> ## Content Index
> Fetch the complete content index at: https://www.marsorganic.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# “Programmers Lie.”
- URL: https://www.marsorganic.com/programmers-lie/
- Published: 2026-10-01T16:40:59.000Z
- Updated: 2026-10-01T16:56:49.000Z
- Author: Mike Tierney

“You’re a really good programmer,” I was once told—sincerely and, I might add, accurately—by a VP at a broadcasting monopoly that shall remain nameless. “Your problem is that you love music too much.” I knew what he meant and took his point. In the moment, I remember telling him something like: “Yeah, I’ve never seen my passion for music as anything but an asset, and I don’t think you can convince me it’s a liability now.” I liked the guy, and he was trying to be helpful. One of his own bosses was notorious for telling his team, as a principle of leadership: “Programmers lie.” Years later that strikes me, as it did then, as (A) too mean, (B) certainly true, and (C) no truer about us than about anyone who wakes up and looks at themselves in the mirror every morning. We lie to ourselves before armoring up for combat in Corporate America. Salespeople tell white lies to close deals. Senior leaders fudge to protect stock prices and their jobs. And more pertinent to my defense against my bosses at CBS (oops): numbers lie too.

The meta point for the suits wasn’t that programmers were dishonest people. It was that eventually we would start saying one thing and doing another for reasons of our own, making decisions subjectively based on our own instincts and tastes, which didn’t scale. Again, entirely true. Ownership could no longer listen to every station it owned, inspect every playlist, or trust the judgment of individual programmers. So it installed a new layer of suits—trusted former program directors who could be bosses for their respective regions, formats, and genres—and henceforth colluded in a tacit conspiracy to place their collective faith exclusively in the data. Programmers need not make taste. Radio would follow the tastes of the listeners, which they would do by proxy: they would follow the data.

It wasn't a bad plan. Of course we should give the listeners what they want. In Corporate America and elsewhere, it’s easier to defend yourself by saying you were following the data than to admit you were trusting your gut, especially when the shit goes south. The same VP at CBS also repeatedly told me—semi-admiringly and, again I must admit, accurately: “You’re a real piece of work.” He didn’t get why I would put my taste and my ass on the line by following my instincts rather than the research. And he was right. But I’m as sure now as I was back then that music programming—curation, as we call it now—was a function of synthesizing multiple inputs, some of which were not data (more on those later).

We all know data can be biased. It won’t surprise anyone to hear that in music, the primary bias trends towards familiar hits and against new music—one excellent use case for subjective taste as a counterweight. It’s human nature: “What I don’t know I fear, and what I fear I must kill (i.e. skip).” 

Curators also can be biased (in AI terms, *we* hallucinate *constantly*—much more on this later as well). It’s my further bias that—like me—most professional curators were listening to music and going to gigs when we should have been doing our homework, so we’re not very good at math. And since we aren’t sophisticated at parsing data, we tend to skim past any nuance and gravitate toward some bottom line data point as a definitive source of truth to determine whether a certain song is “working” or “not working.” In the streaming era, that holy grail stat basically has been “skip rate”—the percentage of times a listener starts a song and then taps “next” before getting through the first thirty seconds—which among other things means the, say, eighty percent of people who might like a song can find their listening dictated by the impulses of the other twenty percent who don’t like it or simply aren’t sure yet.

**Machines Don’t Go To Festivals**  
At any rate, you know who’s excellent at math? Machines, that’s who. Whatever else anyone wants to say about AI, machine learning and personalization now enable people like my old corporate overlords to make their rock ‘n’ roll dreams come true. No more lying programmers, subjective decisions, or unsophisticated parsing of the data. The algorithm will follow the data every single time, treat every customer like the one-off they are, do it better than humans like me ever could, and do it at total scale.

Why do we even need human curators anymore? I think we can all agree that the answer is *not* to be *kinda shitty* at math. It would be ironic if—as I predict self-interestedly—while the machines get better and more efficient at giving fans what they know and love after they already know and love it, the core skill set for human curators reverts to an emphasis on what makes us human. And what makes us uniquely human does not scale: ears, gut, taste, irrational quirks, messy biological drives and urges, connecting to other fans, seeing around corners, navigating by the windshield and not the rearview mirror. You know who can’t go to a festival? Machines, that’s who. As I tell curators all the time: stop worrying about AI and go to a gig.

So the machines have been becoming better and more efficient machines, during the same decades when the human curators have gotten worse at behaving like humans—because we’ve been incentivized to think like machines. The current mistrust of “the algorithm” is hard-earned but it might be misplaced. After all, we realize the algo’s job is just to give us stuff to stream and keep us streaming longer. Save your mistrust for humans who curate as if that’s their sole job too, who don’t go to festivals, who never change their minds, who act as if past performance always indicates future success. 

Trust curation that feels human, that takes chances and makes choices machines never could, and which puts things together that AI never would. Since you probably pay for streaming and skips are free, gravitate toward curation that respects your intelligence enough to give you stuff to skip. There might be an eighty percent chance you’ll like it. The human curation that survives increasingly will be that kind.

**Not For Everybody**  
“I fucks with Mike,” one of my direct reports used to say—accurately and, I have always assumed, affectionately—“but he is *not* for everybody.” Fair. Perhaps the nicest thing anyone has ever said about me. Radiohead are not for everybody. Raw sea urchin is not for everybody. Running marathons isn’t for everybody. And these are a few of my favorite things. I’ve never considered it my job to be for everybody, and it’s for shit sure not the job of my curation. Fancy talk about the monoculture aside, all curation just needs to be for *somebody*—maybe increasingly somebody who’s grown a little tired of the algo. 

I’m going to try to post a playlist every week right here. I’ll give context wherever necessary, but mostly it’ll just be a playlist. Some will be irrationally long; others will be intentionally short enough to burn on a CD. Sometimes the vibes will be obvious, and sometimes they might be obscure.

This week’s playlist “Winning Well” feels indirectly on-theme and should be self-explanatory—or at least I’m not going to overexplain it after a thousand words of this *mishegas*. It's an activity playlist "For leaving the drag field and the own goal scorers, and finding people who can run the plays." As always, my objectives are not to please everyone or be scalable, but to trust my gut, be kinda shitty at math, and put things together in ways that the algo probably never would.

Check it out if ya wanna.