The Word That Built a Murder Conviction

I. The Opening

It was just before half past five in the morning on July 31, 2022, and the car was already split in two.

When the first Strongsville officers reached the brick wall of the PLIDCO building at the corner of Progress and Alameda, what they found did not look survivable. A Toyota Camry, torn into two pieces by the force of hitting that wall near 100 miles per hour. Three young people inside, none of them moving. The officers' first assessment was that all three were dead. Then, as they broke a window and began cutting through the airbags, they heard a sound. Mumbling. The driver was alive.

Her name was Mackenzie Shirilla. She was seventeen years old, pinned beneath the dashboard, badly hurt. The two people in the car with her were her boyfriend, Dominic Russo, twenty, and their friend Davion Flanagan, nineteen. Both were dying. When the firefighters finally worked Shirilla free, one of the first things she said, according to the officer who was there, was a question. She asked how Davion was.

A year later, a judge in Cuyahoga County would conclude that Mackenzie Shirilla drove that car into that wall on purpose. That she meant to do it. That the deaths of Dominic Russo and Davion Flanagan were not a horror that happened to her but a result she intended. The judge convicted her of murder, and Shirilla, now in her twenties, is serving fifteen years to life at the Ohio Reformatory for Women. Her first chance at parole comes in 2037.

I want to be clear at the very start about what this piece is and what it is not. This is not an argument that Mackenzie Shirilla is innocent. I am not going to tell you what happened inside that car, because I do not know, and neither does anyone else who was not in it. Her conviction is final. The Eighth District Court of Appeals affirmed it. The Ohio Supreme Court declined to hear it. The window for most of her challenges has closed, in one instance by a margin of a single day, and we will get to that. When a case reaches this stage, the law stops presuming a defendant innocent and starts presuming the conviction is valid. That is the correct starting point, and it is mine here. Mackenzie Shirilla stands convicted, and this piece treats her that way.

But a conviction being final is not the same as a conviction being beyond examination. And this one is worth examining, because of one word.

The State of Ohio did not convict Shirilla of reckless driving that killed two people. It did not convict her of a tragic, stupid, drug-adjacent teenage catastrophe. Ohio has charges for those things, serious ones, and she was convicted of some of them too. What Ohio convicted her of, at the top of the indictment, was murder. Purposeful murder. Under Ohio law, that word, purposely, means a specific intention to cause a particular result. It means the State proved, beyond a reasonable doubt, that when that Camry accelerated toward that wall, Mackenzie Shirilla specifically intended that Dominic Russo and Davion Flanagan die.

That is an enormous thing to prove. And the State proved it without a confession, without a note, without a witness to the decision, and, by the prosecution's own concession, without any motive, without any evidence that Shirilla was angry or upset with anyone in the minutes before the crash. It proved purpose the way intent is often proved, by circumstantial inference, by asking a factfinder to look at what a person did and conclude what must have been in her mind. There is nothing improper about that. Circumstantial evidence is real evidence, and intent is almost always proved this way. The question this piece asks is narrower and harder. It is whether the particular inference the State built, in this particular case, could honestly carry the weight of that particular word.

Here is why I think the question is fair to ask, and not just fair but necessary. This case was decided by a judge, not a jury. Mackenzie Shirilla waived her right to a jury and put her fate in the hands of a single person, and that choice had a consequence that turns out to matter a great deal. A jury returns a verdict and says nothing. Twelve people file out, and their reasoning goes with them, unspoken and unreviewable. A judge sitting as the factfinder does the opposite. A judge has to explain. And when Judge Nancy Russo explained her verdict from the bench, she said something that, once you understand it, does not sit easily next to a finding of purposeful murder. She said that whether Shirilla intended to kill herself in that same crash could "only be speculated," and she set that question aside as irrelevant.

That sentence is the reason for this piece. Not because it proves Shirilla innocent. It does not. But because of what it reveals about how the certainty of this conviction was built, and where, under pressure, that certainty turns out to be thinner than the verdict's language admitted.

So this is an examination of a process. How Ohio proves that a killing was intentional. How circumstantial evidence is supposed to work and where it strains. What a defense attorney is required to do, and what this one did not. How a filing deadline missed by one day can close a courthouse door for good. The case of Mackenzie Shirilla is a hard case, and hard cases are where you actually learn how the system works, because they are where its machinery is under load.

The honest conclusion, the one this piece will build toward and not flinch from, is not that an innocent girl was railroaded. It is something more uncomfortable and more useful. It is that this conviction is, at the same time, legally sustainable and genuinely contestable. It can stand under the law, and a reasonable person can still doubt it. Both of those things are true, and the space between them is exactly what citizens should understand about the system that decides these cases in their name.

Two young men died against that wall. Davion Flanagan's mother told the sentencing court that her son was so much more than the cargo the State's case sometimes reduced him to, that he was a person who would rescue a friend in the middle of the night, no questions asked. Dominic Russo had a family that has never stopped grieving him. Nothing in this piece is offered to diminish that, and nothing in it should be read that way. They deserved a full and careful accounting of how they died. Whether they got one is the question.

II. What Is Not in Dispute

Before any argument can be fair, the ground has to be clear. So before this piece questions anything, it is worth setting down what there is no reason to question at all. A great deal about the death of Dominic Russo and Davion Flanagan is simply settled. Not contested by the State, not contested by the defense, not contested here. Knowing what is settled matters, because it tells you exactly where the disagreement actually lives. And the disagreement, when you find it, is not about facts. It is about what a set of agreed facts means.

Here is the settled ground.

The people are not in question. Mackenzie Shirilla, seventeen years old at the time, was driving. Dominic Russo, twenty, her boyfriend, and Davion Flanagan, nineteen, their friend, were her passengers. The three of them had been together that night. No one disputes who was in the car.

The crash is not in question. In the early morning of July 31, 2022, on Progress Drive in Strongsville, Ohio, the Toyota Camry Shirilla was driving accelerated to roughly 100 miles per hour and struck the brick wall of the PLIDCO building head-on. Russo and Flanagan died of multiple blunt force injuries consistent with an automobile crash. Flanagan's lethal injuries were to his trunk and abdomen, Russo's to his head. Shirilla survived, pinned under the dashboard, with serious injuries to her leg and arm. None of this is disputed.

The car's data is not in question. The vehicle's event data recorder, its black box, captured the final seconds. In that window the accelerator was pressed fully down, the throttle held at one hundred percent, and the brake was never applied. The data also showed a hard steering input late in the sequence and indicated the gearshift had been moved during the crash. The defense did not dispute that this is what the recorder recorded. What the recorder's readings mean is contested, and we will come to that. That they exist, and say what they say, is not.

The toxicology is not in question, and it is worth pausing on, because it cuts against an assumption people bring to a case like this. Mackenzie Shirilla was not drunk. Alcohol testing was conducted and showed she had not been under the influence of alcohol at the time of the crash. There were drugs in the car, marijuana and psilocybin mushrooms, and the mushrooms were on her person. But the State did not prove at trial that Shirilla was impaired by any substance when the car hit the wall, and it did not build its case on impairment. This matters more than it might seem, and a later section will explain why. For now, simply note it as settled: this was not a conviction for driving under the influence, because the evidence did not establish that she was.

The earlier incident is not in question, at least not as to the fact that something happened. Two weeks before the crash, on or around July 17, 2022, there was an incident on a highway involving Shirilla and Russo, serious enough that Russo's family sent someone to bring him home. What exactly happened on that highway, and what it shows, is very much contested, and it becomes important later. That it happened is not.

The legal history is not in question. Because Shirilla was seventeen, the case began in juvenile court, which held a hearing to decide whether it belonged in adult court. Shirilla then waived her right to a jury. She was tried before a single judge, Nancy Margaret Russo of the Cuyahoga County Common Pleas Court, who is no relation to the victim Dominic Russo. The trial lasted two days in August 2023. The judge found her guilty of murder and other offenses and sentenced her to fifteen years to life. The Eighth District Court of Appeals affirmed the conviction in September 2024. The Ohio Supreme Court declined to review it. A later petition was dismissed as filed too late. None of that procedural history is in dispute.

And one more settled fact, one that does not get enough attention, because it is the quiet hinge of this entire case. When the medical examiner's office first examined the bodies of Dominic Russo and Davion Flanagan, it classified the manner of both deaths as accidental. Russo's autopsy report, dated October 2022, said accidental. Flanagan's, dated September 2022, said accidental. That was the official medical conclusion for months. It was only in March 2023, after the prosecutor's office sent the medical examiner additional materials, cell phone transcripts, photographs, video, witness interviews, that Dr. Joseph Felo amended both reports and changed the manner of death from accident to homicide. That this change happened, and when, and why, is not in dispute. The State concedes it. The amended reports are in the record. Whether the change was done properly is a separate question, and a later section takes it up. But the sequence itself, accident first, homicide second, after the prosecution supplied more material, is simply a fact.

Now look at what that list of settled facts does and does not contain. It contains a driver, a crash, a speed, two deaths, a data recorder, a clean alcohol test, a prior incident, a procedural path, and an autopsy that was changed. What it does not contain, anywhere, is the thing the State actually had to prove. It does not contain intent. Not one item on the settled list establishes, by itself, what was in Mackenzie Shirilla's mind when that car met that wall.

That is the entire shape of this case, and it is worth being precise about it. The dispute here is not factual. Almost nobody disagrees about what happened. The dispute is inferential. It is about what those agreed facts permit a factfinder to conclude. The State looked at the controlled turn, the accelerator, the absent braking, the prior incident, and concluded: purpose, beyond a reasonable doubt. The question this piece examines is whether that conclusion was actually available, whether the settled facts could carry the inference the conviction placed on them, or whether the inference reached past what the facts could hold.

Everything from here forward is about that gap. Not the facts. The leap.

III. The Charge That Decides Everything

Two cars leave the same curb. Both drivers are doing something they should not. One is texting and drifts into oncoming traffic. The other has decided, coldly, to aim the car at a particular person and kill him. Both cause a death. And the law treats those two drivers in completely different ways, gives them completely different charges, completely different sentences, completely different names for what they did. Same dead body. Same wrecked car. Wildly different crime.

That is not a flaw in the law. That is the law working exactly as designed. And understanding why is the only way to understand what was actually at stake in Mackenzie Shirilla's trial.

The criminal law does not just ask what you did. It asks what was in your mind when you did it. Lawyers have a Latin phrase for that mental piece, mens rea, the guilty mind, but the idea behind it needs no Latin. It is something every person already understands. There is a difference between the friend who knocks your phone off the table by accident and the person who grabs it and smashes it against the wall. The broken phone is identical. What separates them is intent. The law takes that ordinary moral intuition and builds an entire structure on it.

Ohio's structure has levels. The Revised Code sorts mental states into a ladder, and a person's place on that ladder, more than almost anything else, decides how serious the crime is and how long the sentence runs. At the top of the ladder is acting purposely. Below it, knowingly. Below that, recklessly. And below that, negligently. Same act, same result, but as you move down that ladder the crime changes and the punishment shrinks.

For this case, two rungs matter, and the whole trial lived in the gap between them.

The top rung is purpose. Ohio defines it directly. A person acts purposely when it is their specific intention to cause a certain result. Read those words slowly, because the State had to prove every one of them. Specific. Intention. To cause. A certain result. Not a vague bad mood. Not reckless disregard for what might happen. A specific intention that a particular result, here, the deaths of Dominic Russo and Davion Flanagan, come to pass. Murder, under R.C. 2903.02(A), is purposely causing the death of another. To convict Shirilla of murder, the State had to prove that she got into that car with, or formed along the way, the specific intention that those two young men die, and then acted to make that happen.

Now the lower rung. Recklessness. Ohio defines that too. A person acts recklessly when, with heedless indifference to the consequences, they disregard a substantial and unjustifiable risk that a result will occur. Sit with the difference. The reckless person is not trying to cause the result. The reckless person sees a serious risk, or should see it, and blows past it anyway. They do not want the bad outcome. They are indifferent to it. They gamble, and they lose, and someone dies. That is a real crime. It is a serious crime. But it is not murder, because the reckless mind and the purposeful mind are not the same mind.

Ohio has a charge built precisely for a death on the road caused by that reckless mind. It is called aggravated vehicular homicide, R.C. 2903.06. It exists for exactly this kind of tragedy, a driver whose conduct was dangerous and inexcusable, whose driving killed people, but who did not get behind the wheel intending to kill. And here is a fact that matters enormously and that often gets lost: Mackenzie Shirilla was convicted of that too. The aggravated vehicular homicide counts were part of the verdict. Nobody in this case, not the State, not the defense, not this piece, disputes that what happened on Progress Drive was at minimum a criminal homicide. The wreck was not lawful. Someone is criminally responsible for it.

So the trial was never about whether a crime occurred. It was about which crime. It was about which rung of the ladder Mackenzie Shirilla was standing on when that car accelerated toward the wall. And those two rungs are not close together. The vehicular homicide path and the murder path are different worlds. They carry different sentences. They tell different stories about who the defendant is. A reckless driver who kills is a person who did something catastrophically dangerous. A purposeful murderer is a person who decided two people should die and used a car to accomplish it. The distance between those two descriptions is the distance between a terrible mistake and an execution.

This is why one word carried the entire case. The State did not have to prove Shirilla was a careful driver, she plainly was not. It did not have to prove the crash was avoidable, it plainly was. It did not even have to prove she was a good person. It had to prove one specific thing about one specific moment: that the mental state behind the act was purpose, the top rung, and not recklessness, a lower rung. If the State proved purpose, it was murder, fifteen years to life. If the evidence showed recklessness, or showed something the evidence simply could not pin down, then it was not murder, whatever else it was.

And notice what that does to the burden. The prosecution did not get to prove "something bad and intentional-ish" and let the court round up. Our system does not work by approximation. The State chose to charge purposeful murder, the most serious available characterization of this crash, and having charged it, the State had to prove that exact mental state beyond a reasonable doubt. Not probably purpose. Not purpose-or-something-near-it. Purpose. The specific intention to cause those deaths, established to the highest level of certainty the law knows.

Hold that standard, because it is the measuring stick for everything that follows. When this piece examines the State's evidence, the controlled turn, the accelerator, the absent braking, the prior incident, the question is never going to be "does this look bad." Of course it looks bad. Two young people are dead. The question is always going to be the precise one the law actually asked: does this evidence prove, beyond a reasonable doubt, the specific intention that they die. That is the bar. It was set that high on purpose, by a system that decided long ago that the difference between a tragedy and a murder is too important to guess at.

The whole case is one word. The rest of this piece is about whether the State reached it.

IV. How the Law Proves What Was in a Mind

There is a problem at the center of every intent case, and the law has never fully solved it. It has only built tools to manage it. The problem is this: no one can see inside another person's head. Intent is invisible. When the State has to prove that a defendant acted purposely, that she had the specific intention to cause a result, it is trying to prove a fact that exists, if it exists at all, only in the privacy of one person's mind at one vanished moment. There is no instrument that reads it. So how does a courtroom ever establish it?

The answer is circumstantial evidence, and the first thing to understand about circumstantial evidence is that it is not second-rate evidence. People hear the word circumstantial and think weak, as if it means the State was reaching. Ohio law says the opposite, flatly. Circumstantial evidence and direct evidence carry equal weight. The Ohio Supreme Court has said so plainly, and it makes sense when you think about it. Direct evidence is a witness who saw the thing happen. Circumstantial evidence is a set of facts from which the thing can be inferred. You wake up, the ground is wet, the sky is gray, the gutters are dripping. No one saw it rain. You know it rained. That is circumstantial evidence, and only a fool would call it weak.

So intent gets proved circumstantially, by inference, all the time. It almost has to be. Defendants rarely announce their mental states. The factfinder looks at what a person did, the surrounding facts, the way the act was carried out, and reasons backward to what the person must have meant. That is legitimate, and it is ordinary, and nothing in this piece disputes it.

But an inference is a chain, and a chain has to actually hold. Which brings us to the specific tool the State used here, a doctrine old enough in Ohio to have collected nearly a century of dust, and important enough that the whole Shirilla case turns on how it is applied. It is called the natural and probable consequences inference. Stated simply, it works like this: when a person voluntarily does an act, the law permits a factfinder to infer that the person intended the natural and probable consequences of that act. If you fire a gun into someone's chest, you do not get to say you only meant to make a loud noise. Death is the natural and probable consequence of shooting someone in the chest, and the law lets a factfinder infer you intended it.

Notice the word permits. This is the most important word in the doctrine, and it is the one most easily lost. The natural and probable consequences rule permits an inference. It does not command one. It does not create a presumption that flips the burden onto the defendant to prove she did not intend the result. It is a door the factfinder may walk through, not a track the factfinder is locked onto. The State still has to prove intent. The doctrine just says that the consequences of a voluntary act are evidence a factfinder may use in getting there. Used carefully, it is a sound and sensible tool. Used carelessly, it becomes something dangerous: a way to skip the proof of intent entirely by pointing at a bad outcome and working backward, as if every terrible result must have been somebody's plan. The line between those two uses is thin, and a later section will show you exactly where this case lands on it.

Now the standard of proof, because an inference does not just have to be available, it has to be strong enough. In a criminal case the State must prove every element of the offense beyond a reasonable doubt. That is the highest standard of certainty the law has. It is not "more likely than not." It is not "probably." It is the standard we reserve for the moments when the state proposes to take a person's liberty, and it is set deliberately high because our system would rather absorb the cost of a guilty person going free than the cost of convicting an innocent one. For Mackenzie Shirilla, that meant the State did not have to make intent look plausible. It had to prove the specific intention to kill, by inference if necessary, to that highest level of certainty.

And here is a piece of the burden of proof that matters in this case and that people routinely get wrong. Reasonable doubt is not a sliding scale. It does not work like a thermostat where, if the State cannot quite reach murder, the factfinder dials the conviction down to something smaller and convicts on that instead, just to be safe. That is not how the burden works. For each charge, the question is binary: did the State prove this offense, this mental state, beyond a reasonable doubt, yes or no. If the evidence genuinely cannot establish what was in a defendant's mind, the answer is not "well, convict her of the lesser thing." The answer is that the State did not meet its burden on the greater thing. Uncertainty does not earn the prosecution a consolation conviction. It is supposed to cost the prosecution the charge it could not prove. Hold onto that, because it becomes very important when this piece reaches the question of what Mackenzie Shirilla's own defense attorney chose to argue.

There is one more thing to understand, and it is the piece that explains why this conviction can be, at the same time, both contestable and standing. It has to do with what an appeals court is actually allowed to do.

When a convicted defendant appeals and says the evidence was not enough, there are two distinct challenges she can raise, and they are not the same, though they sound alike. The first is a sufficiency of the evidence challenge. Sufficiency asks a narrow, almost mechanical question: viewing all the evidence in the light most favorable to the prosecution, could any rational factfinder have found every element proven beyond a reasonable doubt. Read that carefully, because it is doing something specific. On a sufficiency review, the appeals court is not asking whether the evidence convinced the appeals court. It is not weighing anything. It is asking only whether a reasonable factfinder could have gotten there. That is a low bar for a conviction to clear, and it is low by design. The sufficiency standard exists to catch convictions with a true hole in them, the ones no reasonable factfinder could have reached, not to let an appeals court re-decide a close call it simply would have called differently.

The second challenge is a manifest weight of the evidence challenge. Manifest weight is different. Here the appeals court is allowed to weigh. It sits, in the Ohio Supreme Court's phrase, as a thirteenth juror, and it asks whether the factfinder clearly lost its way and created such a miscarriage of justice that the conviction must be reversed. That sounds more powerful, and in theory it is. But it comes wrapped in heavy deference. The appeals court is told to reverse on manifest weight only in the exceptional case, because the original factfinder, the judge or jury who was actually in the room, saw the witnesses and heard the testimony and is presumed to be in the better position to weigh it.

Put those two standards together and you can see the machine the rest of this piece runs into. Both of them are built to defer to the factfinder. Both of them ask not "was this verdict right" but "was this verdict allowable." A verdict can be genuinely debatable, a verdict a careful person would not have reached, and still be a verdict that some rational factfinder could have reached, which means it survives. That gap, between a verdict that is correct and a verdict that is merely permitted, is not a malfunction. It is the deliberate design of appellate review. But it is also the precise space in which a case like this one lives. The chapters ahead are going to argue that the Shirilla conviction is genuinely contestable. The deference built into these standards is a large part of why contestable has not been enough to undo it.

That is the toolkit. Circumstantial evidence, equal in weight to direct. The natural and probable consequences inference, which permits but never commands. Proof beyond a reasonable doubt, which is not a dial. And a system of appellate review built, top to bottom, to defer. Keep all four close. Everything from here forward is the State's case, and then this conviction, run through exactly these tools.

V. The Case the State Built

If you want to understand why a judge convicted Mackenzie Shirilla of murder, you have to be willing to sit with the prosecution's case as the prosecution meant it to be heard. Not the cartoon version. The real one. Because the State's theory was not weak, and pretending it was will not teach anyone anything. The State built a case that a serious judge found convincing beyond a reasonable doubt, and the only honest way to question that result is to first give the case its full weight.

So here is the State's story, told straight.

It begins two weeks before the crash. On a day in July 2022, Christopher Martin was sitting with Dominic Russo's mother when Dominic called, upset. Something was wrong. Martin drove out to find him and located the car Shirilla was driving, with Dominic in the passenger seat, pulled over on the shoulder of Interstate 71. Martin was on the phone with Dominic as the cars came to a stop, and he heard the two of them arguing. And during that argument, Martin testified, he heard Mackenzie Shirilla say five words: "I'm going to wreck this car right now." He watched a tussle inside the vehicle, Shirilla swinging her hands. Dominic got out, got into Martin's car, and Martin drove him home.

Hold that sentence, because the State did. "I'm going to wreck this car right now." Two weeks before Shirilla drove a car into a wall, a witness heard her threaten to do that exact thing, in anger, to a car with Dominic Russo in it. For a prosecutor, that is not a detail. That is a spine.

Then move to the morning of July 31. Around 5:34 a.m., a city surveillance camera captured the Toyota Camry approaching from Pearl Road. Detective Zaki Hazou described what the video showed: not a car already out of control, not a car careening, but a Camry making a controlled turn from Pearl onto Progress Drive. Controlled. The State leaned hard on that word, and you can see why. A controlled turn is a driver in command of the vehicle. Whatever happened next began from a position of control.

What happened next is the part that gave the case its name. Progress Drive is not a road anyone races down by accident. A Strongsville patrol officer who had driven it hundreds of times described it as a rough, older, concrete road with raised sections you can feel through the wheel, a road with a series of curves and a stop sign partly hidden by tree cover. He testified that he could safely drive it at 54 miles per hour, and that even at 54 he had to grip the wheel to stay in control. The speed limit is 35. And on that road, in the dark, the Camry accelerated. A reconstruction expert put its speed at roughly 97 miles per hour before it left the roadway. It went airborne over a curb, came down, crossed an intersection, and struck the brick wall of the PLIDCO building dead-on.

Then the data. The State's most powerful witness was not a person. It was the car's own event data recorder, the black box, and what it recorded was, for the prosecution, the heart of everything. In the seconds before impact, the accelerator was pressed fully to the floor. The throttle held at one hundred percent. And across the entire preserved window, roughly four and a half seconds, the brake was never touched. Not tapped. Not pressed. Never. The car accelerated into a wall and no one tried to stop it. The State's vehicle expert testified that he could not recall ever investigating a crash where the data showed a full accelerator application for the entire duration of an incident with no braking at all. There was more: the gearshift, the data indicated, had been manually moved during the sequence, and a gearshift does not move itself.

Now the State closed the doors. This is the part of the case that prosecutors knew they needed, because a crash like this invites the question of whether something other than intent caused it. So the State brought experts to rule the alternatives out. Medical professionals examined Shirilla and testified they found no evidence of a neurological condition, no seizure disorder, no medical event that could explain the crash. A mechanical expert examined the vehicle and found no defect, no malfunction, nothing wrong with the car that could have caused it to do what it did. No medical cause. No mechanical cause. The State's point was blunt and effective: if it was not her body failing, and it was not the machine failing, then what is left is a choice.

And then the route itself. GPS data from Shirilla's phone placed her in the area of Progress and Alameda, the very area of the crash, three days earlier, on July 28. To the prosecution, that was not a coincidence. That was scouting. A driver who had been to the place where the road ends, who knew the obscure cut-through, who knew where the wall was.

Stack it the way the State stacked it. A documented threat to wreck a car, with this victim, two weeks earlier. A controlled turn onto a road no one accelerates down without meaning to. An obscure route the driver had visited days before. A car taken to nearly 100 miles per hour on a rough industrial road in the dark. An accelerator held to the floor. A brake never touched. A gearshift moved by a hand. Medical and mechanical causes affirmatively excluded by experts. And a driver who, even in those final seconds hurtling toward a wall, made no move to save the two people beside her or herself.

That is the case. And I want to be direct about something, because the rest of this analysis depends on it. That case is not nothing. A judge looked at that evidence and found it amounted to purpose beyond a reasonable doubt, and a reasonable person can understand how she got there. The pattern the State assembled has a terrible internal logic. Each piece points the same direction, and together they tell a story of a deliberate act. When Judge Russo called the crash controlled, methodical, deliberate, she was not inventing something. She was describing what the State's evidence, taken at its strongest, was built to show.

This is why the questions that follow in this analysis are hard questions and not easy ones. If the State's case were thin, picking it apart would be simple, and it would also be unimportant. It is because the case is genuinely substantial that the places where it strains are worth examining with care. A serious case deserves serious scrutiny, not dismissal. The State built something real here. The question for the rest of this piece is whether what it built can actually carry the weight of the word purposely, or whether, examined closely, it carries something less.

VI. The Speculation Problem

Here is the sentence the whole case turns on. When Judge Nancy Margaret Russo delivered her verdict on August 14, 2023, she said that whether Mackenzie Shirilla intended to kill herself that morning could "only be speculated," and that the question, "as a matter of speculation," had "no relevance to the weight of the evidence in this case."

Read that twice. A judge sitting as the sole factfinder in a double murder trial, explaining on the record why she was about to convict, paused on the question of whether the defendant meant to die in the same crash, called the answer speculation, and set it aside as irrelevant. Then she found beyond a reasonable doubt that the same defendant, in the same act, purposely killed the two young men in the car with her.

I want to be careful and exact about what I am arguing, because this is the part of the case that gets mishandled most. I am not saying the judge slipped. I am not saying she let something revealing fall out by accident. She did not. She addressed the suicide question head-on, she labeled it, and she made a deliberate ruling that it did not matter. That was a considered legal decision. My argument is that the decision was wrong, and that the wrongness of it exposes something about the conviction that the verdict's confident language was built to paper over.

Start with what the State had to prove. Murder under R.C. 2903.02(A) requires that a person purposely cause the death of another. Purpose, under R.C. 2901.22(A), is a specific intention to cause a particular result. Not carelessness. Not a terrible choice that ended in death. A specific intention that those people die. By the time of trial, everyone in that courtroom agreed this was the only real question. Shirilla was driving. Two passengers were dead. The crash happened. The single contested element, the hinge of the entire case, was whether she did it on purpose.

Now look at how the State proved purpose, because the method matters. There was no confession. There was no note. The prosecution conceded it could not show Shirilla was angry at Dominic Russo or upset with anyone in the minutes before the crash. What the State had was the physical record: a controlled turn onto an obscure road, an accelerator pressed to the floor, no braking, a car driven near 100 miles per hour straight into a brick wall. From that physical record, the State asked the judge to infer a mind. The reasoning runs through a doctrine Ohio courts have used for almost a century, the natural and probable consequences inference. When a person voluntarily does an act whose natural and probable consequence is death, a factfinder may infer that the person intended that death. Drive a car at a wall at 100 miles per hour and death is the natural and probable consequence. Therefore, the inference goes, the driver intended it.

Hold onto that inference, because everything depends on what it actually says. It does not say "the driver intended to kill the passengers." It says "the driver intended the natural and probable consequence of the act." And here is the fact the verdict could not get around: the act was one act. One car. One wall. One speed. One impact. The driver sat in that car and met that wall at the same hundred miles per hour as everyone else in it. Whatever the natural and probable consequence of that act was, it was the same consequence for every person inside. The physics did not distinguish between the front seat and the driver's seat.

So the inference, honestly applied, runs in every direction the crash ran. If the act was so plainly lethal that we can infer the driver intended the passengers to die, the same act was just as lethal to the driver, and the same inference points just as hard at her own death. You cannot run that inference at full strength toward the passengers and then, facing the driver, suddenly call it speculation. It is the identical inference from the identical act. Either it is strong enough to carry a finding of purpose, in which case it carries that finding for everyone the act endangered, or it is too weak to carry that finding, in which case it does not carry it for the passengers either.

The judge's theory of guilt made this problem unavoidable. She did not describe a moment of rage or a sudden loss of control. She described the opposite. Her words were "controlled, methodical, deliberate, intentional, and purposeful." She said Shirilla "morphs from a responsible driver to literal hell on wheels." She said Shirilla "had a mission" and "executed it with precision," and that "death was the ultimate goal that day." That is a description of a plan. A deliberate, methodical, executed plan to drive a car into a wall at full speed. And a deliberate plan to drive a car into a wall at full speed is, by its own internal logic, a plan in which the driver dies too. You cannot methodically execute that act and methodically exclude yourself from its consequences. The premeditation theory the judge chose did not leave the driver's death off to the side as some separate, optional question. It put the driver's death right in the center of the plan.

Which is why the speculation ruling does what it does. The judge reached the one question her own theory could not survive, the question of whether this methodical executioner also meant to die in the act she was executing, and instead of answering it, she ruled it out of bounds. She called it speculation and declared it irrelevant. And once you see that move for what it is, you see the trouble. The judge was certain enough about the inference to convict on it. She was not certain enough about the same inference, pointed at the driver, to even venture an answer. Those two positions cannot both be true about a single indivisible act. If the evidence genuinely supported a finding of methodical, purposeful intent to cause death by that crash, it supported that finding across the whole crash. If it did not support that finding as to the driver, then what the evidence actually showed was something short of methodical purpose. It showed a catastrophe whose internal mental state the evidence could not fully reach. And something short of methodical purpose is not murder under R.C. 2903.02(A). It is one of the lesser things the law has names for.

This is where I need to be precise about the law, because there is a real counterargument and I am not going to hide from it. The State will say, correctly, that R.C. 2903.02(A) requires purpose to kill another, not purpose to kill oneself. Shirilla's intent toward her own life is not an element of the offense. The judge found purpose as to Russo and Flanagan; the suicide question was about Shirilla herself; therefore, the State will say, the speculation comment was about something the statute does not even ask. As a statement about the elements of murder, that is right, and I will not pretend otherwise.

But my argument was never that self-intent is a missing element. My argument is about the reliability of the inference the State used to prove the element that does matter. The suicide question is not relevant because the law required the State to prove Shirilla wanted to die. It is relevant because it is a test of the inference. The State's whole case for purpose rested on the claim that this act was so unambiguously a deliberate killing that intent could be read straight off the physics of it. The judge's own inability to read that same intent off the same physics, when the person in question was the driver, is direct evidence that the inference is not as clean as the conviction needs it to be. When the factfinder herself cannot follow her own inference to the place it plainly leads, that is not a side issue. That is the inference failing a test in open court.

Ohio law gives a reviewing court the room to see this. A bench verdict is not a black box. When a judge tries a case without a jury and explains her findings on the record, those findings are part of the record, and an appellate court reviewing the sufficiency and the weight of the evidence reviews them. In Parma Heights v. Barber, the Eighth District, the same court that would hear any appeal in this case, reversed a bench conviction for insufficient evidence because the trial judge's own stated findings showed that the specific purpose the offense required had not been proved. In State v. Harshbarger, the Third District refused to simply presume a missing purpose element from a guilty verdict when the judge's own detailed findings did not support it. The principle in both is the same and it is not exotic. A judge's articulated reasoning can reveal that an element was never actually established, and when it does, the verdict cannot stand on the strength of its own confidence.

I want to separate this cleanly from an argument it is not, because the difference is the whole ballgame. This is not a claim that the verdict is inconsistent and therefore void. Ohio law, like federal law, generally does not throw out verdicts for internal inconsistency, and it does not treat a judge sitting as factfinder differently from a jury on that score. If the argument here were "the judge contradicted herself, so reverse," it would lose, and it would deserve to. That is not the argument. The argument is narrower and it is grounded in the one kind of review that always survives: sufficiency. The question on sufficiency review is whether the evidence, taken in the light most favorable to the State, could convince a rational factfinder of every element beyond a reasonable doubt. The judge's speculation ruling is not offered as a contradiction to be punished. It is offered as evidence, from the factfinder's own mouth, on the record, that the inference of purpose could not actually bear the weight the verdict put on it. State v. Walker tells us, by clear analogy, that speculation cannot supply a contested mental element. Here the factfinder used the word herself, about the very inference the conviction depended on.

Strip it down to its frame and the problem is plain. The State proved a physical act and asked that a specific intent be inferred from it. The act was indivisible. The inference, run honestly, reached the driver as surely as it reached the passengers. The factfinder ran it hard in one direction and called it speculation in the other. A conviction for purposeful murder cannot rest on an inference the factfinder herself would only follow halfway. That is the speculation problem, and no amount of certainty in the language of the verdict makes it go away.

VII. The Arguments This Piece Will Not Make

A careful reader, having reached this point, may already be running ahead of the argument. The keystone in the last section rested on the judge's own words, on a finding she made and then set aside. And a reader who knows a little law, or who simply has a good ear for where an argument seems to be heading, might reasonably expect this piece to take one of two further steps. They are tempting steps. They would make the argument feel bigger. And this piece is not going to take either of them, on purpose, for reasons worth laying out in the open, because the reasons are themselves a lesson in how legal argument actually works.

Here is the first place a reader might expect this to go. If the judge found one thing she could "only speculate" about and another thing she was certain of, both from the same act, why not simply argue that the verdict contradicts itself, that it is internally inconsistent, and that an inconsistent verdict cannot stand? It has a satisfying shape. It feels like catching the verdict in a logical trap with no exit.

This piece does not make that argument, because that argument loses, and understanding why it loses is worth more than the argument ever could be. The law does not throw out verdicts for being internally inconsistent. The United States Supreme Court settled the basic rule in Dunn v. United States and reaffirmed it in United States v. Powell: consistency between verdicts is not required. A verdict might be inconsistent because of compromise, or leniency, or some feature of how the decision was reached that no outsider can reconstruct, and rather than guess, the law lets it stand. Powell said something more, and it is the part that matters here. The Court explained that the protection against an unsupported conviction is not inconsistency review. It is sufficiency review. You do not undo a conviction by showing it sits awkwardly beside another finding. You undo it by showing the evidence could not support it.

Ohio follows the same path. In State v. Lovejoy, the Ohio Supreme Court held that the separate counts of an indictment are not interdependent. And there is one more wall the inconsistency argument would hit, the highest one. A reader might think that because a judge decided this case, and because a judge has to explain her reasoning out loud while a jury never does, a judge can be held to a stricter standard of consistency. The Supreme Court rejected that precise idea in Harris v. Rivera, holding that an apparent inconsistency in a judge's findings does not by itself establish any constitutional violation. Ohio's appellate courts have said the same thing more than once, including the Eighth District, the very court that would hear any appeal in this case. Hold a judge to a stricter consistency rule than a jury, and the law tells you no.

So this piece does not argue inconsistency. But notice, carefully, what that means for the argument it does make, because the two look like cousins and they are not. The inconsistency argument says: two findings clash, therefore the verdict is void. The argument in the last section says something different in kind. It says the factfinder's own stated reasoning is evidence, evidence that the element of purpose was never proven to the standard the law demands, and that the conviction therefore fails for insufficiency. That is not inconsistency review. That is sufficiency review, the exact protection the Supreme Court pointed to in Powell as the real one. The judge's speculation finding is not offered here as a contradiction to be punished. It is offered as proof, from the factfinder herself, that the inference of purpose could not carry the weight the verdict placed on it. The cases that bury the inconsistency argument do not lay a finger on that one. They are answering a different question.

Here is the second place a reader might expect this to go, and it is bolder. If the judge said she could "only speculate" about a piece of what happened, why not argue that she effectively acquitted Mackenzie Shirilla in that moment, and that double jeopardy, the constitutional protection against being tried twice for the same offense, should now bar the conviction?

This piece does not make that argument either, and this one fails even faster, because it asks a word to mean something it does not. An acquittal is a specific thing. It is a ruling that the evidence was legally insufficient, a finding of not guilty. The judge in this case made no such ruling. She made the opposite one. She said the evidence "clearly demonstrates" that Shirilla acted "purposely and intentionally," and she entered findings of guilty. There is no acquittal here to build on, in form or in substance. There is a conviction this piece argues is contestable, and a contestable conviction and an acquittal are not the same thing and cannot be made into the same thing. The double jeopardy argument is not weak. It is simply unavailable, because the thing it would need as its foundation was never created.

Now, why spend a section of a long piece explaining two arguments only to set them down? Because the discipline is the point. The space around a case like this fills up with the most dramatic argument anyone can imagine, and the most dramatic argument is almost always the most fragile one. Inconsistency and acquittal are tempting for the same reason: each promises a clean and total victory, a verdict that simply evaporates. Real legal argument rarely offers that, and when an argument promises it, that promise should raise your suspicion rather than your hope.

The argument this piece makes is deliberately narrower, and the narrowness is the strength. It does not claim the verdict is void. It does not claim Shirilla was acquitted. It does not claim she should walk out tomorrow. It claims something smaller and far harder to knock down: that the evidence, measured against the law's actual standard, did not prove purposeful murder beyond a reasonable doubt, and that the factfinder's own words show it. A modest argument that survives scrutiny is worth more than a thrilling one that does not. That principle is true in this case, and it is true in every case, and a reader who carries only that one idea out of this piece will watch the justice system more clearly than most people ever do.

This piece could have gone bigger. It chose to go solid instead. The rest of the analysis is built on that choice.

VIII. The State's Best Answer

A piece that only states the strong version of its own argument is not analysis. It is advocacy wearing analysis as a costume. So before this piece goes further, it owes the reader the prosecution's best response to the keystone argument, stated at full strength, with nothing held back and nothing softened. Because the State does have a response. It is a serious one. And the honest answer to it is not that it is wrong. The honest answer is more interesting than that.

Here is the State's best counter.

The keystone argument leans on the judge's finding that Shirilla's intent to kill herself could "only be speculated." But look closely at the murder statute, the State would say. R.C. 2903.02(A) makes it murder to purposely cause the death of another. Another. Not oneself. The statute does not ask, anywhere, whether the defendant intended to die. A defendant's intent toward her own life is simply not an element of the offense. So when the judge said the suicide question was speculative, she was talking about something the statute does not require the State to prove in the first place. She separately, and clearly, found that Shirilla acted with purpose to kill Russo and Flanagan. That finding stands on its own. The speculation comment was about a different question, a question the law does not even pose, and a finding about a non-element cannot undermine a finding about a real one.

And the State would press further, because the counter has a second move. A human being is fully capable of holding two different mental states toward two different things at the same time. The law has never required a person's intentions to be symmetrical. A person can intend to kill someone else and be reckless, or indifferent, or fatalistic, or simply not thinking clearly about her own survival. People in extreme states are not models of logical consistency. So there is nothing incoherent, the State would argue, in a factfinder concluding that Shirilla specifically intended the deaths of her passengers while feeling something murkier, something the evidence could not pin down, about her own. The asymmetry the keystone argument treats as a crack in the verdict is, the State would say, just the ordinary untidiness of a real human mind.

This is not a weak argument. I want to say that plainly, because the temptation in a piece like this is to set up the other side's position so it can be knocked over. That is not what is happening here. The State's counter is correct about the statute. R.C. 2903.02(A) really does require purpose as to another, and Shirilla's intent toward herself really is not, formally, an element of murder. If the keystone argument were a claim that the State failed to prove a missing element called self-intent, the State's counter would defeat it cleanly, and this piece would have to concede the point and move on.

But that is not what the keystone argument claims. And the distinction is the whole of this section.

The keystone argument is not a claim about a missing element. It is a claim about the reliability of an inference. Go back to how the State proved purpose in the first place. There was no confession. There was no statement of intent. The State proved purpose circumstantially, by asking the factfinder to look at the physical act, the accelerator, the wall, the speed, and infer the mind behind it through the natural and probable consequences doctrine. The entire conviction rests on the strength of that one inference: this act was so unambiguously lethal that intent to kill can be read directly off it.

So the question that actually matters is not "did the State prove self-intent." Of course it did not have to. The question is "how reliable is the inference the State did use." And that is exactly the question the speculation finding speaks to. The judge was asked to read intent off the physics of this crash. When the person whose intent was being read was a passenger, she read it with total confidence, purpose, beyond a reasonable doubt. When the person was the driver, exposed to the identical physics of the identical act, she could not read it at all. She called it speculation.

That is not a problem because self-intent is an element. It is a problem because it is a test of the instrument. The State's instrument for proving purpose was the claim that this act's lethality was so plain that intent could be inferred straight from it. The factfinder then demonstrated, with her own words, that the instrument did not actually work that way, that the same act, the same lethality, yielded a confident reading in one direction and no reading at all in the other. When the factfinder herself cannot run her own inference consistently across a single indivisible act, that is direct evidence, from the most authoritative possible source, that the inference is not as reliable as a conviction for purposeful murder requires. The speculation finding is not offered to prove a missing element. It is offered to show that the inference used to prove the element that does exist could not bear the weight.

So where does that leave the two sides honestly? It leaves them split, and a fair reader should see it that way. On the question of the elements, the formal structure of the statute, the State has the better of it. Purpose as to another is the element, self-intent is not, and that is simply correct. On the question of the inference, whether the circumstantial reasoning that produced the purpose finding was reliable enough to support proof beyond a reasonable doubt, the defense critique has real force, and the factfinder's own speculation finding is the strongest evidence of that force. Both of those things are true at once. A reader does not have to pick a team to see that the State is right about one thing and vulnerable on another.

It is worth noting that courts in other states have wrestled with the underlying scenario, an intentional crash by a driver who also endangered herself, and those cases mostly cut the State's way on the narrow legal point. Courts in Indiana and Arkansas have upheld convictions where a driver who was arguably suicidal still intentionally killed or tried to kill a passenger, reasoning that a person's willingness to die does not negate an intent to kill others. A Texas court reached a more mixed result, accepting that such a crash could support a murder finding while holding the defendant was entitled to have a lesser charge considered. None of these is Ohio law, and none of them is binding here. They matter because they confirm the honest shape of this section: as a matter of formal doctrine, self-endangerment does not defeat a murder charge, and a careful piece says so out loud. The defense argument was never that it does. The defense argument is about whether the inference of purpose, in this specific case, on this specific record, was ever reliable enough to convict. That question those out-of-state cases do not answer, because that question turns on what this factfinder said about this evidence.

The State has a real answer. It is right about the statute. It is just not an answer to the argument actually being made.

IX. What the Evidence Could Not Rule Out

The keystone argument, the one this piece has built and defended, is about a single inference and whether the factfinder's own words showed it could not bear the weight of a murder conviction. That is the sufficiency argument. But there is a second, related way to test a verdict, and Ohio law provides it. It is the manifest weight of the evidence challenge described earlier, the one where a reviewing court is allowed to sit as a thirteenth juror and ask whether the conviction is one the evidence genuinely supports.

This section is about that test. And it asks a question the sufficiency argument does not quite reach: not whether the factfinder's reasoning was sound, but whether the evidence itself, the actual record built at trial, was ever capable of doing what a purposeful murder conviction requires it to do. The answer, examined honestly, is troubling. Because the record is full of doors the State needed to close and did not.

Here is the principle, and it is worth stating plainly before the facts, because the facts only mean something once the principle is clear. To convict of purposeful murder, the State did not merely have to make intent the most attractive explanation. It had to prove intent beyond a reasonable doubt, and proving something beyond a reasonable doubt means excluding the reasonable alternatives. If the same evidence is genuinely consistent with two or three innocent or less-culpable explanations, and the State cannot rule those out, then the State has not proven its single explanation to the standard the law demands. This is critical, and it is the opposite of what intuition suggests: the defense never had to prove what happened in that car. Nobody did. The burden was the State's, and the burden was not "make purpose look likely." It was "exclude what reasonably might have happened instead." Reasonable doubt lives precisely in what the State leaves unexcluded.

So look at what the State left unexcluded. Start with the single most dramatic piece of physical evidence in the case, the hard steering input. The data recorder captured what the State's own reconstruction expert called a hard yank of the wheel, 142 degrees, in the final seconds. The State needed that to be Mackenzie Shirilla aiming the car. But here is what the State's own expert, Mark Sargent, said on cross-examination, under oath. He admitted that the hard-right turn could be consistent with a driver attempting to maintain or regain control of the vehicle. He admitted it could be consistent with a passenger pulling on the wheel. He admitted it could have been an involuntary physical reaction caused by the car hitting the ground after going airborne. Three possibilities, from the State's own witness, and only one of them is the State's theory. The expert who was supposed to make the steering input mean intent instead testified that the steering input, by itself, does not tell you which of those things happened.

That is not a defense lawyer's spin. That is the State's expert, describing the limits of the State's most important physical evidence.

Now add a fact from outside the crash, because it bears directly on one of those three possibilities. The idea that a passenger might have grabbed the wheel is not something invented after the fact to muddy the case. Two weeks before the crash, on the July 17 highway incident, Mackenzie Shirilla contemporaneously claimed, in text messages to Dominic Russo's own mother, that Russo had grabbed her steering wheel. The piece is careful here, and the reader should be too. Russo's mother later told police she believed Shirilla had exaggerated. Russo himself, in texts afterward, apologized for the conflict while denying he had been trying to crash anything. So this piece is not telling you Russo grabbed the wheel on July 17, and it is certainly not telling you he did on July 31. Nobody can tell you that. What the July 17 messages establish is narrower and still meaningful: the question of who controlled the wheel between these two people was already, two weeks earlier, a live and disputed one. So when the State's expert says the 142-degree input is consistent with a passenger grabbing the wheel, that possibility is not exotic. It has a documented history. And the State did not exclude it.

Move to the second pillar of the State's case, the absence of braking. The data showed the brake was never touched. The State needed that silence to mean a conscious refusal to stop, a driver committed to the wall. But silence is not a confession. A brake that is never pressed is equally consistent with a driver who is unconscious. It is consistent with a driver in a panic freeze, a documented human response to sudden terror. It is consistent with a driver whose foot cannot reach the brake. The data recorder recorded that the brake was not pressed. It did not, and could not, record why. And why is the entire case. The State asked the factfinder to fill that silence with the single most culpable explanation available, and an honest look at the record shows the silence could hold several others.

Then there is the medical question, and here the record contains something the State's narrative tends to skip past. When paramedics reached Mackenzie Shirilla, her blood oxygen level was measured at 82. A normal level is 95 or higher. The paramedic who treated her acknowledged, on cross-examination, that 82 is extremely low, the kind of reading typically seen in or just after cardiac arrest. The State's medical witnesses testified they found no neurological condition, and the State leaned on that. But the same witnesses conceded, under questioning, the thing that matters: a person could have suffered a seizure or a similar medical event before the paramedics arrived, and it could have left no trace by the time she was examined. A trauma physician agreed that a seizure or a mini-stroke could resolve within the roughly two-hour window before Shirilla's hospital evaluation. Shirilla had a diagnosed condition, POTS, postural orthostatic tachycardia syndrome, that is among the more common causes of fainting seen in emergency rooms. This piece is not diagnosing her. It cannot, and a later section will explain why the failure to develop this question properly is itself part of the problem. The point here is narrower and it is a manifest-weight point: a medical event at the wheel was a live possibility the record gestures at and the State did not exclude.

And one more door, the quietest one. The interior of that car was not a controlled space with one actor in it. The appellate court's own recitation of the evidence records that the front passenger seat was reclined all the way back, with Dominic Russo in it and Davion Flanagan lying on top of him. Three young people, in a configuration the trial evidence never explained. This piece is not going to tell you how they came to be positioned that way, or build a story from it, because that would be reconstruction, and reconstruction is exactly the sin this piece accuses the State of. The honest point is only this: the State asked a judge to be certain, beyond a reasonable doubt, about what one person in that car intended, when the State's own evidence never resolved what was happening among the three people inside it. Certainty and an unexplained interior are difficult things to hold at the same time.

Step back and look at the pattern, because the pattern is the argument. The State's most dramatic physical evidence, the steering input, was given three innocent-or-neutral explanations by the State's own expert. The State's second pillar, the absent braking, is consistent with unconsciousness and panic, not only with resolve. The medical possibility was conceded as unexcludable by the State's own witnesses. The interior of the car was never explained. None of this proves Mackenzie Shirilla is innocent. This piece has been clear from the first page that it is not making that claim. What this pattern shows is something different and, for a conviction, more corrosive: the evidence the State built was genuinely consistent with more than one account of that morning, and the State did not exclude the others.

A reasonable factfinder is not supposed to resolve that kind of uncertainty by choosing the worst available explanation and calling it proof. When the evidence will support several stories, a conviction for the most serious of them, requiring the highest degree of certainty the law knows, is a conviction standing on ground that will not hold still. That is what a manifest weight challenge is built to identify. And it is why, set beside the keystone argument, this is the second real reason a careful person can look at this verdict and doubt it.

X. The Autopsy That Was Changed

There is a piece of this case that has nothing to do with the crash itself, and it deserves its own section, because of what it shows about how this prosecution was assembled. It concerns a document, and the document is the autopsy.

Recall a fact from earlier, one of the settled ones. When the Cuyahoga County Medical Examiner's office first examined the bodies of Dominic Russo and Davion Flanagan, it classified the manner of both deaths as accidental. Flanagan's report, dated September 2022, said accident. Russo's, dated October 2022, said accident. That was the official medical conclusion of the county's own forensic pathologists, reached after they did their work, and it stood for months.

Then it changed. In March 2023, the prosecutor's office sent the medical examiner a package of additional material. Not new medical findings. Not a second autopsy. The package was cell phone transcripts, still photographs, video, and witness interviews, investigative material the police had generated. Dr. Joseph Felo reviewed what the prosecutor sent him, and he amended both reports. The manner of death for Dominic Russo and Davion Flanagan was changed from accident to homicide. The amended reports stated that the investigation had revealed what they called the driver's intention to inflict harm on herself and on her passengers.

Sit with the shape of that for a moment, because the shape is the point. A forensic determination, the official medical classification of how two people died, was reversed. And it was reversed not because the medicine changed but because a prosecutor preparing a case sent over a folder of investigative documents and asked for reconsideration. The autopsy did not lead the investigation. The investigation rewrote the autopsy.

Now, there is a real legal question buried here, and Ohio law actually speaks to it. Ohio has a statute, R.C. 313.19, that governs the manner and mode of death recorded by a coroner. Under that statute, the coroner's verdict is treated as the legally accepted manner of death, and the statute describes how it gets changed: the common pleas court of the county, after a hearing, may direct the coroner to change the determination. A hearing. A court. That is the path the statute lays out.

That is not the path this case took. There was no court order directing the change. There was no hearing. There was a prosecutor's package and a pathologist's decision. And this is not a defense theory or a stretch. The Eighth District Court of Appeals, the court that affirmed Mackenzie Shirilla's conviction, looked at this directly and said so. In its own opinion, the appellate court stated flatly that the record was void of any evidence that the medical examiner complied with the statute. The court that upheld the conviction acknowledged, in writing, that the statutory procedure was not followed.

So why is Shirilla still convicted, if the court itself said the procedure was not followed? Because of how appellate review works, and this is worth understanding clearly, because it is the same lesson this piece keeps returning to. The defense had not raised this objection at trial. When an issue is not raised at trial, an appeals court reviews it only for what is called plain error, a demanding standard that asks whether the mistake was obvious and whether it changed the outcome. The Eighth District decided it would not find plain error. It reasoned that R.C. 313.19, as that court had read it in an earlier case, was really aimed at situations where some outside third party wants to force a coroner to change a ruling, not at a medical examiner revising his own determination. It concluded the amendment did not change the verdict, because the trial judge had heard the circumstances of the change and other evidence supported the convictions anyway.

A reader could argue with that reasoning, and reasonable lawyers would. But notice what the appellate court did at the end, because it is unusual and it is telling. Having declined to reverse, the court went out of its way to add a caution. It cautioned the State about the need to comply with R.C. 313.19. Courts do not spend words scolding the winning side for no reason. That caution is the sound of a court that affirmed a conviction while being genuinely uneasy about how a piece of the evidence underneath it was produced.

Here is why this section matters, and why it is not a technicality. This piece is not arguing that the autopsy amendment, by itself, should overturn the conviction. Under the plain error standard, it does not, and this piece is honest about that. The reason the changed autopsy belongs in this analysis is what it reveals about the direction the case was moving. A homicide case is supposed to be built from the evidence outward. The physical findings, the forensic determinations, the facts on the ground come first, and the charge follows from them. What the autopsy sequence shows is a case being built the other way. The deaths were officially accidents. Then a prosecution decided they were something else, gathered the material to support that conclusion, sent it to the medical examiner, and the official record of how these two young men died was revised to match. The medicine was made to follow the theory.

That is not, by itself, proof of anything improper about the verdict. But it is a genuine irregularity, flagged by the appellate court in its own words, and it fits a pattern this piece will return to in the sections ahead: a case in which the conclusion seems to have arrived before the proof did, and the proof was assembled afterward to fit. The autopsy is the cleanest, most documented example of it. It is the moment you can watch the case being shaped.

XI. The Defense That Was Never Mounted

Everything to this point has examined the State's case and the judge's verdict. This section turns to the third corner of the courtroom, the one that gets the least attention in coverage of this case and may matter the most. It examines the defense.

But to see what was wrong with the defense, you have to hold onto the word this entire piece started with. Purposely. From the first page, this analysis has had one organizing question: the State convicted Mackenzie Shirilla of purposeful murder, the top rung of the mental-state ladder, and the whole case turns on whether the evidence could carry that word. An earlier section showed why the word decides everything, the gap between purpose and recklessness being the gap between a tragedy and an execution. Another showed that the burden of proving it is not a sliding scale, that uncertainty about a mental state is supposed to cost the State the charge, not earn it a smaller one. Hold all of that here, because the defense Mackenzie Shirilla received cannot be understood without it.

The Sixth Amendment to the United States Constitution guarantees a criminal defendant the assistance of counsel, and the Supreme Court long ago made clear the guarantee means effective counsel, not merely a lawyer with a pulse standing at the table. The governing standard comes from Strickland v. Washington, and it has two parts. A defendant must show first that counsel's performance fell below an objective standard of reasonableness, and second that the deficient performance prejudiced the defense, meaning a reasonable probability the result would have been different. That is a demanding standard, applied with heavy deference to defense lawyers, because second-guessing trial strategy from the armchair of hindsight is easy and usually unfair. This piece will not pretend the standard is lower than it is.

But this piece is going to ask a hard question about the defense Mackenzie Shirilla received, and it is not the question people usually ask. The usual question is whether her lawyer should have hired experts. That matters, and we will reach it. But it sits downstream of a larger question about the basic shape of the defense itself. Before you ask whether a defense was equipped, you have to ask whether it was aimed at the right target.

There was a hierarchy of defenses available to Mackenzie Shirilla, and they were not equal.

The strongest defense available was the one this piece has spent the last several sections demonstrating. It was this: the State cannot prove what was in Mackenzie Shirilla's mind, because the evidence is genuinely consistent with several accounts of that morning and the State excluded none of them, and a conviction for purposeful murder cannot stand on a record like that. That defense does not require proving Shirilla did nothing wrong. It does not require a competing story about the crash. It requires only holding the State to the burden the State always carried, the burden of proving purpose, that one decisive word, beyond a reasonable doubt. Call it the no-reliable-conclusion defense. Its power is that it takes on nothing the defense did not have to take on. It simply insists that the contested word be proven.

The defense Mackenzie Shirilla's lawyer actually mounted was a different one. The defense argued, in substance, that this was reckless homicide rather than purposeful murder. Not "the State cannot establish what was in her mind." Instead, "the State can establish what was in her mind, and what was in there was recklessness, so convict her of the lesser offense rather than the greater."

Sit with the distance between those two, because it is the center of this section. The reckless-homicide defense, by its very nature, concedes the single most contested thing in the entire case. It concedes that a firm conclusion about Mackenzie Shirilla's mental state can be drawn from this evidence. It argues only about which conclusion. And the moment the defense itself tells the factfinder the evidence is good enough to support a definite finding about her mind, the defense has handed the court permission to make one. It has surrendered the contested ground. The judge was then free to agree a firm conclusion was available, decline the defense's preferred version of it, and adopt the State's. That is precisely what happened.

Measure what that surrendered. The keystone of this entire analysis, the argument that the factfinder's own speculation finding shows the inference of purpose could not bear weight, is an argument that the evidence cannot support a confident reading of Shirilla's mind. The manifest-weight argument, that the State left reasonable alternatives unexcluded, is an argument that the evidence cannot support a confident reading of Shirilla's mind. Both of the strongest arguments available in this case run through one point: this record cannot establish, to the certainty a criminal conviction demands, what Mackenzie Shirilla intended. A defense built on conceding that a confident reading is available, and merely bargaining over its content, gives that entire line away before the first witness is sworn. It does not lose the argument about purpose. It forecloses it.

And here is the fact that makes the choice harder to understand, a fact from the case's own history that this piece has not yet placed in front of you. Before this case ever reached Judge Russo, before the bench trial, it passed through juvenile court. Mackenzie Shirilla was seventeen, and Ohio law required a hearing to decide whether the case would move to adult court. At that hearing the State had to show probable cause, a low bar, merely a fair probability, that Shirilla had committed the offenses charged. The State charged her, among other things, with aggravated murder, the most serious homicide Ohio law recognizes, the one that requires prior calculation and design, genuine premeditation.

The juvenile court found the State failed to establish probable cause for aggravated murder. It found probable cause for murder and the other counts and transferred the case, but on the premeditation tier, the aggravated-murder tier, a court looked at the State's evidence and concluded the State had not cleared even the low probable-cause bar. Premeditation could not be shown.

Now set that beside the verdict. When Judge Russo convicted Mackenzie Shirilla, she did not describe a sudden, reckless catastrophe. She described a controlled, methodical, deliberate act. She said Shirilla had a mission and executed it with precision, that death was the ultimate goal. That is the vocabulary of premeditation. That is, in substance, the aggravated-murder theory, the prior-calculation-and-design theory, the very theory an earlier court had already examined and found the State could not support even at the probable-cause threshold. The conviction is for murder, not aggravated murder, so this is not a formal legal contradiction. But it is a revealing one. The case's own history contained a judicial finding that the premeditation account of this crash was not supported. And the trial then produced a verdict whose every adjective described premeditation.

That history is exactly why the no-reliable-conclusion defense was not some abstract possibility. It was the defense the case was begging for. A court had already drawn a line. And the defense that was mounted instead conceded that the factfinder could draw a firm conclusion about Shirilla's mind, the very move that let the trial sail past the line the juvenile court had drawn.

This connects to something this piece has been tracking since the autopsy. The coroner section showed a prosecution that appeared to build backward, a case in which the conclusion seemed to arrive before the proof and the proof was assembled to fit it. The deaths were officially accidents until a prosecution decided they were not. The defense-posture problem is the other half of that same picture. A criminal trial is supposed to be an adversarial test, one side building toward a conclusion, the other side contesting whether the evidence can support it. Here, one side built backward toward a predetermined outcome, and the other side, instead of making that the fight, conceded that an outcome could be concluded at all. The contested word in the middle, purpose, never received the adversarial test the system exists to provide. That is the deeper failure, and it is not a matter of hindsight. It is a matter of where the defense was aimed.

Now the experts, because the expert question is not separate from this. It is the reason the stronger defense could not have been stood up even if counsel had wanted to. The no-reliable-conclusion defense has to be built. It does not assert itself. It requires showing the factfinder, through qualified witnesses, that the State's confident reading of the physical evidence is not actually confident at all.

In a case like this one, that meant several experts. It meant an accident reconstruction expert of the defense's own, to stand against the State's reconstruction and place the competing readings of the data before the court as expert testimony rather than as a lawyer's questions on cross-examination. It meant a human factors expert, qualified to testify about pedal error, the documented and studied phenomenon in which a driver presses the accelerator while believing it to be the brake. That phenomenon has real limits, and an honest expert would have stated them plainly, but it is a genuine phenomenon, the subject of federal safety research, and in a case where the State's whole theory rested on an accelerator pressed to the floor, whether it could explain anything here is exactly the question an expert exists to answer. And it meant a medical expert, a neurologist, to address Shirilla's diagnosed POTS condition, the unexplained 82 percent blood-oxygen reading, and whether a loss of consciousness could have occurred at the wheel.

The trial record shows none of these experts was called. The defense retained no accident reconstruction expert, no human factors expert, and no medical expert. The most medically significant condition in the case, Shirilla's POTS, reached the courtroom only through the testimony of her mother, a lay witness with no medical credentials, an account the State was then able to undercut. The alternatives this piece described in earlier sections, the passenger grabbing the wheel, the medical event, the pedal error, were raised by defense counsel, but raised as suggestions during cross-examination and argument. They were never built into evidence. A suggestion a lawyer floats and an expert opinion the factfinder must weigh are not the same thing, and the difference is frequently the difference between a doubt that registers and a doubt that does not.

After the trial, in the postconviction petition, Shirilla's appellate lawyer did what trial counsel had not. He obtained an affidavit from a neurologist, Dr. Kamal Chemali, who reviewed the post-crash medical data, the elevated lactate, the low blood oxygen, the complete amnesia, the straight-line trajectory into a fixed object with no braking, and opined that the pattern was consistent with a loss of consciousness at the time of the crash, citing published research on the characteristics of seizure-related crashes. Set aside, for now, what happened to that petition; the next section takes that up. The point here is what the affidavit represents. It is a piece of the defense that could have existed at trial and did not. It is evidence of the shape of the defense that was never mounted.

Here the honesty this piece owes the reader cuts in more than one direction. The law does not treat every uncalled expert as a constitutional failure. Ohio courts, in cases like State v. Madrigal, have been clear that an ineffective assistance claim built on a missing expert is speculative unless the defense can show, concretely, what that expert would actually have said. A claim that counsel should have called someone is worth little without proof of what that someone would have testified to. The strength of this entire line of argument therefore depends on the substance of affidavits like Dr. Chemali's, and on what a reconstruction or human factors expert would actually say if retained. The ineffective assistance question is genuine, but it is only ever as strong as the expert record behind it, and this piece will not overstate it.

Two further choices deserve brief mention, and brief is the right length, because they are the weakest part of the ineffective assistance picture. Counsel advised Shirilla to waive a jury and try the case to a judge, and counsel did not have Shirilla testify. Instinct wants to question both. But courts treat the jury-waiver decision and the decision whether a defendant testifies as among the most strategic and most protected choices in all of criminal defense, and no court has come close to holding that advising a bench trial is, by itself, ineffective assistance. There may be a real conversation to have about whether a jury, seeing a seventeen-year-old who was herself nearly killed and who asked "How is Davion?" as she was pulled from the wreckage, might have weighed this case differently than a judge did. But that conversation is speculation, and it is not where this section's strength lies. The strength of this section is structural: the defense conceded the ground it should have fought on, and never built the expert record that fighting on it would have required.

One fact frames all of this without requiring anyone to read a defense lawyer's mind, and this piece will state it and let it sit. The attorney who tried this case, James McDonnell, is not an inexperienced lawyer. His own professional materials describe roughly three decades in the justice system and prior service as a prosecutor, work that means he understands, from the inside, exactly how the State assembles a case built on intent. His listed experience includes homicide and the juvenile transfer process this case ran through. That is the background. And the trial was two days long, with no defense experts, built on a theory that conceded the State could prove a definite mental state, in a case whose own juvenile-court history had already flagged premeditation as unprovable. This piece is not going to tell you why an experienced former prosecutor tried the case that way. It does not know, and neither does anyone who was not in those rooms, making those decisions, with that client. What this piece will do is set the experience beside the trial and let the reader hold both. An attorney with that background. A defense of that shape. The question that contrast raises is one a fair reader is entitled to ask, and entitled to answer in the privacy of their own judgment.

Whether what happened here meets the formal legal standard for ineffective assistance of counsel is a question for a court, and this piece does not pretend to deliver that verdict. What this piece can say is narrower and still serious. Of all the contestable things about the Shirilla conviction, the defense itself, its basic strategic shape and the expert record it never built, is the most genuinely contestable of all. And as the next section explains, it is also the one the system has had the hardest time ever actually examining.

XII. One Day

There is a date in this case that decided more than any piece of evidence did. Not the date of the crash. Not the date of the verdict. A filing date. October 24, 2024.

Why would a single day on a calendar carry that much weight? Start with what was supposed to happen after the trial. A direct appeal, the one the Eighth District decided in 2024, is limited. It can consider only what is already in the trial record. If the most serious problem with a conviction is something the trial record does not contain, and the problem this piece has spent a full section on, the defense that was never mounted, is exactly that kind of problem, a direct appeal cannot reach it. The vehicle built to reach it is a separate proceeding called a petition for postconviction relief. That is where a defendant brings in new material, the affidavits, the expert opinions, the evidence that should have been at trial and was not. For Mackenzie Shirilla, postconviction was not a formality. It was the one road on which the strongest claim she had could actually be heard.

Ohio puts a clock on that road. A petition for postconviction relief must be filed no later than 365 days after the date the trial transcript is filed in the direct appeal. Shirilla's trial transcript was docketed on October 24, 2023. That set the clock running.

And here is where a quiet trap was waiting. 2024 was a leap year. Most people, asked to find a date 365 days after October 24, 2023, would simply move forward one year on the calendar and land on October 24, 2024. In an ordinary year, that instinct is correct, because an ordinary year is 365 days long. But a leap year is 366 days long, and the stretch of calendar between October 2023 and October 2024 contained the extra day, February 29, 2024. That one inserted day pulls the math apart. Counting 365 actual days forward from October 24, 2023 does not reach the anniversary. It runs out a day short, on October 23, 2024. The Eighth District said it in plain terms: counting from October 24, 2023, the 365th day fell on October 23, 2024.

Shirilla's petition was filed on October 24, 2024. The calendar anniversary. The date the instinct reaches for. And, because of the leap day, the 366th day. One day late.

When her lawyers were later forced to defend that filing, they argued that the anniversary date should have controlled, that the deadline in a leap year should stretch the full 366 days to October 24. The Eighth District rejected it without hesitation. The statute, the court held, prescribes a deadline in days, not years. It says "three hundred sixty-five days," and 365 days is a literal count, leap year or not. The anniversary instinct is not the law. The day count is the law.

It is worth being clear about what kind of mistake this was, because it matters to how the rest of this story should sit with you. This was not a lawyer who let a deadline slide and filed late knowing it. The petition was filed on the calendar anniversary of the triggering date, October 24 to October 24, which is the date a 365-day deadline lands on in every year that is not a leap year. Whatever else is true, the filing was made on a date that, in an ordinary year, would have been exactly on time. The error was not indifference. It was a single hidden day on the calendar, the kind of thing a diligent person can miss, missed once, in the one year out of four where missing it is fatal.

And fatal is the right word, because here is the part that should stop a reader cold. That invisible, human-scale error did not cost Mackenzie Shirilla a procedural advantage or a strategic option. It cost her the entire claim. Permanently. With no review of whether the claim had merit. A mistake the size of a leap day produced a consequence the size of a life sentence left unexamined. When the punishment is that total and the error is that small and that understandable, the result does not read as justice administered. It reads as something that a system, if it were paying attention, would want to fix.

And one day was enough to end it. The trial court dismissed the petition, the Eighth District affirmed, and the reason they gave is the part worth slowing down for, because it is harsher than people expect. The court did not weigh the petition and find it wanting. It did not read the medical affidavit and decide the theory was thin. It never reached any of that. The court held that the 365-day deadline is jurisdictional. That word, jurisdictional, is the whole story. It means the deadline is not a guideline a court may stretch for good reason. It means that once the deadline passes, the court has no power to hear the case at all. A late petition is not a weak petition. To the court, it is an absence. There is nothing there to rule on.

Shirilla's lawyers also argued for what is called equitable tolling, a doctrine that lets a deadline bend when fairness genuinely demands it. The Eighth District said no, and was blunt about why. Equitable tolling does not apply to Ohio postconviction petitions. The deadline is set by statute, the legislature wrote only a few narrow exceptions into that statute, and Shirilla's situation fit none of them. The exceptions cover a petitioner who was genuinely prevented from discovering the facts of her claim, or a newly recognized constitutional right, or new DNA evidence of innocence. A petition that is simply one day late, for any reason, is not on that list.

Sit with the weight of that. The single most contestable feature of this entire conviction, the defense that was never built, the experts never called, the strongest ground never fought on, was set to be examined in the one proceeding designed to examine it. That examination never happened. Not because the claim was judged and found wanting, but because a document arrived twenty-four hours after a deadline. The merits were never reached. The door did not close on Mackenzie Shirilla's argument. It closed on the calendar.

Now, a reader who has followed this far will ask the obvious question. There is still the federal system. Every prisoner has heard of federal habeas corpus, the path by which a federal court can review a state conviction for a violation of the federal Constitution. Why is that not simply the answer here?

It is a path. It is a hard one, and an honest accounting has to say so. But it is a real path, and the way it would actually work is more specific, and more interesting, than the phrase "federal appeal" suggests.

Start with the rule that stands in the way, because you cannot understand the path without first understanding the wall. It comes from a 1991 Supreme Court case called Coleman v. Thompson. Coleman holds that when a state court refuses to hear a claim because the defendant broke a state procedural rule, a federal court generally will not hear that claim either. And Coleman goes further, in a way that bites directly here. It holds that the mistakes of a postconviction lawyer ordinarily do not excuse that default, because there is no constitutional right to a lawyer in postconviction proceedings in the first place. Read Coleman alone, and Mackenzie Shirilla's situation looks closed. Her claims were defaulted by her postconviction lawyers missing a deadline, and Coleman says, as a general rule, that her lawyers' mistake is her problem to carry.

But Coleman is not the end of the law. It is the beginning of it. Because in 2012, in a case called Martinez v. Ryan, the Supreme Court carved an exception directly into Coleman. The Court recognized something the older rule had ignored: that it is fundamentally unfair to bar a defendant's claim that her trial lawyer failed her, forever and without review, when the only reason that claim was never heard is that her postconviction lawyer also failed her. When state law forces a defendant to raise an ineffective-trial-counsel claim for the first time in a postconviction petition, and postconviction counsel botches that one opportunity, Martinez says that failure can count as cause to excuse the default. The door Coleman shut, Martinez propped back open, narrowly.

This matters enormously for Mackenzie Shirilla, and it is worth being precise about why, because the precision is the difference between a serious argument and a sloppy one. A federal habeas petition built on this would not ask a federal court to retry her case. It would not ask a federal judge to decide whether her trial counsel was ineffective. It would ask a narrower question: whether her defaulted ineffective-assistance claim is substantial enough that a rigid one-day jurisdictional bar cannot constitutionally be allowed to bury it without any merits review at all. The federal court's job, at that stage, is not to rule on the claim. It is to gauge the claim's weight. And the gravity of the underlying claim does real work there, because a federal court will not reach past a state procedural ruling to rescue a flimsy claim. It is the seriousness of what was lost, an entire expert-driven defense, in a case where the sole contested element was an inference of intent, that gives a federal court the room to act at all.

If a federal court agreed, the remedy would not be Mackenzie Shirilla walking free, and it would not be a federal judge declaring her trial lawyer ineffective. The remedy would be narrower and, in a way, more powerful. It would be an order returning the claim to Ohio to finally be heard on its merits, where the evidence, the affidavits, the expert opinions, could at last enter a state court record. And only then, after Ohio had ruled on a developed record, would the case be positioned to return to federal court if it needed to. That sequence matters, because there is a 2022 Supreme Court decision, Shinn v. Ramirez, that sharply limits a federal court's ability to develop new evidence itself. The route described here does not crash into Shinn. It goes around it, by having the merits record built where it is supposed to be built, in the state court, which is exactly what the law prefers.

Here honesty requires a hard caveat, and this piece will not soften it. Everything just described is an uphill argument. Martinez is a narrow exception, not a wide gateway. Federal courts have disagreed about precisely how far it reaches, and whether and how it applies to an Ohio case is genuinely unsettled. Federal habeas relief for state prisoners is rare, by deliberate design. This piece is not predicting that a federal court will step in. It is describing a path that exists, that is serious, and that is steep. A real argument and a long shot are not opposites. This is both.

But there is something a reader should understand about that path, and it is the part that matters most, because it corrects a mistake people make about the law all the time. People speak of the law as if it were a finished building, every wall in place, every door already either open or shut. It is not. Every case this piece has cited was once a case that did not exist. Coleman did not exist until 1991. Martinez did not exist until 2012, and on the day before it was decided, the exact argument it would adopt was, technically, not the law. Martinez is itself proof of the point. It is an exception that a court created because someone had the nerve to argue that an old rule, Coleman, was producing an unjust result, and a court agreed.

That is how the law actually moves. One court rules one way, another rules differently, the disagreement works its way upward, and eventually the Supreme Court settles it, until some later court finds the courage to revisit it and the cycle begins again. No door in the law is permanently shut until the Supreme Court says it will not hear the issue. What a case like Mackenzie Shirilla's needs, from here, is not a lawyer who accepts the wall as final. It needs a lawyer with the knowledge to see where Martinez might reasonably be argued to reach further, to a defendant who was a juvenile, who was incarcerated, who was wholly at the mercy of the very lawyers whose error defaulted her claim, and with the nerve to stand in a federal courtroom and argue that this is precisely the situation Martinez was meant to protect, even though no court has yet squarely said so.

I know what that kind of lawyer looks like, because I was raised by one. My father, Steven M. Askin, spent a career in West Virginia courtrooms refusing to treat the law as a finished building. He read every ruling. He knew where the walls were, and he knew that walls in the law are built by people and can be rebuilt by people, and he was not afraid to be the one who stood up and said the old way was wrong and argued for a door where others saw only brick. That kind of advocacy is rare, because it is hard and it is thankless and it usually loses before, occasionally, it changes everything. But it is the engine of every protection a defendant has. Every one of those protections was once a novel argument that a courageous lawyer made and a court had never heard before.

So where does the case of Mackenzie Shirilla actually stand, as of the spring of 2026. Her direct appeal failed. The Ohio Supreme Court declined to review it. Her postconviction petition was dismissed for being one day late, and the Eighth District affirmed that dismissal. Three separate challenges, three denials. Her legal team has continued to pursue avenues of relief, and outside observers are divided on her odds, with some experienced voices suggesting her realistic options are nearly spent and others pointing to the federal route still open in principle. The honest summary is that the road is narrow, it is steep, and whether anyone ever travels it successfully depends entirely on the skill and the courage of the lawyers who carry it.

Step back from the doctrine and look at what this section has really described. A young woman was convicted on a contested inference of intent. The defense that might have tested that inference was, this piece has argued, never properly built. The proceeding designed to examine that failure was closed before it began, by one day. And the federal backstop, while real, is hemmed in by doctrines that make the path genuinely hard. None of this is a conspiracy. Every piece of it, the jurisdictional deadline, the Coleman default rule, the limits on federal review, is ordinary law, working as written. And that is the uncomfortable lesson. The system did not fail here through malfunction. It produced this result through its rules functioning normally. A case can become nearly unreviewable not because anyone decided it should be, but because a series of neutral procedural rules, each sensible on its own, stacked up in front of one defendant until the merits could no longer be reached.

One day. That is what the rules, working exactly as designed, came down to. And whether that is the end of the story depends now on whether a lawyer with enough knowledge and enough courage decides it should not be.

XIII. The Verdict That Is Both Sound and Contestable

So what do we do with all of this.

A piece this long, that has pulled a conviction apart inference by inference, owes the reader a clear ending. Not a dramatic one. A clear one. And the honest ending of this case is not the ending people usually want, because it does not resolve into a clean verdict. It resolves into two things being true at the same time, and the reader has to be willing to hold both.

The first truth is that Mackenzie Shirilla's conviction is legally sustainable. This piece is not going to pretend otherwise. Under the deferential standards that govern appellate review, the standards laid out back in the early pages of this analysis, a reviewing court asks not whether a verdict was correct but whether a rational factfinder could have reached it. Measured by that question, this conviction survives. The Eighth District said so. The Ohio Supreme Court let that stand. There was a documented prior threat, a controlled turn, an accelerator held down, a brake never touched. A rational factfinder could assemble those into a finding of purpose. That is what the law asks, and by that measure the answer is yes. Anyone telling you this conviction is an obvious miscarriage that any honest court would erase is overselling. It is not that.

The second truth is that the conviction is genuinely contestable, and not in a trivial way. This is not a case where the doubt has to be manufactured. The doubt is structural, and it sits in the record itself. The factfinder convicted on an inference of purpose and then, addressing the same indivisible act, called a piece of that very inference speculation and ruled it irrelevant. The State's own expert gave three innocent or neutral explanations for the most dramatic physical evidence in the case. The medical question was never developed. The defense conceded the one thing most worth fighting. The autopsy was changed from accident to homicide after a prosecutor sent over a folder. And the one proceeding built to examine the weakest link, the defense itself, was closed by a single day on a calendar. None of that is nothing. Stacked together, it is a serious, sober case for doubt.

Here is the part that matters most, the part this whole piece has been built to deliver. Those two truths are not in contradiction. People assume that if a verdict is contestable it must be wrong, and if it is not wrong it must not be contestable. That assumption is itself the misunderstanding. A verdict can be genuinely debatable and still be a verdict the law will not disturb, and the reason is the deference this piece described in its early pages. The gap between a verdict that is correct and a verdict that is merely permitted is not an accident or a flaw. It is the deliberate architecture of appellate review. The law decided long ago that the factfinder who was in the room gets the close calls, and that an appeals court does not reverse simply because it would have decided differently. So the fact that this conviction is both standing and contestable is not the system malfunctioning. It is the system doing exactly what it was designed to do. The uncomfortable truth is that the design itself permits a debatable conviction to stand. That is not a glitch in the machine. That is the machine.

Which brings this piece to the thing it must say plainly, because it is the heart of why the case matters and it is the place where commentary on this case most often goes wrong.

There is a version of the defense story, repeated by people discussing this case, that holds that Mackenzie Shirilla actually received a capable defense. The argument goes that her trial lawyer raised the alternatives, the medical possibility, the wheel grab, the chance of an accident, through his questioning, that he cross-examined the State's witnesses and got real concessions out of them, and that this amounted to a real defense. It sounds reasonable. It is wrong, and the reason it is wrong is a principle every citizen should carry out of this case.

Cross-examination is not evidence. When a defense lawyer asks a pointed question and an expert concedes a point, that concession can expose a weakness in the State's case, and that has value. But it does not put an affirmative case into the record. It does not give the factfinder a qualified expert opinion to weigh. A lawyer's question is a lawyer's question. The factfinder cannot convict on a lawyer's questions, and a factfinder will rarely acquit on them either. Raising a possibility in cross-examination and proving it with evidence are different acts, and the difference is frequently the difference between a doubt that holds and a doubt that evaporates. To call what happened at this trial an effective defense because the lawyer asked good questions is to mistake the gesture of a defense for the substance of one.

And underneath that mistake is a deeper one, the deepest one in the whole case, and this piece will end on it because the American legal system is built on it.

When a trial ends and the evidence does not establish guilt beyond a reasonable doubt, the correct outcome is not a conviction on a lesser charge as a kind of compromise. It is not a conviction that splits the difference because something bad clearly happened and someone should answer for it. When the State has not proven its case to the required certainty, the correct outcome is an acquittal. Not guilty. Even when the defendant may well have done something wrong. Even when two young people are dead and a community wants an answer. That is not a loophole in the system. That is the system. The people who built this country's criminal law made a deliberate and difficult choice. They decided that it is worse to convict on insufficient proof than to let an unproven case fail, and they wrote that choice into the burden of proof itself. The burden does not move because the crime was terrible. It does not lower because the public is angry. It does not bend because the only survivor is the easiest person left to charge.

That is the lesson of the Shirilla case, and it is why this piece was worth writing about a conviction that will, in all likelihood, stand. The point was never to free Mackenzie Shirilla. This piece does not know what happened inside that car, and it has refused, from the first page, to pretend otherwise. The point was to show how a purposeful murder conviction gets built on an inference, how an honest look at that inference reveals it straining, how a defense can fail before it begins, and how a courthouse door can close on a single day. The point was to teach the machinery, because a public that understands the machinery is the only real check on it.

Two young men should be kept in view as this closes, because they are the reason any of this carries weight. Davion Flanagan was nineteen. His mother told the sentencing court that her son was the kind of person who would get up in the middle of the night to rescue a friend, no questions asked, and that he was so much more than the case sometimes made him out to be. Dominic Russo was twenty, and the family that loved him has carried his loss every day since. Davion Flanagan's family has said they long for the day the appeals are finally over, so they can begin to heal, and that wish deserves to be honored as the real and human thing it is. Nothing in this analysis is offered against them. They lost everything that morning. They were owed a careful, rigorous, fully tested accounting of how it happened.

The hardest question this case leaves behind is whether they got one. A trial that lasted two days. A defense with no experts. A conviction resting on an inference the factfinder herself would not follow all the way. A postconviction door shut by a calendar. The families of Dominic Russo and Davion Flanagan deserved a process beyond reproach, every bit as much as Mackenzie Shirilla did. The unfinished business of this case is that it is genuinely hard to say, hand on heart, that anyone got that.

A conviction can be sound enough to stand and troubled enough to keep you up at night. This is one of those. The work of a citizen is not to pick the comfortable half and forget the other. It is to hold both, and to keep watching, because the process only works as well as the public insists that it work.

That is the whole of it. Justice is a process. When the process is sound, it earns the result. When it is not, the result is just a verdict wearing the clothes of justice. Telling the difference is the entire job. It was my father's life's work, and it is the reason this piece exists.

— END ANALYSIS —