As longtime readers know, I’ve never seen Donald Trump as an even remotely appealing or minimally qualified candidate for president, but over the course of the past week, he’s struck me as more pathetic than usual. Last Thursday, in Houston, Ted Cruz and Marco Rubio combined to make him look pitiable. On Tuesday, he gave a noticeably grim-faced victory speech after the Super Tuesday primaries, although at that point in the evening he had only lost two, Texas and Oklahoma, to Cruz. (By the end of the night, Cruz had also emerged victorious in Alaska, and Rubio had posted his first win, in Minnesota.) And last night, in Detroit, Trump offered a debate performance that was somehow even more woeful than his previous public appearances.

As usual, Trump lied with abandoned, weaseled when confronted, yelled when challenged, interrupted on impulse, complained that he was being treated unfairly, threatened various forms of retribution, and demonstrated his ignorance of policy, economics, and the United States Constitution. This time, though, he also conceded his own weakness, which I can’t remember him having done in the past. “Just, for the record, I have won 10. He has won three or four,” said Trump, referring to Cruz, proceeding to reference a number of polls in such detail that it was obvious he knows exactly how many times he’s been beaten. When Rubio asked him whether he would commit to stop producing his clothing line in Mexico and China, Trump declared himself helpless: “They make it impossible for clothing-makers in this country to do clothing in this country.” When Megyn Kelly presented him with evidence that he had not, in fact, been a staunch opponent of the Iraq War, Trump explained that he had been intimidated by Howard Stern: “I said very meekly, long before we went in, I said very meekly, well, maybe, maybe, I don’t know.”

On the eve of Super Tuesday, polls suggested that Trump was the likely winner in every state other than Texas. The fact that he ultimately lost three other states as well suggests that the Houston debate actually did him some damage. If so, last night’s thrashing can’t have been good for his campaign, especially since it was countered by a strong performance by Cruz, the only candidate with a chance of overtaking him.

Even so, Trump remains the clear frontrunner for the Republican nomination, and so I’d like to address a matter of grave importance to the nation and the world: There’s a very serious risk that this manifestly ill-tempered adult, if nominated, will win the general election.

Most Democrats would disagree with that, as would most Republicans. I understand their reasoning, and I hope they’re right. But I see several flaws in the argument for calm, and I’d rather be unduly alarmist than say nothing until it’s too late.

I’ll begin by noting that the conventional wisdom, that Trump would be a poor candidate in the general election, is backed by some of the available evidence and a number of wholly reasonable premises about contemporary American politics. Trump continues to have a significant “favorability” gap, although it’s less yawing than it once was. General-election matchups suggest that he would be a weaker candidate, against either Hillary Clinton or Bernie Sanders, than either Ted Cruz or Marco Rubio.

And—in addition to the fact that Trump has yet to secure the nomination—his footing among Republicans is less impressive than he would like. His margin through the first fifteen primary contests isn’t as commanding as the one Mitt Romney or John McCain had posted at this point in 2012 and 2008 respectively, and polls show that he has yet to elicit the support of a majority of Republican voters. A number of conservative voices and elected officials have publicly said that they will never support Trump, and polls show that about a third of likely Republican voters can’t see themselves voting for him; the dissatisfaction is so widespread that Chris Wallace’s final question, in last night’s debate, referenced the #NeverTrump movement: “Can you definitively say you will support the Republican nominee, even if that nominee is Donald J. Trump?” Perhaps even more telling than the question itself was the dismay from many Republicans, and the derision of many Democrats, after all four candidates–Trump, Rubio, Cruz, and Ohio governor John Kasich–answered in the affirmative.

That intraparty anhedonia would put Trump at a relative disadvantage if he makes it to the general election. The Democrats are likely to have a nominee with flaws of her own—Clinton is a polarizing figure whose long career in public life has left her with no shortage of baggage—and the Democratic primary has had some contentious moments. Still, there’s been no occasion to clarify whether Clinton and Sanders would ultimately support the other as the party’s nominee, and I haven’t heard anyone express indecision about the hypothetical choice between Clinton and Trump.

Plus, although Democrats have yet to turn their focus to defeating Republicans in the general election–thus far they seem to have concluded, not unreasonably, that Republicans can handle that themselves–it’s safe to assume that Democrats will actively oppose the GOP nominee. I also take some comfort in a possibility that Ilya Gerner suggested to me the other day on Twitter, that the Democratic Party is more functional, as a party, than the national GOP. As a journalist based in Texas–this is going to sound more snarky than I intend it to–I’m really poorly equipped to assess the Democratic Party’s institutional strengths and weaknesses, but the Republican Party could hardly have handled its Trump problem less competently. And although there’s plenty of blame to go around, at this point, the Republican party wouldn’t be on the cusp of nominating Trump if not for its own pre-existing pathologies; if that wasn’t true, then I wouldn’t have had occasion to raise the alarm about how serious the GOP’s Trump problem was back in August, much less to describe its nature.

All of that’s reassuring, because we’ve known since the Iowa caucus that Trump isn’t invincible, and his travails over the past week suggest that he might have been defeated months ago had the media not been waiting for him to self-destruct, and had his rivals not been counting on someone else to tackle him.

At the same time, it’s not just Republicans who have been slow to see this Trump train hurtling down the tracks. I know that because over the past six months I’ve offered a series of Trump-related predictions that were, in retrospect, well ahead of the curve. I knew in November that Cruz would beat Trump in Iowa, and on the eve of the caucus I predicted that Trump would underperform: that’s exactly what happened, and most observers were surprised. I warned in September that Republicans needed to take a different approach to Trump, and added that Rubio would be well equipped to puncture his self-aggrandizing mythos; Rubio did so brilliantly in Houston, and again last night, and had he done so several months ago, he might now be among the two candidates left standing in a far saner two-man race. In some cases, the national commentariat has seemed willfully obtuse about this whole debacle of a primary. I’m pointing this out because I really don’t think Trump would be a good president.

And so I hope Democrats reading this will, at least, consider the following concerns seriously before proceeding on the assumption that beating Trump in the general election would be a breeze.

First, although Trump’s favorability ratings and the current poll matchups are a valid proxy for public opinion at this moment in time, we’ve already seen that public opinion on Trump is mutable. It’s March. Eight months ago a majority of Republicans had an unfavorable opinion of Trump. The general election is in November, eight months from now.

Second, we’ve already seen that Trump himself is mutable. Obviously, he’s identifying as a Republican these days. But he has historically affiliated with the Democrats, more or less, and his Democratic tendencies still exist. They can be discerned in his comments about Planned Parenthood and Obamacare’s individual mandate, or the fact that he’s campaigning on strongly worded, if poorly specified, support for protectionist policies more commonly associated with the left than the right. (That explains one of the odder epiphenomena of the 2016 cycle: voters, including former Minnesota governor Jesse Ventura, who are torn between Trump and Sanders.) Further, since launching the campaign, Trump has already begun to pivot to the center on some issues, with illegal immigration being the obvious example. Last night, for example, he flip-flopped on H1B visas during the debate, then issued a statement flip-flopping the flip-flop; I’m actually not sure where he ended up on the issue, or where he started, because I don’t think it matters, because none of his stated views can be considered binding.

Trump has also begun to draw distinctions between himself and the presumptive Democratic nominee. During the debate prior to the South Carolina primary, for example, Trump proclaimed himself a visionary who had opposed the Iraq War from the beginning. As Kelly pointed out last night, the claim doesn’t check out. But it’s true that Trump didn’t vote for the war as a member of Congress. And if he’ll accuse George W. Bush of lying us into war on the eve of the South Carolina Republican primary, he’ll do so in the general. Clinton’s support for the Iraq War was a liability in the 2008 Democratic primary, and it’s been an issue in this year’s primary too: it could easily become an issue if she faces Trump in the general.

Democrats might be skeptical, on the basis that regardless of whether he was ambivalent about the Iraq War, Trump has taken plenty of positions that are anathema to Democratic voters. But complacency is ill-advised, because we already know some Democrats support Trump. And that brings us to the biggest problem with the assumption that Trump can’t win the general election: we already know that the electorate is mutable. The first fifteen primaries have shown that Trump’s appeal isn’t confined to a bloc of Republican voters delineated in the traditional ways, with reference to characteristics like region, ideology, socioeconomic status, race, or religious affiliation. His support is strongest among some subsets of the Republican electorate, like blue-collar white men, but he nonetheless performs pretty well among college-educated professionals too. He’s been successful among voters who would have logical reasons to reject his candidacy, such as southern evangelicals. He’s even been successful at animating support from the most unlikely subset of all: voters who don’t usually bother to vote.

To put it differently: Trump has created a new coalition, one that doesn’t map onto the traditional categories. Who are Trump’s supporters? I’ve given a lot of thought to that question, and it will be the subject of my next post, because they are, as yet, being mischaracterized by most observers, and clarity on the subject is crucial. Trump supporters may not be a silent majority of the electorate as we know it today, and their commitment to Trump isn’t inviolable: Cruz has proved that much, and he still has a chance to prevail. If not, Trump will become the nominee. If Democrats can get a clear picture of his supporters, they’ll be able to counter his message effectively. If not, Trump supporters will end up picking the next president eight months from now.