Matt Bruenig is a socialist lawyer who loves using AI. You might even say he’s the prototypical example of a lefty who takes the contrarian position on AI among those he’s allied with politically. Earlier this year, he told Transformer’s Dan Kagan-Kans, for a piece titled “The left is missing out on AI,” that he was worried about “the left-wing abstaining from using the tools when the right-wing does not.”
But it wasn’t abundantly clear to me at the time how it follows that leftist goals like unionizing workplaces, building a working-class movement, or getting leftist politicians elected were supposed to benefit from AI.
I still wouldn’t say it’s abundantly clear, but Bruenig has certainly made an intriguing YouTube video about putting Claude Code to leftist uses. He’s a policy wonk, so the video is not for anyone who might be squeamish about watching a 51-minute video about statistical weighting, but if you can get over that hump, I recommend it:
The background is that a recent Wall Street Journal article profiled young people moving in with their parents. It framed the idea not as the hardship it once was, but almost as a frugal no-brainer, and a strategy smart young people are employing to, perhaps, regroup following a breakup, or “save money for a starter home.” I heard some drive time DJs talking about this viral article on the radio earlier this month, and while part of it was about the undeniably true fact that housing is too expensive in America, the part about a cultural change just didn’t feel accurate.
But the marquee statistic sounded undeniable:
“Last year, 49% of adults under age 30 said they lived with a parent, up 12 percentage points from 2019, according to the Federal Reserve’s latest Survey of Household Economics and Decisionmaking.”
12 percentage points is kind of a breathtaking six-year jump, reflective not just of the Covid-era shock, but of a sudden and widespread change in attitude.
And according to Bruenig, apparently with the help of Claude Code, it’s simply a bad conclusion drawn from statistical crud in the Survey of Household Economics and Decisionmaking (SHED) survey data.
In the video, we see that Bruenig had already loaded up some folders (in Yazi) with survey data, and he uses Claude Code (armed with the Ghostty terminal emulator and running tmux to create panes in Ghostty). In a previous video he showed how he uses this setup to perform microdata analysis.
And while I’m not a stats guy, and I hesitate to just agree that the Journal’s conclusion is bunk based on one YouTube video, it’s impressive to see Bruenig prompt Claude by basically saying, verbally, that the stat looks a little fishy, and then have Claude Code go into bloodhound mode and quickly jump all over what seems to be some pretty clear bad math.
Allegedly, the illusion of a sudden 12% jump is caused by a “surveying artifact.” In one year’s survey, SHED’s 18-29 tranche has too few teens—the cohort likely to live with their parents—and in another survey the ages in that tranche are more balanced. On a graph, the result of sampling in this way would look like a big jump in the 18-29 group living with their parents, but it’s an outlier caused by statistical noise, and the real recent jump is probably around 3%.
This would probably mean there wasn’t actually some big attitude change, and that people probably just live with their parents because that’s what they’re stuck doing. Same as it ever was.
Using AI to apparently prove this point with minimal effort is cool. Does it mean the left should embrace AI wholeheartedly? That’s a debate for a different time. But, assuming the math can be independently validated, it’s an example of AI sniffing out the truth, and if you’re someone who longs to, say, leftistly prove the objective truth of the maxims of historical materialism, maybe it’s worth at least considering something like Bruenig’s AI tool as a helper. Or not. Don’t get mad at me.