With all respect for my colleagues in the majority, I cannot concur in their conclusion that a Brady violation occurred in this case. To be sure, the facts portrayed by the majority regarding the conviction of Brady George Spicer for the brutal beating and attempted murder of Francis Den-vir suggest troubling questions that strike at the heart of our criminal justice system. Our system, however, has been and remains the best ever devised for ferreting out and punishing the guilty, while vindicating the innocent.
*563In its pursuit of the guilty, the government often enlists the assistance of brigands and blackguards to ensure that the worst among them receive their just desserts. Time and again we have acknowledged the risk that inheres in harboring our hopes for a true and honorable adjudication in those who are, by nature and inclination, generally untrustworthy and dishonorable.
But our criminal justice system has always willingly accepted that risk, confident that the greater good — 'the public interest — is thereby better served. Now, faced with an unpleasant consequence of the choice that society has made, the majority tailors from whole cloth a remedy for the injustice it perceives, i.e., that Brady Spi-cer may have been wrongly convicted. From the tapestry of that remedy, however, hangs a loose thread — an unprecedented expansion of Brady and the unwarranted issuance of the Great Writ.
I.
At the heart of every Brady dispute lies the conduct of the prosecuting attorney. The majority bases its grant of relief in this case on the supposed neglect of Steven Sindler — the Assistant State’s Attorney for Anne Arundel County who prosecuted Spicer — to inform Spicer’s lawyer, James Saltón, that witness Larry Brown had made “inconsistent” statements concerning his contact with Spicer in the days surrounding the attack on Denvir at Armadillo’s. It should be instructive, then, to examine the relevant facts from Sindler’s point of view.
A.
Sometime during the summer of 1990, Sindler was contacted by Gary Christopher, an attorney in the Anne Arundel County office of the Maryland Public Defender. Christopher was interested in working out a deal for Brown, his client, who had recently been indicted on drug charges. Sindler was told that Brown “could give him the perpetrator” of the crimes committed against Denvir about six months previously. J.A. 601 (state post-conviction testimony of Gary Christopher).
Within a week or two, after Sindler had consulted with his boss to confirm the state’s interest in a potential deal with Brown, he met again with Christopher, who made the following proffer:
He said that Larry Brown had been working downtown in the downtown area in the market house shucking oysters, and that Mr. Spicer had been hanging out down there, running errands and I believe he said sleeping on the street and everything.... [Brown] was friendly with Brady Spicer, or acquainted with him through this running of errands, and ... Mr. Spicer had asked him things about Armadillos.... Then the incident occurred and that sometime after that incident there was conversation between the two, something to the effect that gee, thanks for being — that Mr. Spicer said to Larry Brown, well, thanks for being cool or something like that, something in — that general type of conversation.
J.A. 847-48 (state post-conviction testimony of Steven Sindler).
Sindler “didn’t really pay much attention” to the details of Christopher’s proffer,1 because he wanted to talk with Brown himself before finalizing any plea bargain: “[B]efore I would act upon anything that a snitch would tell me, ... I would speak to that person and know firsthand. Generally ... an attorney will say [only] that their client has information.” J.A. 851.2
*564B.
About a month later, Brown arrived in Sindler’s office to discuss the incident at Armadillo’s directly with Sindler. Significantly, Christopher was absent from this meeting. Although he had just accompanied Brown to a plea proceeding in court, Christopher left his client alone with the prosecutors after, in his words, “deliver[ing] him to their hands.” J.A. 604.3
Brown told his story to Sindler, filling in the details of Christopher’s sketchy proffer. With regard to the attack on Denvir, which occurred as Brown was shucking oysters across the street from Armadillo’s, Brown said that he had seen Spicer running from the restaurant. Sindler “didn’t think anything” of Christopher’s earlier failure to mention that Brown had actually been an eyewitness, because Christopher’s proffer had only been a “description of what his client might say.” J.A. 851, 867 (emphasis added). As a result, Sindler “didn’t think that [the statements] were contradictory.” J.A. 867.4
Indeed, consistency — not contradiction — was the hallmark of Brown’s account of Spicer’s connection to the Denvir beating. With Sindler present, Brown repeated the same story under oath to the grand jury, and again at Spicer’s trial. On the strength of all the evidence, including the testimony of Henry Connick and Sam Novella, the jury found Spicer guilty.
II.
The majority notes the applicable standard of review, ante at 554, but it bears repeating: a federal court is prohibited from granting habeas relief on any claim that was adjudicated at the state level unless the state court either (1) unreason*565ably determined the relevant facts or (2) unreasonably applied the law to the facts. See 28 U.S.C. § 2254(d). We must evaluate what is “reasonable” in light of all the evidence (with regard to the facts), and within the context of clear Supreme Court mandates (with regard to the law). See id.
The prosecution’s failure to disclose specific evidence to the accused does not violate Brady unless the evidence is both favorable and material. Strickler v. Greene, — U.S. -, 119 S.Ct. 1936, 1948-49, 144 L.Ed.2d 286 (1999). Christopher’s proffer to Sindler fails to satisfy either criterion.
A.
Though evidence impeaching a prosecution witness is doubtlessly “favorable,” United States v. Bagley, 473 U.S. 667, 676, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985), impeachment with the witness’s prior inconsistent statement presupposes two things. First, the earlier statement must indeed be that of the witness. Second, the statement must actually be inconsistent.
The prior statement that Sindler failed to disclose was not that of his witness, Brown, but that of Brown’s lawyer. Nothing that Brown testified to before the grand jury or at Spicer’s trial was inconsistent with anything that he had ever said in Sindler’s presence; the purported discrepancy is instead between what Brown told Sindler directly, and what Christopher represented that Brown had said earlier.
The lack of identity between the speakers is crucial to this appeal, particularly considering that Christopher’s statement was given in the context of a proffer to Sindler. Proffers by counsel attempting to negotiate a plea, immunity, or other benefit for their clients are, by their very nature, unreliable for ascertaining specific facts. The following is an apt description of the proffer process:
Through incremental steps, we get information about what the witness can tell us, and we are willing to say at each step what our reaction is to the information we have received.... Counsel will come in for a witness and advise us that he has a witness who is prepared .to cooperate.... At that point, we will usually take a hypothetical proffer from the attorney that identifies general areas of subject matter, timing, sometimes specifics ..., sometimes more general information.... If, at that point, the proffer is generally acceptable to the prosecutor, we will advise the attorney that if the testimony comes in along those lines, we would be interested in accepting a fuller proffer.
Robert E. Bloch, et al., Representing Corporate Employees During an Antitrust Grand Jury Investigation, 56 Antitrust L.J. 901, 920 (1988) (statement of Judy L. Whalley, Deputy Director of Operations, Antitrust Division, United States Department of Justice).
Even after defense counsel has submitted a “fuller proffer,” it is virtually always the case that the prosecutor requires the witness to make a personal statement. Ah examination of this process makes it clear that the preliminary, hypothetical representations of counsel are not considered to be those of the witness:
There is no way that we can fully judge the candor, credibility, and cooperativeness of a witness without meeting with that witness directly.... In order to grant such reassurances as we can in that situation, we use proffer letters.... The proffer letter states that during the witness interview, loe ivill agree not to use thé statements of the witness directly against the witness in the future.
Id. at 920-21 (emphasis added); see also Fed.R.Crim.P. 11(e)(6) (statements made to government attorneys during the course of failed plea discussions not generally admissible against the defendant in subsequent proceedings). Obviously, if the parties to- the negotiation process in any way imagined that the prior statements of counsel could potentially incriminate the client-witness, those statements would be *566included within the scope of the proffer letter. It therefore speaks volumes that the typical letter omits any reference to counsel’s initial overtures.
It should be clear from the foregoing that proffers made by defense attorneys to prosecutors rarely encompass certainty. Most often, the proffer and negotiation process instead resembles a poker game, rife with understatement, bluff, and bluster. Lawyers for criminal defendants are understandably leery of turning up their hole cards, ie., their clients’ knowledge of other crimes, unless it is likely that they will garner a few chips in return. Most significantly, neither side contemplates that the informal proffer will ever be used by either side, or anyone else, for any purpose.
In practice, the process often does not entail even the limited structure and formality of the one described above. Many times the prosecutor and defense counsel will be familiar with each other from their past professional dealings. The attorneys may have developed a cordial relationship, and, away from the office, they may be friends. A chance encounter in the courthouse hallway can, in a matter of moments, migrate from amiable banter to discussion of a potential plea.
Any agreement arising out of this type of impromptu negotiation is bound to be sketchy, and may amount to nothing more than a tacit understanding to talk again later. Neither side will walk away from the meeting knowing precisely what it has bargained for, but each will be confident that it has given away little of substantive value.
Such are the realities of the modern plea negotiation process, and it was in light of these realities that the state habeas court specifically found that “Brown never made any inconsistent statement to the State regarding his being an eyewitness to the chase.” J.A. 990 (Memorandum of Opinion and Order of August 9, 1996, denying Spi-cer’s petition for post-conviction relief). This finding properly focuses on Brown’s statements, and not those of Christopher in his proffer to Sindler, rendered unreliable by the context in which they were made.
The state habeas court concluded that “neither Brady nor the line of cases following Brady required the State to inform defense counsel of a potential discrepancy between what Brown’s attorney indicated Brown knew and what Brown actually told the Prosecutor.” Id. The state court’s application of the law to the facts before it was not only “reasonable” within the meaning of § 2254(d), it was unassailable. The Supreme Court has never invoked Brady to grant habeas relief on facts remotely similar to those in this case.
In characterizing the difference between the statements of Brown and his lawyer as a “potential discrepancy,” the state habeas court acknowledged the reality that negotiation is something less than an exact science. Absent Christopher’s eventual testimony to the contrary, it is easy to imagine that he might have held back key pieces of information in order to gauge Sindler’s response to the tidbits already on the table. Under that likely scenario, Brown and Christopher would have shared a common understanding regarding the matters to which Brown could testify. Consequently, Christopher’s proffer would have been, at most, an incomplete account of the truth related to him by his client.5 That being the case, it could not be credibly argued that Sindler would have any con*567ceivable duty to disclose to Spicer or Sal-kin the details of the proffer.
Yet the majority holds that Sindler violated his duty in this case because, as it turned out, Brown’s testimony did not comport with Christopher’s understanding. Such Monday-morning quarterbacking unfairly makes a scapegoat of Sindler, who could quite reasonably assume that the story he was hearing from Brown was the same one that Christopher had been told, and that any variation between what Brown claimed to know and Christopher’s account of the same was wholly attributable to the latter’s negotiation tactics. Indeed, from Sindler’s perspective, it would have made little sense for Brown to have told him more than Brown had revealed to his own lawyer.
There is simply no way that Sindler could have known of Brown’s embellishment without investigating the matter further, i.e., contacting Christopher for the purpose of confronting Brown. It is well-established, however, that a prosecutor’s duty to the defendant does not extend so far. See United States v. Walker, 559 F.2d 365, 373 (5th Cir.1977) (“While Brady requires the Government to tender to the defense all exculpatory evidence in its possession, it establishes no obligation on the Government to seek out such evidence.”); United States v. Beaver, 524 F.2d 963, 966 (5th Cir.1975) (“Brady clearly does not impose an affirmative duty upon the government to take action to discover information which it does not possess.”).
Without the necessary follow-up, it was impossible for Sindler to conclude that Brown’s statements in his presence were in fact inconsistent with what Brown had told Christopher. Without the requisite inconsistency, Christopher’s bare proffer had no impeachment value, and was therefore not “favorable” to Spicer’s defense. Evidence that is not favorable to the accused need not be disclosed under Brady, as the state habeas court correctly observed. The state court’s interpretation of Brady in this case was not an “objectively unreasonable application of established [legal] principles to new facts,” requiring our correction under § 2254(d). Green v. French, 143 F.3d 865, 870 (4th Cir.1998), cert. denied, — U.S. -, 119 S.Ct. 844, 142 L.Ed.2d 698 (1999).
B.
The state habeas court’s irreproachable finding that the evidence of Christopher’s proffer was not favorable to Spicer’s defense is a sufficient basis, standing alone, to deny the relief requested in this case. For the sake of completeness, however, I will briefly address two other reasons why Spicer’s Brady claim must fail: (1) the proffer’s lack of materiality; and (2) Spi-cer’s ready access to the evidence of Brown’s supposed mendacity from a source other than Sindler.
1.
Evidence is “material” for Brady purposes “if there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” Strickler, 119 S.Ct. at 1948 (quoting Bagley, 473 U.S. at 682, 105 S.Ct. 3375) (additional citation omitted). The Supreme Court has noted that “evidence” that is inadmissible is not evidence at all, and thus cannot affect the outcome of trial. Wood v. Bartholomew, 516 U.S. 1, 6, 116 S.Ct. 7, 133 L.Ed.2d 1 (1995) (per curiam).
Christopher’s proffer to Sindler was plainly inadmissible under Maryland law. See Elmer v. State, 353 Md. 1, 724 A.2d 625, 630 (1999). Elmer involved the trial of two defendants for an unlawful shooting and related offenses. At issue was the prosecutor’s cross-examination of one defendant who had testified that it was he, and not his co-defendant, who had fired the weapon that had injured the victim. Attempting to impeach this testimony, the prosecutor asked the witness whether he had previously asserted the opposite, i.e., that his co-defendant had pulled the trig*568ger. The basis for the prosecutor’s question was a proffer by the witness’s lawyer during plea negotiations, in which counsel indicated that his client would testify that the co-defendant was the triggerman.
The Maryland high court held that the prosecutor’s attempt to get the substance of the proffer before the jury was improper, noting that “the entire area of inquiry was infused with the client/attorney privilege [and] the inadmissibility of plea bargaining discussions.... ” Id. at 12, 724 A.2d 625. The import of this holding is obvious: Christopher’s proffer in this case, like the proffer in Elmer, is not competent evidence. Because any testimony pertaining to Christopher’s proffer was required to have been excluded from Spicer’s trial, such evidence could not — under Bartholomew — be material within the meaning of Brady.6 Sindler’s failure to disclose the substance of the proffer wasj therefore, not a constitutional violation.7
2.
Even if Christopher’s proffer could somehow be classified as Brady material, the law is clear that Sindler need not have disclosed it to Salkin if it were “available to the defendant from other sources” through Salkin’s reasonable diligence. See United States v. Wilson, 901 F.2d 378, 380 (4th Cir.1990) (quoting United States v. Davis, 787 F.2d 1501, 1505 (11th Cir.1986)); accord, Barnes v. Thompson, 58 F.3d 971, 975 & n. 4 (4th Cir.1995).
In this case, Salkin actually spoke with Christopher prior to Brown’s appearance at Spicer’s trial. J.A. 618, 641 (state post-conviction testimony of James Salkin). Salkin testified that, during his considerable tenure as a criminal defense lawyer, he had occasionally questioned attorneys for adverse witnesses about specific statements their clients had previously made, for possible use on cross-examination. J.A. 621.
*569And indeed, this particular courthouse encounter began with a hint of revelation: Christopher mentioned “in a vague way” that Saltón had “better watch out for Larry Brown.” J.A. 618. Saltón believed that Christopher was trying to indicate to him that Brown “was a person you had to be wary of.” J.A. 621.
Revelation, however, was not to be had. Notwithstanding Christopher’s willingness to come forth and Salkin’s own past practice, Saltón did not inquire of Christopher as to the latter’s expectations regarding Brown’s testimony. Instead, he merely attempted to have Christopher elaborate on his rather nebulous warning:
[Saltón]:.... My recollection is I asked him what’s the problem, why are you sort of alerting me, and he was — I think he told me he couldn’t tell me.
Q: Did you ask him what it was that Mr. Brown was going to testify to at the trial?
A: I don’t know if I asked him that question, but I thought I knew the answer.
Q: What did you think the answer was?
A: That my client was the person he had seen run out of the building and run up the street.
Q: You never—
A: And that he knew him from before.
Q: But you never asked him that.
A: No. I just asked him what — you know, why — you know, what is the problem or what is the real — what is really going on.
J.A. 641-42.
Had Saltón asked Christopher a specific question regarding Brown’s probable trial testimony, he would have quickly discovered that Christopher’s impression was different from his own. Christopher, who “came up to [Saltón] and started talking” (J.A. 641), very likely would have been receptive to such an apparently innocuous question; as far as Christopher knew, his understanding of Brown’s story comported with his client’s official statement, the substance of which was already known to Sal-tón from the case file or otherwise.
Although the information that Saltón could have obtained from Christopher was of doubtful utility, the point is that it should have been obtained from Christopher; Sindler was not the sole available source. Our precedents simply do not permit the finding of a Brady violation under such circumstances.
III.
The majority’s grant of habeas relief in this case rests upon its conclusion that a prosecutor violated his duty to disclose to the defense that one of his witnesses made a “prior inconsistent statement” that was (1) not the witness’s statement; (2) not inconsistent; (3) not admissible in any event; and (4) reasonably available from an alternate source. The majority’s willingness to find a Brady violation on these unprecedented facts appears to stem from its concern that there is a “grave question over the fairness and accuracy” of the trial accorded Brady George Spicer. Ante, at 550.
The “grave question” concerning the reliability of the jury’s verdict centers on the testimony of an incarcerated drug dealer who purchased his freedom at the cost of Spicer’s. Brown was given the opportunity to avoid a potential twenty-year prison sentence because the Anne Arundel County prosecutors were more interested in solving a high-profile case involving the savage beating of a popular restaurateur in broad daylight.
There was nothing wrong with that decision; it is the job of prosecutors to ensure that the brunt of the state’s retributive power is brought to bear on those who commit the most egregious breaches of the peace. Often, however, the only way that justice can be served is by offering sufficient inducement to a minion of the dark (but notoriously disloyal) fraternity of hoodlums and thieves to throw his brother to the wolves. Before we can punish, we *570must prove, and such proof is rarely within the domain of the virtuous.
Our system of justice depends on the ability of those charged with its administration to wield the occasional carrot along with the stick. In order to remove from our midst the most serious offenders, prosecutors must be given the leeway to entice lesser wrongdoers to bear witness. On occasion, the enticement proves too strong, and these witnesses lie. Others, at a later date, merely claim to have lied; these recantations, not made under oath, are quite rightly viewed with a jaundiced eye.
The majority knows all this, of course. Yet it tinkers dangerously with a system proven to be highly reliable, out of fear that it may not be infallible. The inevitable result of this kind of tinkering is a system that may or may not be more reliable, but also one that will surely be less efficient, and perhaps even unworkable.
The contours of the new responsibilities thrust upon prosecutors by the majority are, at best, uncertain. How should a prosecutor determine whether a witness’s statement is actually “inconsistent” with an earlier proffer made by counsel? Are all potential discrepancies, regardless of their evidentiary value, now material for Brady purposes? Should prosecutors, as a prerequisite to negotiating a plea, require full written disclosure of everything to which a potential witness can testify? If so, will defense lawyers be willing to accept such a condition?
Assuredly, some lawyers will not be willing. As a result of the additional burdens on the bargaining process, it is likely that fewer pleas will be negotiated. The majority’s decision will thus ensure that while our system will trap its share of minnows, hooking the bigger fish will be ever more difficult. Although it today heeds those who claim that the net was cast too broadly, the majority may well be constrained tomorrow to lament the “one that got away.”
I would reverse the district court’s grant of the writ of habeas corpus in this case.8 I therefore respectfully dissent from the majority’s affirmance of the same.
. Neither Sindler nor Christopher were sufficiently interested in the details of the proffer to write anything down. Neither the prosecutor's file nor that of the witness's lawyer contained any writing confirming the contents or parameters of the proffer.
. Indeed, recalling his initial contact with Sindler regarding Brown, Christopher acknowledged that he was “not sure whether I told him what I knew at that time or whether I held back on that...." J.A. 601.
. I am dismayed at Christopher's apparent abandonment of his client at such a critical stage of the proceedings. Had his interview with Sindler not borne fruit, Brown faced the prospect of twenty years in prison. More importantly, had Christopher been present at Brown’s interview, any misunderstanding between lawyer and client concerning the extent of the latter’s involvement in the Spicer case would have been easily and immediately rectified.
Christopher’s conduct following Spicer's trial is even more troubling, and is not to be encouraged. Christopher acted in a manner that was consistently disloyal to his client. To begin with, he should have been — but was not — present while Brown testified at Spicer’s trial. Afterward, upon perceiving that Brown’s testimony was "inconsistent” with his recollection of what Brown had told him (notwithstanding his advice that Brown not provide him with "all the details,” see infra note 5), Christopher went directly to Spicer's attorney.
What Christopher failed to do, however, speaks much more loudly: (1) he failed to confront Brown with the supposed inconsistency, thereby giving his client an opportunity to clarify the point; (2) he failed to seek Brown's permission to disclose to others his confidential client communications; (3) he failed to make any effort to approach the prosecutor, Sindler, to clear up the issue; and (4) he failed to seek out the trial court for possible remedial steps, or to secure the court's authority to disclose his client’s confidential communications.
As a result of his actions (and inactions), Christopher found himself on the witness stand seeking to contradict his client’s trial testimony, accusing his client of a criminal act, and making improper disclosure of confidential client communications:
Once an attorney-client relationship has been established and privilege has attached to confidential communications between the attorney and client, the privilege is absolute, continuing ... to protect those communications as long as the confidentiality is preserved. Once protected by the privilege, a communication may not be the subject of compelled disclosure regardless of the need or good cause shown for such disclosure.
1 Paul R. Rice, Attorney-Client Privilege in the United States § 2:5 (2d ed. 1999); see Maryland Rules of Professional Conduct Rule 1.6(a) ("A lawyer shall not reveal information relating to representation of a client unless the client consents after consultation....”). Defendants and witnesses in criminal proceedings are entitled to greater loyalty from their lawyers — appointed or retained — than that demonstrated by Christopher in this case.
. Sindler explained that ”[i]t wasn't relevant to me at the time [of Christopher’s proffer] whether Brown would be an eyewitness ... or not. What was relevant was that Larry Michael Brown gave me the name of Mr. Spicer.” J.A. 874.
. The record does not indicate that Christopher at any time revealed to Sindler his belief that Brown did not witness Spicer’s alleged flight, or that Sindler was ever informed that Brown had denied being an eyewitness. Likewise, there is no suggestion that Christopher even once represented to Sindler that the proffer was a complete account of the evidence available from Brown. Indeed, it seems unlikely that Christopher believed that he had a full account from his client, inasmuch as he had advised Brown beforehand not to provide him with "all the details." J.A. 479.
. The Court in Bartholomew dismissed as "mere speculation” the reasoning of the court of appeals that the results of a polygraph test administered to a key prosecution witness, inadmissible at trial, were nonetheless subject to disclosure on the theory that the information “might have led [the defendant's] counsel to conduct additional discovery that might have led to some additional evidence that could have been utilized.” Id. at 6, 116 S.Ct. 7 (emphasis added). Similarly, in this case, it would be mere speculation to posit that the disclosure of Christopher's proffer would have led counsel for Spicer to discover admissible evidence useful for impeaching Brown’s credibility. As stressed by the majority, ante at 560-61, the entire case against Spicer consisted of the testimony of three eyewitnesses, including Brown. An investigation of Brown’s credibility would therefore have been crucial to Spicer’s defense in any event. It is difficult to see how such an important preparation would be conducted with any more care or diligence by counsel armed with the supposition that Brown intended to fabricate at least part of his testimony. Indeed, counsel had to assume that Brown was lying, insofar as his story was irreconcilable wiih Spicer's protestations of innocence.
. In its discussion of the materiality of the "impeachment evidence” that it hypothesizes would have negated Spicer's eyewitness testimony, the majority downplays the significance of the remaining evidence before the jury. Ante, at 561 ("If the jury doubted that Brown was an eyewitness, it would be left without any conclusive, or perhaps even persuasive, identification evidence”). Spicer, however, does not assert in this proceeding that the evidence was insufficient to convict him, and the majority apparently concurs in that assessment.
In my opinion, the evidence of Spicer’s guilt was substantial. Even if the jury had viewed with skepticism Brown’s account of the events on the day of the beating, it could yet credit Brown's testimony that Spicer had approached him afterward to thank him for not giving his name to those investigating the crime. Moreover, Henry Connick, the bartender at Armadillo’s, positively identified Spicer as the culprit in the attack on Denvir; although it is true that Connick's physical description of the man he chased did not match that of Spicer, such incongruities are not uncommon among persons who have undergone the stress of witnessing a crime from an uncomfortably short distance. Finally, Connick's identification was bolstered by that of Sam Novella, who witnessed the chase and testified that Spicer looked "very, very familiar.”
. I agree with the majority's conclusion that the district court inappropriately granted relief on the alternative ground that Spicer's