruce’s testimony. The court agreed with the prosecutor, and indicated that Garner should testify. In one of this record’s many unexplained gaps, however, this issue was never resolved: Garner did not testify, and no one mentioned the matter again at trial.

The court allowed DeBruce to make a direct statement to the jury. DeBruce asked the jury to spare his life and allow him to see his son grow up. Id. at 1169— 70. Thereafter, DelGrosso presented De-Bruce’s closing argument to the jury. DelGrosso appealed to the jury as parents, asking them to view DeBruce as a “kid” who did not deserve to die just because he had done something “stupid.” Id. at 1172.

The jury recommended the death penalty by a vote of 11 to 1, and the court scheduled the sentencing for March 13, 1992, ordering the preparation of a presen-tence report in the interim. At sentencing, Mathis argued that the court should sentence DeBruce to a term of life without parole so that he could continue to see his family. The court then addressed De-Bruce, and inquired about his health. De-Bruce said it was “fair.” Mathis said that DeBruce was suffering from “extreme stomach problems,” which his incarceration was aggravating. Id. at 1214. The court sentenced DeBruce to death. The court found the two aggravating circumstances the State relied upon, and one statutory mitigator: DeBruce’s age of twenty.

II.

DeBruce makes the startling assertion that his attorneys failed “to conduct any investigation into available mitigating evidence,” Pet’r’s Br. at 63 (emphasis in original), but his claim is demonstrably false. Mere paragraphs later, in the same brief, DeBruce concedes that, actually, his lawyers interviewed him and interviewed his mother (though DeBruce won’t say what they talked about) and also admits his lawyers arranged for a mental examination and report by a court-appointed psychologist. When the attorneys requested that examination, they told the trial judge it “would be for the purposes of the punishment phase of the trial.” State Court—Trial Transcript, Vol. 1, at 12-13. Thus, even on the scant record before us, it’s clear the lawyers were aware of and acting on “the need for defense counsel to conduct some sort of mitigation investigation,” though the scope of that investigation is unclear. Porter v. McCollum, 558 U.S. 30, 40, 130 S.Ct. 447, 453, 175 L.Ed.2d 398 (2009) (per curiam) (emphasis in original).

The majority ultimately endorses a watered-down version of DeBruce’s claim: that the attorneys should have done more to investigate mitigating evidence. But, incredibly, DeBruce can’t prove that his attorneys didn’t do more, nor can he prove that further investigation would have told the lawyers anything they didn’t already know. The majority thus transgresses a basic principal of habeas: “[t]he petitioner carries the burden of proof.” Pinholster, — U.S. at -, 131 S.Ct. at 1398. This burden has teeth, and DeBruce’s failure to produce a thorough record is—by itself—a robust barrier to relief. See Titlow, — U.S. at -, 134 S.Ct. at 19 (Sotomayor, J., concurring) (“Had [the petitioner] made a better factual record—had she actually shown, for example, that [the defense attorney] failed to educate himself about the case ... then she could well have prevailed.”).

Here, the majority overlooks three enormous gaps in the record. First, we do not know anything about the second-chair attorney, DelGrosso, who may have made investigations not reflected in the record. Second, DeBruce’s case against Mathis is selective and at best circumstantial, with no inquiry into what the attorneys knew about their client or about their overall mitigation strategy. And third, the lawyers were never questioned about information in the Garner Reportan omission that does not stop the majority from concluding, with zero evidence, that the attorneys never read it.

A.

The first and most serious hole in the record is DeBruce’s omission of half his defense team. DeBruce’s briefing implies that Mathis was his only lawyer, when in fact there was a second attorney, William DelGrosso, whom DeBruce has tried to erase from history. DeBruce did not call DelGrosso to testify at the Rule 32 hearing, did not present evidence about DelGrosso’s pretrial activities, did not question Mathis about DelGrosso’s involvement, and does not even mention DelGrosso in his brief, except in the prefatory “Certificate of Interested Persons,” where DelGrosso is identified as “co-counsel for Derrick De-Bruce at trial.” Pet’r’s Br. at C-2. This omission is inexcusable, particularly because it was DelGrosso, not Mathis, who delivered the opening statement and closing argument to the jury at the penalty phase of trial—the phase during which DeBruce says his, attorneys were ineffective. See State Court—Trial Transcript, Vol. 7, at 1105-10, 1171-73.

DeBruce’s failure to account for DelG-rosso is fatal to his ineffectiveness claim. “It should go without saying that the absence of evidence cannot overcome the ‘strong presumption that counsel’s conduct [fell] within the wide range of reasonable professional assistance,’ ” and here we have no evidence at all. Titlow, — U.S. at -, 134 S.Ct. at 17 (citation omitted). For all we know, DelGrosso did everything DeBruce says his attorneys failed to do, and there may be other mitigation investigations we know nothing about. Indeed, Strickland requires that we presume DelGrosso acted reasonably and then look to DeBruce to rebut that presumption. “[W]here the record is incomplete or unclear about [the attorney’s] actions, we will presume that he did what he should have done.” Williams v. Head, 185 F.3d 1223, 1228 (11th Cir.1999). DeBruce cannot rebut Strickland’s presumption because he says nothing at all about DelGrosso, and the inquiry is therefore at an end. The State court saw this clearly:

DeBruce was represented by two attorneys at trial; however, only one attorney testified at the postconviction hearing. DelGrosso did not testify or execute an affidavit for purposes of the postconviction proceedings. We do not know the extent of his involvement in the case or ivhat investigation he conducted in preparation for trial. Neither was Mathis questioned about what investigations that he and DelGrosso conducted.

DeBruce, 890 So.2d at 1085 (emphasis added).

The State court’s findings are more than reasonable; under AEDPA, they are bulletproof. AEDPA only permits federal ha-beas relief in two circumstances: if the State court’s decision was “based on an unreasonable determination of the facts,” or if the State court’s decision was “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by” the Supreme Court. § 2254. DeBruce cannot meet either condition. The State court’s fact-finding is obviously correct: there is simply no evidence, one way or the other, about DelGrosso’s mitigation activities. And it therefore cannot have been unreasonable, as a legal matter, to say that DeBruce failed to carry his burden of proof concerning DelGrosso. There is thus no lawful avenue for federal habeas relief.

The majority sets the State court ruling aside without even mentioning, much less applying, § 2254. Instead, the majority undertakes what it calls “our review of the entire state court record” and constructs, on its own initiative, an argument on De-Bruce’s behalf. That is worth repeating. DeBruce does not argue that DelGrosso rendered ineffective assistance; instead, he simply ignores his second attorney, and hopes that we do the same. Instead of acknowledging this critical flaw in De-Bruce’s petition, the majority — on what is supposed to be our highly deferential posture of AEDPA review — scours the record, sua sponte, and makes the argument for him. That is simply out of bounds. What’s more, the argument the majority writes for DeBruce is exceedingly weak, and would not even rebut the Strickland presumption on de novo review. I will consider first the facts, and then the law.

1.

The State court did not make “an unreasonable determination of the facts” concerning DelGrosso. § 2254(d)(2). The court said — accurately—that “[w]e do not know the extent of his involvement in the case or what investigation he conducted in preparation for trial.” DeBruce, 890 So.2d at 1085. The majority chooses to discount this hole in the record for reasons that do not withstand scrutiny.

First, the majority, having undertaken a “review of the entire state court record,” says that “Mathis was lead counsel and DelGrosso played only a minor role” in DeBruce’s defense. Maj. Op. at 1273. As a threshold matter, I do not know where the majority’s “review of the entire state court record” standard comes from, but it sounds an awful lot like de novo review. It certainly isn’t AEDPA, because § 2254 only asks whether the State court’s factual finding was reasonable, not whether we would make the same finding were we to consider the issue afresh. “[A] state-court factual determination is not unreasonable merely because the federal habeas court would have reached a different conclusion in the first instance.” Wood v. Allen, 558 U.S. 290, 301, 130 S.Ct. 841, 849, 175 L.Ed.2d 738 (2010). In applying its own “review of the entire state court record” instead of § 2254, the majority asks the wrong question.

The majority then delivers the wrong answer, because its conclusion — that “Mathis was lead counsel and DelGrosso played only a minor role” — is an obvious non sequitur. Even assuming DelGrosso played a minor role in the overall defense, it does not follow that he did nothing to prepare specifically for the penalty phase, which is the subject of DeBruce’s Strickland claim. I suppose a placekicker plays “only a minor role” in the grand scheme of a football game, but his efforts nonetheless matter a great deal. Similarly, DelGrosso appears to have focused his energies on the penalty phase, as evidenced by the fact that he delivered the opening statement and closing argument — though of course we can’t be sure, because DeBruce presents no evidence about DelGrosso. In any event, the question under Strickland is not whether DelGrosso’s role was a “minor” or “major” one, but whether he contributed to the attorneys’ mitigation investigation, and if so, how. On that question — which is to say, on the relevant question — the record is totally silent.

Protesting perhaps a little too much, the majority submits no fewer than nine numbered facts to support the (uneontrover-sial) proposition that DelGrosso was the second-chair attorney on the defense team. The majority’s list actually serves to underscore how little we know about DelG-rosso: eight of the nine items are about Mathis, not DelGrosso, and not one of the facts addresses the actual question, which is what DelGrosso did to prepare for the penalty phase of trial. The majority’s “proof’ is thus a parade of irrelevancies: “Mathis’s name alone appears on all of DeBruce’s pretrial motions “Mathis, not DelGrosso, cross examined all of the state’s guilt phase witnesses”; “Mathis presented DeBruce’s case-in-chief during the guilt phase ”; and so on. Maj. Op. at 1273 (emphasis added). Irrelevant facts, delivered in bulk, are still irrelevant. Nothing here establishes, or even suggests, that DelGrosso was ineffective. As the State court rightly concluded, we simply do not know.

The majority’s other argument for ignoring DelGrosso rests on a flagrant misreading of the record. The majority says we needn’t worry about DelGrosso because Mathis testified at the Rule 32 hearing “that DelGrosso’s assistance in investigating for either the guilt or penalty phases consisted of helping Mathis sort through a banker’s box of discovery material.” Maj. Op. at 1272 (emphasis added). The word “consisted” does all the work in that sentence: the reader is meant to understand that Mathis said DelGrosso did nothing beyond looking through the banker’s box. But Mathis never said that; nobody even asked that question. Here is the transcript:

[THE STATE]: Mr. Mathis, did you have the assistance of-you did have the assistance of co-counsel in this case?

[MATHIS]: I got Bill DelGrosso, and he was a tremendous help. [THE STATE]: To what extent did you divide your duties in preparing for the case?

[MATHIS]: You grab a pile and I grab the other. I don’t think there was any particular division of duties. We were just working together trying to get all of this done.

[THE STATE]: Did he look through discovery also?

[MATHIS]: Ma’am, I’m sure he did. We were right there in the same building. We had the same office. We’re in the same building, and each of us had equal access.

[COUNSEL FOR DEBRUCE]: I would object. I would ask that the witness answer whether he knows for a fact or not whether Mr. DelGrosso actually did look through the discovery material.

THE COURT: If he doesn’t remember exactly, he can give you a judgment. Did /all work together on the case?

[MATHIS]: I’m sure he looked through it, Judge.

THE COURT: Tall worked together on other cases?

[MATHIS]: Yeah.

THE COURT: What kind of experience did he have?

[MATHIS]: He’s probably tried as

many capital murder cases as I have.

THE COURT: At that time?

[MATHIS]: Oh yeah.

State Court — Collateral Appeal Transcript, Vol. 15, at 179-80. There is no basis in this testimony for the majority’s claim that DelGrosso’s work “consisted,” in its entirety, of a review of discovery materials. Mathis was responding to a specific question: he was “sure” DelGrosso reviewed the box of documents. But that does not establish — and Mathis did not say — -that this was the only thing DelGros-so did. Nobody asked about that.

In sum, even after its improper “review of the entire state court record,” the majority comes up empty on the question of DelGrosso’s performance. The State court was manifestly correct in saying that we simply do not know what he did.

2.

The State court’s decision was not contrary to, or an unreasonable application of, Supreme Court precedent. § 2254(d)(1). The legal principle here is not complicated or controversial: a habeas petitioner who fails to document his attorney’s actions cannot successfully claim that the attorney was ineffective. The Supreme Court just last term rejected an ineffectiveness claim because of the petitioner’s “failure to present enough evidence of what [the attorney] did or did not do.” Titlow, 134 S.Ct. at 18 (Sotomayor, J., concurring); see also Burger v. Kemp, 483 U.S. 776, 793, 107 S.Ct. 3114, 3125, 97 L.Ed.2d 638 (1987) (refusing to draw an inference on petitioner’s behalf where petitioner “has submitted no affidavit from that lawyer establishing that he would have offered substantial mitigating evidence if he had testified.”); Kimmelman v. Morrison, 477 U.S. 365, 384, 106 S.Ct. 2574, 2588, 91 L.Ed.2d 305 (1986) (“Counsel’s competence ... is presumed, and the defendant must rebut this presumption by proving that his attorney’s representation was unreasonable under prevailing professional norms and that the challenged action was not sound strategy.” (emphasis added, citations omitted)).

This rule is a familiar one to Alabama judges. “It is extremely difficult, if not impossible, to prove a claim of ineffective assistance of counsel without questioning counsel about the specific claim, especially when the claim is based on specific actions, or inactions, of counsel that occurred outside the record.” Broadnax v. State, 130 So.3d 1232, 1255 (Ala.Crim.App.2013). And I know for a fact that this is the rule our court applies: “An ambiguous or silent record is not sufficient to disprove the strong and continuing [Strickland ] presumption.” Chandler, 218 F.3d at 1314 n. 15.

Sometimes this rule is harsh. In one ease, the petitioner’s attorney died before the collateral hearing, and the petitioner was thus physically incapable of questioning the attorney about his actions at trial. See Callahan v. Campbell, 427 F.3d 897, 933 (11th Cir.2005). We nonetheless denied relief: “Because [the lawyer] passed away before the Rule 32 hearing, we have no evidence of what he did to prepare for the penalty phase of [the petitioner’s] trial. In a situation like this, we will presume the attorney did what we should have done, and that he exercised reasonable professional judgment.” Id. (footnotes and citations omitted). One judge on the Callahan panel expressed serious reservations about the adequacy of counsel’s mitigation investigation, noting “[t]here is no evidence [the lawyer] performed any substantial investigation into [the petitioner’s] background, or attempted to call family members other than [the petitioner’s] aunt,” but there was no evidence either way, and “since we are bound by our circuit precedent, we presume [the lawyer] did” what he should have done. Id. at 938-39 (Wilson, J., concurring) (emphasis added). The same reasoning applies to DelGrosso in this case.

B.

We know more about Mathis than we do about DelGrosso, but we do not know enough to overcome the Strickland presumption. DeBruce’s entire case against Mathis is drawn from about a page and a half of superficial Rule 32 testimony. Mathis was asked about three things he did not do, but was never asked about his overall strategy for the penalty phase, or about how much he learned about De-Bruce’s life from other sources (like De-Bruce himself, or DeBruce’s mother), or even whether he would have put on a different mitigation case if he could do it all over again. The majority nonetheless treats this testimony as incontrovertible proof of ineffectiveness, and adopts De-Bruce’s theory of the case wholesale.

DeBruce’s theory is as follows: ‘When asked to explain his failure to investigate, Mathis cited time and resource constraints — not strategic concerns.” Pet’r’s Br. at 67 (emphasis added). DeBruce then cites the Rule 32 transcript. A casual reader would take “failure to investigate” to refer to the whole scope of an attorney’s investigation; DeBruce means us to infer that behind the citation we will find Mathis both conceding a total “failure to investigate” and attributing it to a lack of time. Here, then, is the entirety of the testimony that supposedly establishes a total “failure to investigate”:

Q: And do you normally have a social worker or investigator or other specialist help you gather evidence?

[MATHIS]: We normally have an investigator who does that, an investigator plus the other people in my office.

Q: And does that person normally gather the client’s school records?

[MATHIS]: Yes. We would normally subpoena-we would find out from the client which schools he attended and immediately issue subpoenas for all the school records, and we would do that and set a date way ahead of trial so that the stuff would be brought to the courthouse so we could get it.

Q: And does that person normally interview siblings and pastors and teachers and other litigation witnesses?

[MATHIS]: Yes.

Q: And did you have such a person working with you on this case?

[MATHIS]: No. We just didn’t have time. We did not have the time. We didn’t do it.

Q: Did you have time to gather Mr. DeBruce’s school records yourself?

[MATHIS]: None.

Q: You didn’t have a strategic reason for not doing so?

[MATHIS]: None.

Q: Did you have time to gather medical records yourself?

[MATHIS]: We got in it too late. We did good to prepare with what we did do.

Q: You didn’t have a strategic reason?

[MATHIS]: No.

State Court—Collateral Transcript, Vol. 15, at 136-37.

Now, bearing in mind that this is the only testimony about the mitigation investigation, we have three factual claims: (1) Mathis did not hire an outside investigator; (2) Mathis did not collect school records; and (3) Mathis did not collect medical records. DeBruce cites (and the majority credits) this testimony as establishing a complete “failure to investigate,” on all fronts, by both attorneys. Pet’r’s Br. at 67.

Plainly, those things are not the same. Are they close? Maybe. Relevant? Surely. But dispositive, such that we can say that DeBruce carried his burden? Absolutely not. First of all, the argument is highly selective: it fails to account for the possibility that the attorneys took steps other than the three Mathis was asked about, or that the attorneys learned of the relevant information from other sources. Second, DeBruce’s questioning of Mathis sheds no light on the attorneys’ strategy for the penalty phase, or on whether De-Bruce’s supposedly “new” mitigation evidence would have caused them to change their strategy.

1.

If DeBruce wants to hang his hat on the claim that his attorneys did nothing, Strickland requires that he actually account for their time. It is not enough to cherry-pick particular omissions and then extrapolate from those a total “failure to investigate.” See Pet’r’s Br. at 67. “The [Strickland] presumption impacts on the burden of proof and continues throughout the case, not dropping out just because some conflicting evidence is introduced.” Chandler, 218 F.3d at 1314 n. 15. The need for reviewing courts to be “vigilant and independent,” Harrington, 562 U.S. at -, 131 S.Ct. at 780, in holding petitioners to their burden is especially acute when we evaluate attorney-ineffectiveness claims. No one is better positioned to document the activities of a defendant’s attorneys than the defendant himself: he has privileged access to their work product, and his recollections can be crucial if—as frequently happens—new attorneys take over the case on collateral attack. Equally obvious is the danger that, if petitioners are not required to produce a thorough record, they might be tempted to omit information that weakens their claim of ineffectiveness. Strickland’s requirement that we presume effectiveness until we are convinced otherwise plays an important role in protecting the integrity of habeas proceedings: by enforcing it, courts ensure honest and meaningful review. Lax enforcement, on the other hand, “would encourage the proliferation of ineffectiveness challenges. Criminal trials resolved unfavorably to the defendant would increasingly come to be followed by a second trial, this one of counsel’s unsuccessful defense.” Strickland, 466 U.S. at 690, 104 S.Ct. at 2066.

Here, brief testimony that Mathis did not take a particular action does not establish that the attorney (or his unmentioned co-counsel) took no action at all, nor does Strickland allow us to extrapolate from a decision not to take action X that the attorney must have also not taken action Y or Z. Mathis’s testimony certainly does not establish that “Mathis [or DelGrosso] did not interview any members of DeBruce’s family,” Pet’r’s Br. at 64, or that “Mathis [or DelGrosso] never examined the [Garner] report,” id. at 65, claims that DeBruce asserts (and the majority accepts) without citation. Nor does the testimony address the possibility that the attorneys pursued avenues of investigation other than the three narrow ones Mathis was asked about. See Titlow, — U.S. at - -, 134 S.Ct. at 18 (“The record does not reveal how much [the defense attorney] was able to glean about respondent’s case from other sources; he may well have obtained copies of the critical materials from prosecutors or the court.”). Mathis was not even able to say definitively that the unnamed “other people in [his] office” did not help him prepare for trial: “I think at that time all we had was me and Bill [DelGrosso]. I don’t think we had anybody else helping us on this case.” State Court— Collateral Transcript, Vol. 15, at 180-81 (emphasis added). The majority-in a neat demonstration of its methodology — says that Mathis’s hazy recollections make it “clear that the only person Mathis had helping him was DelGrosso,” Maj. Op. at 1272 (emphasis added), even though the Supreme Court rejected an identical argument in Pinholster: “[A] handful of post-hoc nondenials by one of [petitioner’s] lawyers” is not enough to overcome Strickland’s presumption that both lawyers performed effectively. See — U.S. at -, 131 S.Ct. at 1406.

The attorneys may well have gotten everything they needed from their client, or their client’s mother. We know that De-Bruce met with his lawyers “many times,” but we do not know what they talked about. See State Court-Trial Transcript, Vol. 1, at 12. The content of these conversations is important because the reasonableness of a lawyer’s investigation is shaped by information and guidance from his client. See Strickland, 466 U.S. at 691, 104 S.Ct. at 2066 (“In particular, what investigation decisions are reasonable depends critically on [information supplied by the defendant].”). DeBruce offers no documentary record of his attorneys’ investigation — no billing records, no time sheets, no notes, nothing- — -and so he has to rely on testimonial evidence. The universe of possible witnesses to the attorney-client conversations is a limited one: it could come either from the two lawyers, from DeBruce’s mother (who retained Mathis), or from DeBruce himself.

Neither DeBruce nor his mother testified at the Rule 32 proceeding. In a criminal case, of course, DeBruce’s failure to testify would be irrelevant. See U.S. Const, amend. V; Griffin v. California, 380 U.S. 609, 85 S.Ct. 1229, 14 L.Ed.2d 106 (1965). But habeas corpus is a civil action, Hilton v. Braunskill, 481 U.S. 770, 776, 107 S.Ct. 2113, 2119, 95 L.Ed.2d 724 (1987) (“Our decisions have consistently recognized that habeas corpus proceedings are civil in nature.”), and the burden of proof shifts from the State to the petitioner, who has to prove “that counsel made errors so serious that counsel was not functioning as the ‘counsel’ guaranteed the defendant by the Sixth Amendment,” Strickland, 466 U.S. at 687, 104 S.Ct. at 2064. DeBruce claims that his attorneys knew nothing about his background when they made decisions about mitigation, but to prove that he would need to shed some light on what he (or his mother) told the attorneys.

Presumably, DeBruce chose not to testify at the Rule 32 hearing because he wanted to avoid cross-examination about other crimes — like the attempted armed robbery of the City National Bank in Sylacauga— and because Alabama does not have a statute of limitations for violent crimes. See Ala.Code § 15 — 3—5(a)(2). Nonetheless, DeBruee knows what he said to his lawyers and what they said to him, and he has not shared that information with the Rule 32 court or with us. Nor did he call his mother or DelGrosso as witnesses, or ask Mathis about the conversations they had before trial. It was within DeBruce’s power to document these critical attorney-client conversations, and he failed to do so.

At common law, a plaintiff who failed to call an essential witness made himself vulnerable to the so-called “missing witness rule,” whereby a court can draw an inference against a party who inexplicably fails to produce a witness whose testimony would be crucial to resolving the case. See Graves v. United States, 150 U.S. 118, 121, 14 S.Ct. 40, 41, 37 L.Ed. 1021 (1893) (“The rule, even in criminal cases, is that, if a party has it peculiarly within his power to produce witnesses whose testimony would elucidate the transaction, the fact that he does not do it creates the presumption that the testimony, if produced, would be unfavorable.”); Creel v. Comm’r, 419 F.3d 1135, 1142 (11th Cir.2005) (“It is well settled that the production of weak evidence when strong is available can lead only to the conclusion that the strong would have been adverse.”) (quoting Raley, Inc. v. Kleppe, 867 F.2d 1326, 1329 (11th Cir.1989)).⁴ Strickland embodies a modern missing witness rule insofar as it operates as a kind of evidentiary presumption: “[CJounsel is strongly presumed to have rendered adequate assistance and made all significant decisions in the exercise of reasonable professional judgment,” and a petitioner cannot prevail unless he convincingly proves otherwise. 466 U.S. at 690, 104 S.Ct. at 2066.

Because DeBruee did not produce the available witnesses — himself, his mother, or DelGrosso — or ask Mathis about his mitigation strategy, DeBruee is unable to rebut the possibility that “[t]rial counsel could reasonably surmise from his conversations with [the defendant] that character and psychological evidence would be of little help.” See Strickland, 466 U.S. at 699, 104 S.Ct. at 2071. Likewise, he is unable to substantiate the claim that his attorneys were “unaware of [his] upbringing,” Maj. Op. at 1276, because he presents no evidence about what his attorneys knew before trial. DeBruce’s pre-trial interactions with his attorneys are a black box, and his inability — or unwillingness — to document those interactions makes it impossible to rebut the “strong presumption that [the attorneys acted] for tactical reasons rather than through sheer neglect.” See Yarborough, 540 U.S. at 8, 124 S.Ct. at 5 (citing Strickland, 466 U.S. at 690, 104 S.Ct. at 2066).

2.

DeBruce’s questioning of Mathis at the Rule 32 hearing is deficient in another respect: it sheds no light on the attorneys’ strategy for the penalty phase. An omission, standing alone, is meaningless under Strickland; what matters is the reasonableness of the omission, which turns “on the facts of the particular case, viewed as of the time of counsel’s conduct.” 466 U.S. at 690, 104 S.Ct. at 2066 DeBruce never asked Mathis about strategy, and is thus unable to disprove the presumption that the lawyers had good reasons for acting as they did. Cf. Massaro v. United States, 538 U.S. 500, 505, 123 S.Ct. 1690, 1694, 155 L.Ed.2d 714 (2003) (“The appellate court may have no way of knowing whether a seemingly unusual or misguided action by counsel had a sound strategic motive or was taken because the counsel’s alternatives were even worse.”). Without a sufficiently developed factual record, a petitioner is merely asking the court to speculate about an attorney’s activities or motives, an exercise forbidden to us by Strickland and AEDPA. Cf. Guinan v. United States, 6 F.3d 468, 473 (7th Cir.1993) (Easterbrook, J., concurring) (“[T]he absence of a complete record prevents definitive action. No matter how odd or deficient trial counsel’s performance may seem, that lawyer may have had a reason for acting as he did.”) (quoted with approval in Massaro, 538 U.S. at 505, 123 S.Ct. at 1694). Here, we cannot say the attorneys acted unreasonably because we do not know anything, one way or the other, about their reasoning.

C.

Instead of holding DeBruce to his burden of proof and critically examining whether what DeBruce says the evidence shows is congruent with what the evidence actually is, the majority accepts the claims in DeBruce’s brief in their entirety. Actually, the majority goes further, inventing “facts” that even DeBruce did not assert. According to the majority, the defense attorneys “fail[ed] to take even the first step of asking DeBruce and his mother about the information in the [Garner] report.” Maj. Op. at 1274. This is total speculation: there is exactly zero testimony to that effect, and DeBruce did not make the claim in his brief. The exasperating thing is that the majority could be right — but if they are, it is because they guessed, not because DeBruce presented any evidence establishing the point.

Alternatively, the majority could be dead wrong: maybe Mathis and DelGrosso did ask DeBruce and his mother about information in the report. Maybe, as the State suggests, DeBruce and his mother actively misled the attorneys about their client’s background and told Mathis that the report was nonsense, leading Mathis to “tailor[] his investigation decisions to the information he received from his client and his client’s mother,” see Gov’t. Br. at 45-46, something the Supreme Court in Strickland explicitly said attorneys should do, 466 U.S. at 691, 104 S.Ct. at 2066 (“Counsel’s actions are usually based, quite properly, on informed strategic choices made by the defendant and on information supplied by the defendant.”), and something our circuit has found dispositive in dismissing ineffective-assistance claims, see, e.g., Williams v. Head, 185 F.3d 1223, 1237 (11th Cir.1999) (“An attorney does not render ineffective assistance by failing to discover and develop evidence of childhood abuse that his client does not mention to him.”).

Or, as long as we’re indulging baseless speculation, maybe Mathis and DelGrosso asked DeBruce and his mother about the report and learned information that led them to believe a mental health mitigation strategy would be counterproductive. See Strickland, 466 U.S. at 691, 104 S.Ct. at 2066 (“[W]hen a defendant has given counsel reason to believe that pursuing certain investigations would be fruitless or even harmful, counsel’s failure to pursue those investigations may not later be challenged as unreasonable.”). Or maybe the attorneys followed up on the report only to conclude—given what they knew of the jury, or of their client, or of the evidence they expected to uncover — that it would make for a weak argument and that they would be better off trying something else. See id. (“[W]hen the facts that support a certain potential line of defense are generally known to counsel because of what the defendant has said, the need for further investigation may be considerably diminished or eliminated altogether.”); see also Harrington, 562 U.S. at -, 131 S.Ct. at 788 (“Unlike a later reviewing court, the attorney observed the relevant proceedings, knew of materials outside the record, and interacted with the client, with opposing counsel, and with the judge.”).

We just don’t know: Mathis was not asked about the Garner Report, DelGrosso did not testify, and DeBruce does not offer any other information. That should be the end of it. Instead, the majority rewards DeBruce for his incomplete presentation and overturns a reasoned State court decision on nothing more than the majority’s self-proclaimed powers of divination. We should be more than a claim-laundering service for underdeveloped habeas petitions; if we were going to let DeBruce write his own habeas writ, we could have done that without two decades of inconclusive litigation.

III.

The majority today has misapplied AEDPA in a general sense by failing to hold DeBruce to his burden of proof under § 2254(d), but the majority also misapplies AEDPA in the specific context of Strickland. The Court of Criminal Appeals concluded that DeBruce’s attorneys pursued a “beg for mercy” strategy at the sentencing phase of trial. See DeBruce, 890 So.2d at 1093-94. Because the majority believes, in its independent judgment, that a competent lawyer would have gone all-out in an effort to humanize DeBruce and generate sympathy for him in the eyes of the jury, the majority reverses a State court decision as “unreasonable” only in the sense that the majority would have decided De-Bruce’s case differently had they been sitting on the State bench.

The majority’s casual second-guessing of the Court of Criminal Appeals ignores clear Supreme Court guidance on the application of AEDPA to Strickland claims. Our task is first to make a sincere effort “to reconstruct the circumstances of counsel’s challenged conduct, and to evaluate the conduct from counsel’s perspective at the time,” Strickland, 466 U.S. at 689, 104 S.Ct. at 2065, followed by an equally sincere effort to “determine what arguments or theories supported” or “could have supported [ ] the state court’s decision,” Harrington, 562 U.S. at -, 131 S.Ct. at 786 (emphasis added). We are “required not simply to give [the] attorneys the benefit of the doubt, but to affirmatively entertain the range of possible reasons [the petitioner’s] counsel may have had for proceeding as they did.” Pinholster, — U.S. at -, 131 S.Ct. at 1407 (first alteration in original, quotation marks and citations omitted). I will first evaluate DeBruce’s petition according to this guidance from the Supreme Court, and then show how the majority bypasses these steps by substituting its own Strickland analysis for that of the State court.

A.

Reconstructing events from the attorneys’ perspective at the time, we start with Mathis’s testimony that — based on his interactions with his client — he did not believe anything was mentally or physically wrong with DeBruee. “[H]e seemed about as attentive as people would [be] under the circumstances,” Mathis said. State Court-Collateral Transcript, Vol. 15, at 173. “We would all of us criminal defense lawyers on a capital murder case like to have a client with 120 IQ who could talk to us like we could talk to them, but unfortunately that never happens. So you get used to it in this business. And I assumed that Derrick was not real bright, that he could get by, but that he was not real bright, that I was going to have to do most of this myself, and reacted accordingly so.” Id. at 174. The trial judge corroborated this observation at the Rule 32 hearing: “I sat here and watched [DeBruee], and he wasn’t in pain when I was trying him that week. He would rather have been somewhere else I’m sure, but there wasn’t physically anything wrong with him. I observed him throughout the trial. He even got up and made part of the closing argument in the sentencing phase. Didn’t appear to have any difficulty.” Id. at 276.

But rather than relying on his observation of DeBruee or his conversations with his client (Mathis said that he “spoke with [DeBruee] many times,” State Court — Trial Transcript, Vol. 1, at 12), Mathis asked the court to have DeBruee evaluated. The record shows that Mathis was motivated to ask for the evaluation because of “some information I was given by this defendant’s mother prior to having the opportunity to speak at length with him.” Id. Mathis was plainly attentive to the possibility that whatever “some information” was would be relevant to the penalty phase, because he made clear that the evaluation — what would become the Garner Report — -“would be for the purposes of the punishment phase of the trial ... if it comes down to it.” Id. at 12-13. Because there is no evidence to the contrary, Strickland requires us to assume that Mathis and DelGrosso reviewed the Garner Report and considered its usefulness as they planned for trial. See 466 U.S. at 690, 104 S.Ct. at 2066.

Here the attorneys would come to a fork in the road, and we have to ask: having reviewed the report, could reasonable attorneys make the decision to pursue a beg-for-mercy strategy in the penalty phase, as opposed to a humanization defense or a defense based on mental health? The Supreme Court, our court, and the Alabama Supreme Court have all held that they could. Attorneys are not required to exhaust every line of investigation before making tactical decisions. “There comes a point where a defense attorney will reasonably decide that another strategy is in order, thus ‘mak[ing] particular investigations unnecessary.’” Pinholster, — U.S. at -, 131 S.Ct. at 1407 (alteration in original, quoting Strickland, 466 U.S. at 691, 104 S.Ct. at 2066). The Supreme Court in Burger noted that “[t]he record at the habeas corpus hearing does suggest that [the defense attorney] could well have made a more thorough investigation than he did,” but the Court held he was not required to. 483 U.S. at 794-95, 107 S.Ct. at 3126. “[T]here comes a point at which evidence ... can reasonably be expected to be only cumulative, and the search for it distractive from more important duties.” Bobby v. Van Hook, 558 U.S. 4, 11, 130 S.Ct. 13, 19, 175 L.Ed.2d 255 (2009).

Our court has said that “[b]y its nature, ‘strategy’ can include a decision not to investigate ... [and] a lawyer can make a reasonable decision that no matter what an investigation might produce, he wants to steer clear of a certain course.” Rogers v. Zant, 13 F.3d 384, 387 (11th Cir.1994). We do not require attorneys to “pursue every path until it bears fruit or until all hope withers.” Williams, 185 F.3d at 1237 (quoting Foster v. Dugger, 823 F.2d 402, 405 (11th Cir.1987)). “Counsel is not required to present every nonfrivilous defense; nor is counsel required to present all mitigation evidence, even if the additional mitigation evidence would not have been incompatible with counsel’s strategy.” Chandler, 218 F.3d at 1319; see also Tarver v. Hopper, 169 F.3d 710, 715 (11th Cir.1999) (noting that lawyers are “not required to investigate and present all available mitigating evidence to be reasonable”); Waters v. Thomas, 46 F.3d 1506, 1511 (11th Cir.1995) (en banc) (“Our decisions are inconsistent with any notion that counsel must present all available mitigating circumstance evidence, or all available mental illness or impairment evidence, in order to render effective assistance of counsel at the sentencing stage.”). If we ourselves have given such wide latitude to defense counsel, and if the Supreme Court has repeatedly rejected “the notion that the same investigation will be required in every case,” Pinholster, — U.S. at -, 131 S.Ct. at 1406-07, then it cannot have been unreasonable for the Court of Criminal Appeals to say here that DeBruce failed to show that his attorneys were incompetent.

Things might be different if a client is “fatalistic and uncooperative.” Porter v. McCollum, 558 U.S. 30, 40, 130 S.Ct. 447, 453, 175 L.Ed.2d 398 (2009) (per curiam). If an attorney has reason to think his client is not forthcoming, he might need to take further steps to uncover things his client is trying to conceal. Certainly, he will need to have more than “one short meeting” with his client before sentencing, as the inexperienced defense lawyer in Porter did. Id. at 39, 130 S.Ct. at 453. But DeBruce was not fatalistic or uncooperative; the evidence we have shows him participating extensively in his own defense. He met many times with Mathis before trial. He addressed the court directly on a mid-trial suppression motion. And he addressed the jury on his own behalf in the penalty phase of the trial. The Court of Criminal Appeals could have reasonably believed that DeBruce’s lawyers, in consultation with their client, considered and rejected the mitigation strategy that DeBruce now, in retrospect, urges as the only viable path forward. Debruce presents no evidence to prove otherwise.

Bearing in mind that “[t]here are countless ways to provide effective assistance in any given case,” Strickland, 466 U.S. at 689, 104 S.Ct. at 2065, we also must ask whether Mathis and DelGrosso might have had reason to be wary of focusing on De-Bruce’s background at sentencing. Competent lawyers could have many reasons for not taking the humanization route. Mathis and DelGrosso — both experienced attorneys familiar with local conditions and juries — might have concluded that the jury would not find a defense based on De-Bruce’s background persuasive. Take, for instance, DeBruce’s new mental health mitigation evidence. At the Rule 32 hearing, DeBruce contended that he suffers from brain damage and offered an expert, Dr. David Williamson, to testify that he found “clear evidence for brain dysfunction from Mr. DeBruce.” State Court—Collateral Transcript, Vol. 17, at 32. The State responded with its own expert, Dr. Glen King, who testified that DeBruce “suffers from no serious mental illness or mental defect,” id. at 232, and that, in Dr. King’s opinion, DeBruce “was malingering psychological and psychiatric symptoms,” id. at 237' — 'that is, that DeBruce was faking it to evade punishment.

“Evaluating] the conduct from counsel’s perspective at the time,” Strickland, 466 U.S. at 689, 104 S.Ct. at 2065, the lawyers would have been aware of the risks of a mental health mitigation defense. Maybe the attorneys would get lucky, and the jury would patiently endure long stretches of tedious cross-examination as the parties labored to discredit the other side’s expert. See, e.g., State Court-Collateral Transcript, Vol. 17, at 259-86. Maybe the jury would gladly dive into the details of the Minnesota Multiphasic Personality test, id. at 233, the Instructed Interview Reported Systems test, id. at 237-38, and the Wechsler Adult Intelligence Scale, id. at 236. And maybe, after extended squabbling about standardization procedures and norming and clinical validity, the jury would credit the defense expert over the State expert and conclude that DeBruce is mentally impaired.

But maybe they wouldn’t. Maybe the jury would find the State expert more credible, or maybe the jury would simply find the battle of the experts exhausting and default to their own observations of DeBruce in the course of trial. Now the defense could be in real trouble: they have exposed their client, whom the jury just convicted of a cold-blooded murder, to accusations of faking brain dysfunction to avoid responsibility for his crime. In doing so, the attorneys might undermine their own credibility in front of the jury and undercut whatever sympathy jurors might feel for their client. The mental health mitigation strategy that DeBruce advocates in hindsight would not have been risk-free from the perspective of his attorneys at the time. See Harrington, 562 U.S. at -, 131 S.Ct. at 790 (finding that counsel was not ineffective for foregoing expert testimony where “there was the possibility that expert testimony could shift attention to esoteric matters of forensic science, distract the jury ... or transform the case into a battle of the experts.”).

Obviously, different lawyers will evaluate these risks differently; “[e]ven the best criminal defense attorneys would not defend a particular client in the same way.” Strickland, 466 U.S. at 689, 104 S.Ct. at 2065. That is why Strickland does not allow us to assume that certain kinds of mitigation strategies are obviously superior to others. “The current infatuation with ‘humanizing’ the defendant as the be-all and end-all of mitigation disregards the possibility that this may be the wrong tactic in some cases because experienced lawyers conclude that the jury simply won’t buy it.” Pinholster, - U.S. at -, 131 S.Ct. at 1408 (quoting Pinholster v. Ayers, 590 F.3d 651, 692 (9th Cir.2009) (en banc) (Kozinski, C.J., dissenting) rev’d — U.S. at -, 131 S.Ct. at 1388). “[A]pplying a heavy measure of deference to counsel’s judgments,” Strickland, 466 U.S. at 691, 104 S.Ct. at 2066, we cannot discount the possibility that DeBruce’s attorneys considered these risks and decided against the approach DeBruce and the majority now endorse. And because DeBruce offers no evidence to rebut that possibility, we certainly cannot say the Court of Criminal Appeals was unreasonable for thinking that “counsel made such a reasoned decision in this case.” Pinholster, — U.S. at -, 131 S.Ct. at 1408.

DeBruce’s lawyers appear to have decided that the best course was to try and elicit sympathy not for their client but for his family. See id. at -, 131 S.Ct. at 1407 (“[I]t certainly can be reasonable for attorneys to conclude that creating sympathy for the defendant’s family is a better idea because the defendant himself is simply unsympathetic.”). Thus, DeBruce’s lawyers offered testimony from the defendant’s mother: “I’m asking y’all to please to think about me and my child. I love him and you love yours and I’m begging y’all try not to give him the death penalty. Let him live.” State Court — Trial Transcript, Vol. 7, at 1136-37.⁵ DeBruce’s own statement to the jury centered on his affection for his mother and his own child: “I have a son who is one year old and my mother. I love my mother. I love my son.” Id. at 1169. In his closing, DelGrosso appealed to the jurors as parents: “How many of you have kids that have done the most absolute stupid things? They go out and do it and it is stupid. It’s just stupid, but do we kill that kid? No, we don’t kill him.” Id. at 1172. The prosecutor certainly understood this as a family sympathy defense. He began his closing statement by acknowledging that the jury had “just witnessed as good a guilt trip, as heavy a burden that anybody can place upon your shoulders.” Id. at 1174. He responded directly to DeBruce’s appeal to family: “I love my mother; I love my son.... All of those things apply to Doug Battles [sic]. All of those things apply to the Battles [sic] family.” Id. at 1176.

Focusing on DeBruce’s family also deprived the State of the opportunity to respond with additional evidence in rebuttal. See Wong v. Belmontes, 558 U.S. 15, 25, 130 S.Ct. 383, 389, 175 L.Ed.2d 328 (2009) (“The type of ‘more-evidence-is-better’ approach advocated by [defendant] ... might seem appealing — after all, what is there to lose? But here there was a lot to lose. A heavy-handed case to portray [defendant] in a positive light ... would have invited the strongest possible evidence in rebuttal.”). Expert testimony about DeBruce’s mental condition would invite rebuttal evidence from the State’s expert that De-Bruce was faking his condition to evade punishment. Testimony about DeBruce’s background and upbringing would open the door to testimony about DeBruce’s earlier criminal activities, including the spree of robberies or attempted robberies leading up to Douglas Battle’s murder. Testimony about DeBruce’s family might focus attention on DeBruce’s siblings, many of whom were also incarcerated and who might not make ideal character witnesses, or who might undercut sympathy for DeBruce’s mother by describing their childhoods as hellish. DeBruce’s sister Ja-quelline, for example, testified at the Rule 32 hearing that their mother approved of the eldest sister Juanita’s heavy-handed disciplining of the other children: “My mother said she left [Juanita] in control over us and she can chastise us as she choose.” State Court — Collateral Transcript, Vol. 15, at 355.

The defense team had to pick: they could appeal to the jury’s sympathy for DeBruce’s family, suggesting that as parents they should spare this “kid” who had done something “stupid,” State Court—Trial Transcript, Vol. 7, at 1172, or they could try to manufacture sympathy for DeBruce himself, thereby inviting the State to introduce evidence that this “kid” was in fact a committed criminal. But the defense team could not do both. They made a tactical decision to take one road instead of the other, and the fact that DeBruce now, in hindsight, wishes they made a different call does not mean that the attorneys were constitutionally ineffective. See Chandler, 218 F.3d at 1319 (“Considering the realities of the courtroom, more is not always better. Stacking defenses can hurt a case. Good advocacy requires ‘winnowing out’ some arguments, witnesses, evidence, and so on, to stress others.”). It was not unreasonable for the Court of Criminal Appeals to find that the attorneys’ “strategy in the penalty phase was to beg for mercy,” DeBruce, 890 So.2d at 1093, and Strickland and AEDPA insulate that strategic decision from second-guessing.

B.

Rather than “affirmatively entertaining] the range of possible reasons [petitioner’s] counsel may have had for proceeding as they did,” Pinholster, — U.S. at -, 131 S.Ct. at 1407, the majority “treat[s] the unreasonableness question as a test of [their] confidence in the result [they] would reach under de novo review.” Contra Harrington, 562 U.S. at -, 131 S.Ct. at 786. The majority opinion advances two theories of ineffectiveness: first, that the attorneys must have failed to conduct an adequate investigation because they began their representation shortly before trial, and second, that the attorneys must have failed to pursue the investigative leads in the Garner Report relating to DeBruce’s background and mental health.

Neither theory has definitive support in the record. The majority nonetheless presents its speculation as certainty, and then goes further and says that it would be objectively unreasonable to disagree. “Because the Court of Appeals had little doubt that [the petitioner’s] Strickland claim had merit, the Court of Appeals concluded the state court must have been unreasonable in rejecting it.” Id. This approach reflects unbounded review, heedless of the limitations of Strickland and § 2254(d). “It bears repeating that even a strong case for relief does not mean the state court’s contrary conclusion was unreasonable.” Harrington, 562 U.S. at -, 131 S.Ct. at 786.

The majority is forced to proceed by supposition because the facts in evidence do not get DeBruce very far. After the Rule 32 hearing, the only specific claims are that the attorneys did not hire an investigator and did not retrieve DeBruce’s school or medical records. The majority does not say—and indeed cannot say — that this is prima facie evidence of ineffectiveness, because doing so would treat those omissions as mechanical, per se requirements for attorneys. Strickland, 466 U.S. at 689, 104 S.Ct. at 2065; see also Chandler, 218 F.3d at 1313 (“To state the obvious: the trial lawyers, in every case, could have done something more or something different. So, omissions are inevitable. But, the issue is not what is possible or ‘what is prudent or appropriate, but only what is constitutionally compelled.’ ” (quoting Burger, 483 U.S. at 794, 107 S.Ct. at 3126)). And DeBruce has no evidence to rebut the possibility that the attorneys engaged in other, undocumented investigation activities, see Titlow, — U.S. at -, 134 S.Ct. at 17, because he failed to question Mathis “about what investigations that he and DelGrosso conducted.” DeBruce, 890 So.2d at 1085. That is why the majority has to argue ineffectiveness by supposition.

The majority’s first theory of ineffectiveness makes an inference about time. Advancing a sort of res ipsa loquitur approach to Strickland, the majority claims that “counsel’s testimony during the state collateral proceedings made clear that his failure to conduct adequate investigation was not the result of a strategic decision, but rather a result of lack of time as he scrambled to prepare for DeBruee’s capital trial with less than four weeks’ time while also preparing for and conducting other trials.” Maj. Op. at 1275.

By now, we know that Mathis did not actually say that. Mathis said only that he did not hire an outside investigator or gather academic or medical records. State Court — Collateral Transcript, Vol. 15, 136— 37. Mathis did not say — because he was not asked — whether he (or DelGrosso) took other steps “to investigate DeBruce’s mental health,” nor was he asked about his overall strategy with respect to the penalty phase of his client’s trial. What the majority presents as fact here is just conjecture: because the attorneys were retained so close to trial, the majority infers that they could not possibly have had time to conduct an adequate investigation or make an informed decision.

Of course the majority cites no Supreme Court case to support this temporal rule of attorney performance, because there is none. Under Strickland, our analysis would be the same whether the attorneys were retained one month before trial or one hour before trial. We would still need to make a reasonableness determination, which would still require us to know more about what the attorneys did before we could call them ineffective. The majority cannot hurdle the gaps in the evidence by drawing an inference from the amount of time the lawyers spent preparing for the case; the Strickland presumption points in the other direction, and to rebut it the majority needs more than guesswork and innuendo.

The majority’s second theory of ineffectiveness — that the attorneys ignored the Garner Report — requires more attention. The majority believes that the Garner Report is their smoking gun, the red flag that would induce even a minimally competent attorney to probe for more information. We are told that the attorneys ignored the Garner Report by “failing to take even the first step of asking DeBruce and his mother about the information in the report,” Maj. Op. at 1274, and that even if the attorneys had not ignored it, their investigation would still have been inadequate because “no lawyer could reasonably have made a strategic decision to forego the pursuit of mitigation evidence based on” the Garner Report. Id. at 1274. Either way, the majority reasons, the lawyers must have performed ineffectively, and so surely the Court of Criminal Appeals was unreasonable to find otherwise.

For simplicity, I will divide the majority’s theory into its factual and legal components. The factual claim — that the lawyers outright ignored the Garner Report— suffers from the now-familiar burden of proof problem: DeBruce presents no evidence one way or the other about how the attorneys used (or did not use) the Garner Report, and without that evidence, Strickland requires us to assume — as the State court did- — -that the attorneys used the report properly. The legal claim — that a pre-trial competency evaluation could never be an adequate basis on which to make mitigation decisions — is not supported by any particular Supreme Court holding but merely by the majority’s “own sense of ‘prudence.’ ” Pinholster, — U.S. at -, 131 S.Ct. at 1407 (citation omitted). That approach ignores that “it is not ‘an unreasonable application of clearly established Federal law’ for a state court to decline to apply a specific legal rule that has not been squarely established by [the Supreme] Court.” Knowles, 556 U.S. at 121, 129 S.Ct. at 1419 (quotation marks omitted). That the majority would evaluate this case differently under Strickland does not mean that the Court of Criminal Appeals applied Strickland unreasonably.

Turning first to the suggestion that the attorneys ignored the Garner Report, the only testimony remotely on point was elicited by the attorney for the State, who asked Mathis “whether [he] saw any reason to pursue a psychological defense in this case.” Mathis replied that he had not, because “the [Garner Report] indicated that [DeBruce] was all right and we should go forward, and so we went forward from that point.” State Court — Collateral Transcript, Vol. 15, at 174. No one stopped to clarify whether “psychological defense” referred to the guilt phase, the penalty phase, or both, or whether by “indicated” Mathis meant he had personally read the report or learned of its contents some other way.

This is the only testimony available to us, and the best we can say is that it is ambiguous. The majority reads Mathis to be saying that when DeBruce was found competent to stand trial, his lawyers simultaneously abandoned any hope of a mental health mitigation defense. See Maj. Op. at 1273-74. If you assume, as the majority does, that Mathis and DelGrosso were buffoons, then you might agree here that Mathis betrayed a “failure to understand the law” of mitigation. Id.

But if instead you start with the “strong presumption that [the attorneys acted] for tactical reasons rather than through sheer neglect,” Yarborough, 540 U.S. at 8, 124 S.Ct. at 5 (citing Strickland, 466 U.S. at 690, 104 S.Ct. at 2052), you would not assume that Mathis- — -an experienced and successful capital-defense attorney — was somehow unaware of the elementary legal distinction between guilt-phase competence and penalty-phase mitigation. That assumption is obviously false: this is one of the few things on which we actually have contemporaneous record evidence, in the form of Mathis telling the trial judge that the Garner Report “would be for the purposes of the punishment phase of the trial ... if it comes down to it.” State Court — Trial Transcript, Vol. 1, at 12-13. Mathis plainly understood what the penalty phase of a trial looks like, and he was thinking early on about how to prepare. He said outright he was requesting the Garner Report as part of his penalty-phase investigation, “as opposed to me thinking that there is any way on God’s earth that [DeBruce] is going to be determined by some expert to be insane.” Id.

The majority’s interpretation of Mathis’s ambiguous testimony at the collateral hearing is thus both inconsistent with Strickland’s presumption of competence and contradicted by the trial record. It is just as plausible, if not more so, to interpret Mathis as saying that he saw no reason to pursue a “psychological defense” at the guilt phase, as the Garner Report “indicated that [DeBruce] was all right and we should go forward” to trial. See State Court — Collateral Transcript, Vol. 15, at 174. DeBruce had the chance to clear this up at the Rule 32 hearing — Mathis was, after all, his witness — and he did not.

But, true to form, the majority is ready with the inverted burden of proof: “[T]here is no indication that Mathis’s failure to develop the troubling leads in DeBruce’s competency report was supported by a professional judgment not to pursue a mitigation investigation.” Maj. Op. at 1275. We need to take a step back here, because that is not how Strickland works: the burden is on DeBruce to demonstrate what his attorneys did, or did not do, with the Garner Report. The majority instead takes absence of evidence as evidence of absence, and assumes that the lack of testimony ought to count in DeBruce’s favor. This ambiguity is then breezily converted into a conviction that Mathis was “unaware of DeBruee’s upbringing.” Id. at 1276; contra Titlow, — U.S. at -, 134 S.Ct. at 17-18. Never mind that the Court of Criminal Appeals found as a factual matter that Mathis had reviewed the Garner Report, DeBruce, 890 So.2d at 1092, meaning the burden is on DeBruce under § 2254(d)(2). DeBruce had no evidence to challenge the court’s finding because he never asked Mathis about the Garner Report at the Rule 32 hearing, despite every opportunity to do so. We are required to assume, both by § 2254 and by Strickland, that the attorneys reviewed and thought about the report. Without unambiguous evidence to the contrary, we have no basis for calling the Court of Criminal Appeal’s finding unreasonable.

Undaunted by the lack of evidence and indifferent to our posture of “doubly deferential” review, Yarborough, 540 U.S. at 6, 124 S.Ct. at 4, the majority goes on to lecture the Court of Criminal Appeals about its alleged misapplication of Strickland: “Because no lawyer could reasonably have made a strategic decision to forego the pursuit of mitigation evidence based on the [Garner] report governing competency to stand trial, the Alabama Court of Criminal Appeals’ conclusion to the contrary constitutes an unreasonable application of Strickland’s performance prong.” Maj. Op. at 1274. The claim here seems to be that because the Garner Report was nominally a competency evaluation rather than a full-bore social history, it was by definition an inadequate basis on which to make strategic decisions regarding mitigation.

Even assuming the majority is right about this as a factual proposition — that is, even if we assume, contra Strickland, that the Garner Report was the only source of information relied on by the attorneys in planning for the penalty phase, see Titlow, — U.S. at -, 134 S.Ct. at 18-the Supreme Court has never endorsed a per se rule that attorneys are unreasonable for curtailing an investigation after reviewing something like the Garner Report, simply because that report is called a competency evaluation. Indeed, if anything, the Court has said the opposite. In Pinholster, the Ninth Circuit relied on Williams v. Taylor, 529 U.S. 362, 120 S.Ct. 1495, 146 L.Ed.2d 389, Wiggins v. Smith, 539 U.S. 510, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003), and Rompilla, and

drew from those cases a “constitutional duty to investigate,” and the principle that “[i]t is prima facie ineffective assistance for counsel to ‘abandon[] their investigation of [the] petitioner’s background after having acquired only rudimentary knowledge of his history from a narrow set of sources.’ ” The [Ninth Circuit] explained that it could not “lightly disregard” a failure to introduce evidence of “excruciating life history” or “nightmarish childhood.”

Pinholster, — U.S. at -, 131 S.Ct. at 1406 (alterations in original, citations omitted).

The Ninth Circuit was reversed. Their reasoning “misapplied Strickland and overlooked ‘the constitutionally protected independence of counsel and ... the wide latitude counsel must have in making tactical decisions.’ ” Id. (quoting Strickland, 466 U.S. at 689, 104 S.Ct. at 2065). “The Court of Appeals erred in attributing strict rules to [the Supreme] Court’s recent case law,” id. at -, 131 S.Ct. at 1407, and failed “to affirmatively entertain the range of possible ‘reasons [petitioner’s] counsel may have had for proceeding as they did,’ ” id. (citations omitted). “There comes a point where a defense attorney will reasonably decide that another strategy is in order, thus ‘mak[ing] particular investigations unnecessary.’ ” Id. (alteration in original, quoting Strickland, 466 U.S. at 691, 104 S.Ct. at 2066)).

The majority opinion does not engage with the Supreme Court’s most recent decisions applying AEDPA to Strickland claims, like Pinkolster and Harrington. Instead, the majority relies principally on Williams, from which they draw the principle that “decisions about whether to present a mental-health defense during the guilt phase are separate issues from the decision whether to investigate mitigation evidence during sentencing.” Maj. Op. at 1273 (emphasis in original). This platitude is both obvious and totally unhelpful; of course there’s an analytical distinction between preparing for the guilt phase and preparing for the penalty phase, but that tells us nothing about what these attorneys did in this case, or whether those actions were reasonable. That the distinction exists, in the abstract, does not also establish that Mathis and DelGrosso failed to appreciate it or act on it. See Roe v. Flores-Ortega, 528 U.S. 470, 478, 120 S.Ct. 1029, 1035, 145 L.Ed.2d 985 (2000) (“The Court of Appeals failed to engage in the circumstance-specific reasonableness inquiry required by Strickland, and that alone mandates vacatur and remand.”). Williams certainly does not contain the categorical language the majority tries to squeeze from it; even the majority’s preferred language from Williams — that mitigating evidence “may alter the jury’s selection of penalty, even if it does not undermine or rebut the prosecution’s death-eligibility case,” 529 U.S. at 398, 120 S.Ct. at 1516 (emphasis added) (quoted at Maj. Op. at 1273) — is phrased conditionally, and does not resemble a mechanical rule that attorneys must conduct a particular kind of investigation before they make tactical decisions about mitigation.⁶

The majority bandies the distinction between guilt-phase competence and penalty-phase mitigation around like it is some kind of profound insight lost on the Court of Criminal Appeals. The majority opinion asserts that the State court “erroneously] conflate[ed] ... the issues of guilt-phase mental health defenses and competence to stand trial with the separate issue of whether to conduct a mitigation investigation.” Maj. Op. at 1273. As a preliminary matter, that statement would probably come as a surprise to the author of the Court of Criminal Appeals opinion, Judge Wise. She wrote, “we hold that Mathis was not ineffective for failing to investigate further into DeBruce’s mental health,” DeBruce, 890 So.2d at 1093, in a section of the opinion that begins “De-Bruce also argues that his counsel was ineffective in the penalty phase for failing to investigate.” Id. at 1088 (emphasis added). Judge Wise went on to discuss other cases involving “a claim that counsel’s performance was deficient for failing to present mitigation evidence of the defendant’s abusive childhood,” id. at 1092, and ended by considering the application of Wiggins “when reviewing a claim of ineffective assistance of counsel at the penalty phase of a capital trial,” id. at 1093 (emphasis added). To put it simply, Judge Wise understands what the penalty phase of a trial looks like, and understands the distinction between competence and mitigation. To put it even more simply: Judge Wise is not a dolt, and the majority’s eagerness to assume otherwise is not consistent with AEDPA. See Titlow, — U.S. at -, 134 S.Ct. at 13 (“[AEDPA] and Strickland ... do not permit federal judges to so casually second-guess the decisions of their state-court colleagues or defense attorneys.”) (citations omitted); see also Woodford v. Visciotti, 537 U.S. 19, 24, 123 S.Ct. 357, 360, 154 L.Ed.2d 279 (2002) (“This readiness to attribute error is inconsistent with the presumption that state courts know and follow the law.”).

The majority’s real disagreement with Judge Wise is not about her “conflating” elementary legal concepts, but over the content of her court’s judgment. The Court of Criminal Appeals believed that DeBruce and his mother actively misled the defense attorneys (and maybe everyone else) about DeBruce’s background, and relied on the long line of cases— including Strickland — holding that a defendant cannot dissuade his attorney from investigating a particular line of defense and then later, in a collateral proceeding, turn around and challenge the attorney’s conduct as unreasonable. See 466 U.S. at 691, 104 S.Ct. at 2066 (“Counsel’s actions are usually based, quite properly, on informed strategic choices made by the defendant and on information supplied by the defendant. In particular, what investigation decisions are reasonable depends critically on such information.”) (emphasis added). Judge Wise noted that DeBruce said one thing to Gary Garner (resulting in the Garner Report, which the majority assumes contains the truth), another thing to the probation officer who prepared the post-trial, presentence report, see DeBruce, 890 So.2d at 1092 (“[DeBruce] reported a good relationship with both of his parents and his numerous siblings. There were no significant problems noted during his formative years.... DeBruce claims to have attended high school at Ramsey, in Birmingham, Alabama[,] and was in the ‘Gifted’ program. The Probation Office report from Birmingham indicated that DeBruce finished high school at Hayes High School. DeBruce then attended the University of Alabama in Birmingham for one semester.”) (quoting the Probation Office’s post-trial presentence report), and said perhaps a third thing to his attorneys. Of course, we can’t be sure about that last one because DeBruce offered no evidence at the Rule 32 hearing about the substance of his conversations with his lawyers. “Given DeBruce’s mother testimony [sic], DeBruce’s own comments, and the conflicting expert testimony about DeBruce’s mental health, [the Court of Criminal Appeals held] that Mathis was not ineffective for failing to investigate further into De-Bruce’s mental health.” Id. at 1093.

The majority does not even pretend to engage with that court’s reasoning. Instead, mimicking the error for which the Sixth Circuit was reversed in Harrington, the majority “treat[s] the unreasonableness question as a test of its confidence in the result it would reach under de novo review.” 562 U.S. at -, 131 S.Ct. at 786. Hence the majority’s conviction that the attorneys “never confronted DeBruce or his mother with these inconsistencies [in DeBruce’s account of his background] in order to discover the truth, nor did [they] follow up on the [Garner Report].” Maj. Op. at 1274. At bottom, what the majority tries to portray as a legal disagreement is really a factual one: the majority reads the record differently, and then conflates this with the reasonability determination required by AEDPA. See Knowles, 556 U.S. at 126, 129 S.Ct. at 1421 (“Here, the Court of Appeals failed even to mention the clearly-erroneous standard, let alone apply it, before effectively overturning the lower court’s factual findings related to counsel’s behavior.”).

All this eventually returns us to the burden of proof, which rests on DeBruce: if he wants to claim that his attorneys never discussed his background with him, or never asked him about the contents of the Garner Report, he needs to actually prove that those conversations never happened. Otherwise, Strickland’s strong presumption of competence requires us to assume, as the Alabama court did, that the defense lawyers reviewed the Garner Report, discussed it with their client, and made a tactical decision to go in a different direction. DeBruce had every opportunity to rebut that presumption during his direct examination of Mathis at the Rule 32 hearing; it would have been simple to ask Mathis about the Garner Report, about DelGrosso’s involvement, about the attorneys’ interactions with their client and his family, and so on. But those questions were not asked.

Where does that leave us? The answer should be easy: DeBruce’s petition turns on the substance of conversations between DeBruce and his attorneys that we know nothing about, and this “absence of evidence cannot overcome” Strickland’s presumption of effectiveness, Titlow, — U.S. at -, 134 S.Ct. at 17, let alone overcome the deference we owe to the State court under AEDPA. But instead of applying AEDPA, we rewrite the record to include “facts” to which nobody testified, contra id.; we announce new rules of attorney performance, shakily derived from the dicta of cases decided years after DeBruce’s trial, contra Pinholster, — U.S. at -, 131 S.Ct. at 1406-07; and we ignore the Court of Criminal Appeals’ decision almost entirely, unless it is to insinuate that our State-court colleagues do not know what the penalty phase of a trial looks like, contra Harrington, 562 U.S. at -, 131 S.Ct. at 786. I would have thought that seeing our sister circuits get reversed for these errors would give us some pause before rolling them all into one opinion.

IV.

It is hard to say what principle guides the majority’s prejudice analysis. As far as I can tell, today’s opinion holds that if a petitioner can produce mitigating evidence in a collateral proceeding that was not presented at trial, then federal judges should find prejudice if they are convinced — in their independent judgment— that the new evidence is sufficiently “powerful” or “compelling.” Maj. Op. at 1272. Of course it is not so simple.

The Alabama court was reasonable in finding no prejudice. Though the majority does its best to wring every last drop of pathos from the testimony, the new mitigation case is merely (1) that DeBruce was diagnosed with a painful intestinal disorder two years after his conviction, (2) that DeBruce is a person of low-to-average intelligence, and (3) that he grew up poor in a bad neighborhood. Those circumstances are, indeed, unfortunate. On the other hand, many individuals in equally unfortunate circumstances do not go on to murder AutoZone customers. The majority never spells out exactly how DeBruce’s “powerful” and “compelling” mitigation evidence would have helped him at trial. Even DeBruce does not argue that his intelligence or his childhood explain, or lessen his culpability for, the murder of Douglas Battle. They are, I suppose, just reasons to feel sorry for him. But that is not enough.

The Alabama court conducted the prejudice analysis required by Strickland, “reweigh[ing] the evidence in aggravation against the totality of available mitigating evidence,” Wiggins, 539 U.S. at 534, 123 S.Ct. at 2542, and concluded that there was no reasonable probability of a different outcome. Repeating its conclusion from the direct appeal of DeBruce’s conviction, the court wrote:

Although we have reviewed ‘worse’ crimes of robbery-murder, the apparent cold-blooded, senseless, and unmerciful manner in which the defendant executed this crime impresses this Court as it did the trial judge. Other than the general arguments for not imposing a punishment of death in any ease, we find no reason to spare the defendant’s life.

DeBruce, 890 So.2d at 1091 (quoting DeBruce v. State, 651 So.2d 599, 623 (Ala.Crim.App.1993)).

Just as the majority ignored AEDPA in evaluating Strickland’s performance element, the majority proceeds to examine the prejudice element as though § 2254 does not exist. The majority sets out its own prejudice analysis — relying principally on Eleventh Circuit cases instead of the Supreme Court cases relevant to § 2254— and then concludes that the Court of Criminal Appeals was “unreasonable” because they reached a conclusion different from the majority’s. But the Supreme Court has “said time and again that ‘an unreasonable application of federal law is different from an incorrect application of federal law.’ Even if the Court of Appeals might have reached a different conclusion as an initial matter, it was not an unreasonable application of our precedent for the [State court] to conclude that [the petitioner] did not establish prejudice.” See Pinholster, — U.S. at -, 131 S.Ct. at 1411 (citations omitted, quoting Harrington, 562 U.S. at -, 131 S.Ct. at 785). The majority “d[oes] not observe this distinction, but ultimately substitute^] their own judgment for that of the state court, in contravention of § 2254(d).” Woodford, 537 U.S. at 25, 123 S.Ct. at 360.

In particular, the majority relies on various Eleventh Circuit cases for the principle that “the primary purpose of the penalty phase is to ensure that the sentence is individualized by focusing on the particularized characteristics of the defendant,” Brownlee v. Haley, 306 F.3d 1043, 1074 (11th Cir.2002), and says that a petitioner is necessarily prejudiced when omitted mitigation evidence “paints a vastly different picture of [the petitioner’s] background than that created by” the evidence that was presented, Williams v. Allen, 542 F.3d 1326, 1342 (11th Cir.2008). For what it is worth, I think these sweeping statements — which reflect the humanization-above-all-else ethos rejected in Pinholster, see — U.S. at -, 131 S.Ct. at 1410—are inconsistent with Strickland’s contextual, case-by-case standard; surely these statements cannot be taken literally, because a clever lawyer could “paint a vastly different picture” in a way that benefits his client by excluding even worse background information from the jury’s consideration. See Wong, 558 U.S. at 25, 130 S.Ct. at 389. But that is not worth much, because Brownlee and Williams are not worth anything for AEDPA purposes: § 2254(d)(1) is concerned with “clearly established Federal law, as determined by the Supreme Court,” not by panel decisions of the Eleventh Circuit. See Parker v. Matthews, — U.S. -, 132 S.Ct. 2148, 2155, 183 L.Ed.2d 32 (2012) (“The Sixth Circuit also erred by consulting its own precedents, rather than those of the [Supreme] Court, in assessing the reasonableness of the [State court]’s decision.”) (emphasis added). That the majority relies principally on our cases, rather than those of the Supreme Court, underscores how little basis the majority has for not deferring to the State court’s decision, as AEDPA requires.

A.

Regarding evidence of DeBruce’s gastrointestinal problems, the District Court concluded, with some understatement, that the claim “does not require extensive analysis.” DeBruce, No. 04-2669, ECF Doc. 31, at 61. It is hard to imagine that the jury, having just convicted DeBruce of a cold-blooded murder, would be inclined to give him a pass because he suffers bad stomach aches. DeBruce’s claim here is not that his diagnosis has any bearing on the murder for which he was convicted; it’s just a reason to feel sorry for him. The Court of Criminal Appeals was not unreasonable in concluding that this diagnosis would not have changed DeBruce’s sentence.

B.

DeBruce’s second category of mitigating evidence is that he is not particularly bright. The Court of Criminal Appeals described evidence on this point as “conflicting,” DeBruce, 890 So.2d at 1085, because at the Rule 32 hearing the State’s expert disagreed with DeBruce’s expert about whether DeBruce has brain damage. On the separate question of DeBruce’s IQ, the experts were more closely aligned: the State’s expert “had evaluated DeBruce and had conducted several IQ tests and determined that DeBruce’s verbal IQ was 84,” id. at 1084, but agreed on cross-examination that “scores in the 70s are consistent, and it would be consistent also with my observations of him,” State Court — Collateral Transcript, Vol. 15, at 240. At the Rule 32 hearing, then, there were two issues in play: whether DeBruce had brain damage, and whether DeBruce had a low-to-average IQ. The Court of Criminal Appeals believed evidence on the first issue was inconclusive — -the State’s expert testified that DeBruce was malingering — and believed that the second issue, standing alone, was not mitigating enough to warrant a different sentence.

The majority says that the court was unreasonable because the inconsistency in the expert testimony “favors DeBruce.” Maj. Op. at 1277. I cannot imagine what the majority means by that, because I cannot imagine how testimony that De-Bruce “suffers from no serious mental illness or mental defect” and was “malingering psychological psychiatric symptoms” could possibly “favor” him. The word I would use is “rebut,” as in: “To rebut DeBruce’s evidence, the State offered the testimony of its own expert at the postcon-viction hearing.” DeBruce, 890 So.2d at 1084. That expert “said that it was his opinion that DeBruce was not mentally retarded but was malingering.” Id.⁷ If that kind of testimony “favors” the petitioner, what would damaging testimony look like?

The majority simply disagrees with the Court of Criminal Appeals about the weight of DeBruce’s IQ scores, and conflates that disagreement with AEDPA’s reasonableness determination. The majority accuses the State court of “discounting” evidence of DeBruce’s “mental impairments,” Maj. Op. at 1276, but the court did no such thing:

As directed by the Supreme Court in Wiggins, we have reweighed the omitted mitigating evidence — evidence indicating that DeBruce had Crohn’s disease and conflicting evidence of DeBruce’s mental health — against the aggravating circumstances that were proven in this case— the fact that the murder was committed during a robbery and the fact that De-Bruce had previously been convicted of a crime of violence. Given the aggravating circumstances presented here and the cold blooded manner in which the killing occurred, we believe that De-Bruce’s death sentence was appropriate.

DeBruce, 890 So.2d at 1093 (emphasis added). The court did not “discount” De-Bruce’s mental health argument; they considered it and, in light of testimony that DeBruce “suffers from no serious mental illness or mental defect” and was “malingering psychological and psychiatric symptoms,” the court accurately characterized the evidence as “conflicting.” See State Court—Collateral Transcript, Vol. 15 at 231, 237. The majority, obviously, would read the transcript more favorably for De-Bruce, but that is not the question under AEDPA. Section 2254 asks whether it was unreasonable for the State court to find as it did. See Pinholster, — U.S. at -, 131 S.Ct. at 1411. The Court of Criminal Appeals was not unreasonable for believing that DeBruce would not benefit from a battle of the experts over his mental health.

C.

The final category of mitigating evidence concerns DeBruce’s background and upbringing, which the majority says was “heavily disadvantaged.” Maj. Op. at 1276. Most of this evidence is drawn from the testimony of two of DeBruce’s sisters, Linda and Jacqueline DeBruce, who were both in jail at the time of the Rule 32 hearing. The sisters’ testimony suggests that the DeBruce children grew up poor in a bad neighborhood. Their father “just went to work, came home. That’s about it.” State Court — Collateral Transcript, Vol. 17, at 347. Their mother “worked the majority of the time.” Id. And their neighborhood “was like normal housing projects, violence, fighting, stabbing, stuff like that.” Id. at 336. Jacqueline testified to “occasions where a lot of guys jumped on Derrick, for what reason I don’t know,” id. at 338, but clarified on cross-examination that “we didn’t see it, but we know that he got beaten up because my mother had made a police report on the guys who jumped on Derrick,” id. at 354.

Some of the sisters’ testimony is not mitigating at all, and some of it is of only debatable usefulness to DeBruce. When asked whether the sibling she grew up with had any mental problems, the best Jacqueline DeBruce could come up with was an anecdote about DeBruce forgetting to buy bread at the grocery store. Id. at 343. Likewise, she recalled that the eldest DeBruce child, Juanita, would discipline DeBruce by withholding his favorite foods, “because he really liked to eat what he wanted to eat.” Id. at 347.

The majority makes much of “the daily beatings that DeBruce suffered as a child at the hands of his older sister [Juanita].” Maj. Op. at 1276. These “daily beatings” sound more like spankings in Jacqueline’s telling — “[Juanita] used to whoop him all the time____[S]he used to whoop him with extension cords, switches and twist them together.” State Court — Collateral Transcript, Vol. 15, at 347 — and it seems odd that this supposed sadist would also be the sibling most invested in helping DeBruce with his homework: “[Juanita] used to try to help him, or we used to have to help him also, but she used to like be a little hard on him because he told her that he didn’t know how to do it, and she wanted to instill in his head that he know how to do it. She used to make him do it.” Id. at 348. Indeed, at several points Jacqueline testified that the DeBruce children were close-knit and looked out for one another: “I used to have to dress him and tie his shoes. I used to have to go to the classroom door and get him all the time when the rest of us used to meet up together somewhere.” Id. at 342.

The majority says that all this “is the kind of troubled history that the [Supreme] Court has ‘declared relevant to assessing a defendant’s moral culpability,’” Maj. Op. at 1276 (quoting Wiggins, 539 U.S. at 535, 123 S.Ct. at 2542), but the majority is careful not to go into much detail about the troubled histories actually described in cases like Porter, Wiggins, and Williams, because — as the District Court recognized — “DeBruce’s childhood ... pales in comparison to the nightmarish childhoods experienced by the petitioners in the line of Supreme Court eases he relies upon to support a prejudice argument.” DeBruce, No. 04-2669, ECF Doc. 31, at 66. DeBruce did not “routinely witness[ ] his father beat his mother, one time so severely that she had to go to the hospital and lost a child.” Porter, 558 U.S. at 33, 130 S.Ct. at 449. His father never tried to shoot him. Id. His parents did not “frequently [leave him] and his siblings home alone for days, forcing them to beg for food and to eat paint chips and garbage.” Wiggins, 539 U.S. at 516-17, 123 S.Ct. at 2533. His mother did not have “sex with men while her children slept in the same bed” or “force[ ] [his] hand against a hot stove burner.” Id. at 517, 123 S.Ct. at 2533. DeBruce was not “repeatedly molested and raped” by a foster parent. Id. He was not “gang-raped ... on more than one occasion” or “sexually abused by his supervisor” at work. Id. His parents were not “imprisoned for the criminal neglect” of their children, Williams, 529 U.S. at 395, 120 S.Ct. at 1514, nor did DeBruce grow up in a house with “several places on the floor where someone had had a bowel movement,” where “[u]rine was standing in several places in the bedrooms,” where “it was impossible to step any place on the kitchen floor where there was no trash,” or where “[t]he children had to be put in [the hospital], as four of them ... were definitely under the influence of whiskey.” Id. at n. 19. “DeBruce’s childhood simply does not carry with it the same horrors as these cases. Nor are there any hospital, social services or school records that corroborate [Jaqueline] DeBruce’s contention that De-Bruce was physically and emotionally abused, malnourished or suffered from black outs or seizures.” DeBruce, No. 04-2669, ECF Doc. 31, at 67. In sum, DeBruce’s case looks nothing like the cases on which he and the majority rely. Compared side-by-side with those cases, his prejudice argument is weak.

The majority also ignores the possibility that extended testimony about DeBruce’s background would have been more hurtful than helpful. See Wong, 558 U.S. at 25, 130 S.Ct. at 389 (“A heavy-handed case to portray [the petitioner] in a positive light ... would have invited the strongest possible evidence in rebuttal.”). The humanizing strategy the majority now advocates “would have put into play aspects of [petitioner’s] character that would have triggered admission of” rebuttal evidence. Id. at 22, 130 S.Ct. at 387-88. For instance, claims about DeBimce’s “resistance to joining gangs,” Maj. Op. at 1276, would obviously open the door to testimony about DeBruce’s participation in numerous robberies as part of a criminal gang, including the attempted armed robbery of the Syla-cauga bank that Mathis had lobbied so vigorously to exclude in the guilt phase of the trial.

Indeed, following DeBruce’s mother’s testimony at the penalty phase of trial, the prosecutor expressed frustration at not being allowed to rebut her testimony: “I’m sitting back with witnesses for rebuttal.” State Court — Trial Transcript, Vol. 7, at 1144. The prosecutor also wanted to call Gary Garner: “The issue [of mental health] is interjected, Judge. You’ve got the defendant saying [in the Garner Report] that he denies any prior mental health treatment or psychiatric hospitalization, and that’s in direct conflict to the testimony [of DeBruce’s mother] they put on.” Id. at 1151.

The defense lawyers objected. DelGrosso argued that the prosecutor “had every opportunity in the world during the course of this trial to introduce this defendant’s statement, and he didn’t do it. Now he’s trying to put on evidence that he didn’t put on in his case in chief and it is not in rebuttal to anything that [DeBruce’s mother] said.” Id. at 1148. “I don’t think we can go into any of it. [The prosecutor] is talking about putting [Garner] on to testify about statements that [DeBruce] made to impeach [Jessie DeBruce] ... [the prosecutor] can’t impeach her by saying what [DeBruce] said as far as hearsay goes.” Id. at 1150. There is some irony in the fact that DeBruce’s lawyers worked so hard to exclude the testimony that DeBruce now, in hindsight, calls mitigating.

It also is not obvious that a greater focus on the DeBruce household would have been helpful. For one thing, “some of the new testimony would likely have undercut the mitigating value of the testimony by [petitioner’s] mother.” Pinholster, — U.S. at -, 131 S.Ct. at 1410. Jacqueline DeBruce testified that De-Bruce’s mother was aware of the supposedly heinous abuse Juanita was inflicting on the children: “My mother said she left [Juanita] in control over us and she can chastise us as she choose.” State Court— Collateral Transcript, V ol. 15, at 355. Additionally, a focus on the DeBruce family could draw lingering attention to the other DeBruce siblings’ legal problems; Mathis tried to blunt this in his direct examination of DeBruce’s mother by acknowledging that most of the DeBruce sons either were, or had been, in prison. See State Court— Trial Transcript, Vol. 7, at 1127-29. It is not clear how a sustained focus on the DeBruce family would have helped DeBruce. See Pinholster, — U.S. at -, 131 S.Ct. at 1410 (“The new evidence relating to [petitioner’s] family ... is also by no means clearly mitigating, as the jury might have concluded that [petitioner] was simply beyond rehabilitation.”).

In any case, the Alabama Court of Criminal Appeals considered all of the evidence — the old and the new, the good and the bad — and determined that none of the new mitigating evidence was weighty enough to show a “substantial” likelihood of a different sentence. See Harrington, 562 U.S. at -, 131 S.Ct. at 792. They were right. Stripped of rhetorical embellishment, DeBruce’s new evidence is simply not the harrowing stuff of cases like Porter, Wiggins, or Williams. Some of it might have even backfired and opened the door to additional damaging testimony. The majority is wrong to find prejudice here, but it is especially wrong to set aside the State court’s contrary judgment simply because the majority feels differently.

V.

While I join the majority in concluding that DeBruce failed to prove that his lawyers were ineffective in the guilt phase of his trial, I would reject his claim that his lawyers were ineffective in the penalty phase. DeBruce had the responsibility— and opportunity — to prove that his lawyers were genuinely unaware of his background at trial. DeBruce responded with speculation and bluster. Because that is not enough to warrant habeas relief, I respectfully dissent.

1 . I agree with the majority’s rejection of De-Bruce’s guilt-phase ineffectiveness claim.

2 . Mathis testified that he charged DeBruce’s family $10, 000 for the representation, and "they paid about $1, 500.” State Court— Collateral Transcript, Vol. 15, at 61. Curiously, other members of the DeBruce family continued to use Mathis's legal services even after Derrick DeBruce's conviction. The State court found it “interesting that Mathis testified at the postconviction evidentiary hearing that since DeBruce's trial he has represented 'several' of DeBruce’s 10 siblings.” DeBruce, 890 So.2d at 1081 n. 3 (quoting State Court— Collateral Transcript, Vol. 15, at 173).

3 . Gary Garner did not testify at trial or at the Rule 32 hearing. At the end of his report, he identified himself as "Gary L. Garner, M. S.,” and listed his title as "Co-ordinator Talladega Mental Health Center.”

4 . See also Sparkman v. Comm’r, 509 F.3d 1149, 1155 n. 5 (9th Cir.2007) ("Where the burden of production rests on a party, a court may, at its discretion, presume or infer from that party’s failure to call a witness that the testimony the witness would have offered would not favor that party.”); United States v. Pagan-Santini, 451 F.3d 258, 267 (1st Cir.2006) ("[A missing witness] instruction may be given where a party controls or has peculiar access to a witness and, in the circumstances, it may be reasonable to suppose that the party would produce the witness unless the testimony was unfavorable.”); Streber v. Comm’r, 138 F.3d 216, 221-22 (5th Cir.1998) ("In general, a court may draw a negative inference from a party's failure to produce a witness 'whose testimony would elucidate the transaction.'" (quoting Graves v. United States, 150 U.S. 118, 121, 14 S.Ct. 40, 41, 37 L.Ed. 1021 (1893))); Gumbs v. Int'l Harvester, Inc., 718 F.2d 88, 96 (3d Cir.1983) ("The unexplained failure or refusal of a party to a judicial proceedings to produce evidence that would tend to throw light on the issues authorizes, under certain circumstances, an inference or presumption unfavorable to such party.”); United States v. Tucker, 552 F.2d 202, 210 (7th Cir.1977) (“Like all sound inferences, the missing witness inference is rooted in notions of common sense, specifically that where a party fails to call an available witness with important and relevant knowledge, it may be that he has something to fear in the witness' testimony.”); see generally 2 McCormick on Evidence § 264 (7th ed.) ("When it would be natural under the circumstances for a party to call a particular witness, or to take the stand as a witness in a civil case, or to produce documents or other objects in his or her possession as evidence^] and the party fails to do so, tradition has allowed the adversary to use this failure as the basis for invoking an adverse inference.”) (footnotes omitted).

5 . The majority thus misses the point when it describes Jessie DeBruce’s testimony as "grossly inaccurate.” Maj. Op. at 1274. That criticism only makes sense if you believe that humanization is the only appropriate defense strategy; it ignores the possibility that the lawyers intended to arouse sympathy not for DeBruce but for DeBruce's family. If the lawyers could present DeBruce as the wayward son of a hardworking, close-knit, blue collar family, the jurors might take mercy on the family and spare them the pain of their son’s execution. This is "the sort of calculated risk that lies at the heart of an advocate’s discretion.” Yarborough, 540 U.S. at 10, 124 S.Ct. at 6.

6 . I also wonder why the majority is talking about 'Williams at all in the part of their opinion about attorney performance. Williams was decided on the prejudice prong of Strickland: the Virginia Supreme Court was held to have misinterpreted the application of Lockhart v. Fretwell, 506 U.S. 364, 113 S.Ct. 838, 122 L.Ed.2d 180 (1993), to Strickland’s prejudice standard. See Williams, 529 U.S. at 391, 120 S.Ct. at 1512. There was no actual dispute in Williams about performance — the state judge who sentenced Williams had found counsel’s performance deficient, id. at 370, 120 S.Ct. at 1501, the Virginia Supreme Court "assumed, without deciding, that trial counsel had been ineffective,” id. at 371, 120 S.Ct. at 1501, and the State "barely disputed” that point in its brief, id. at 395, 120 S.Ct. at 1514—and so the Court’s disquisition about performance was dicta: it says nothing about a case like DeBruce’s, where we must apply AEDPA when a state court finds an attorney competent and the petitioner disagrees. Coincidentally, Williams was also the case establishing that AEDPA is concerned with “the holdings, as opposed to the dicta, of [the Supreme] Court's decisions as of the time of the relevant state-court decision.” Id. at 412, 120 S.Ct. at 1523 (emphasis added). And so, by relying on dicta from a case, decided on a different posture of review, that also post-dates DeBruce’s trial by about a decade, the majority pulls off a kind of dubious hat trick, getting Williams wrong in just about every relevant way.

7 . The majority stresses that the State expert testified that DeBruce "was giving reasonable effort” during his evaluation. Maj. Op. at 1277 (quoting State Court — Collateral Transcript, Vol. 16, at 236). And that is true. What the majority leaves out is that Dr. King was responding to a question about whether he felt "that [DeBruce] was putting forth maximum effort.” State Court — Collateral Transcript, Vol. 16, at 236 (emphasis added). "Reasonable effort” looks less impressive in context, so the majority just omits the context.