There are a ton — a ton — of startups vying for to be your music streaming service of choice. Beyond Pandora, Spotify, Rdio and MOG, they drop off in usage quickly, but suffice it to say there’s no shortage of smaller competitors. So many that it doesn’t really matter which one is the best, or which one came first, or which one has the most songs — it’s all about the marketing at this point. Whichever service gets the most users, and therefore, the most leverage with the labels, and the biggest audience for advertisers is going to win.

Breaking through the noise with an innovative service is difficult enough, but the real problem is that the likes of Spotify, MOG and Rdio have spoiled users into expecting on-demand streaming. Costly as those deals are, those are the table-stakes. It’s like what Zappos and free-shipping did to ecommerce. For any startup to pretend to matter it has to actually improve on the ability to play whatever we want, whenever we want, however many times we want, and in the order we dictate.

The problem is they can’t. Songza, 8tracks, Fuzz.com, and others rely on users to make playlists for other users to stream, because the standard licensing rules –the ones Pandora and everyone else are subject to — outlaw on-demand. That’s why you can play a Mumford and Sons channel on Pandora but only hear three of their songs per hour.

The only way to skirt this is with direct licensing deals with the labels, which is a near-impossible undertaking. Spotify’s Daniel Ek managed to pull it off by working tirelessly to win them over, at one point sleeping outside of an office building on the street for two months to get a meeting, persisting in his attempts for two-and-a-half years in order to get special deals with the labels (see a video of his describing the experience from our recent PandoMonthly below). “I used to have hair before I started this, and now I don’t,” he slyly joked of the harrowing experience.

Two years is long enough for most startups to run out of money. MOG and Rdio had to fight just as hard for their licensing deals. So did Turntable.fm, which it secured just in time to fall out of relevance.

Most of the music startups in the market know they can’t get these same deals. This is ultimately why Muxtape was shut down, why Imeem failed, why Mixwit, Seeqpod, Grooveshark, and Skeemr didn’t work out. And many more before them. So those currently ambitious enough to be in the market have to work around it, with user-generated playlists or, like Pandora, robot-generated playlists.

But both of those “solutions” have problems too. The problem with user-generated playlists is that users can’t even listen to their own playlists unless they already own the music. Only other people can. And even then, they aren’t allowed to see a whole playlist before it plays because it’s a “radio” license. If someone makes a playlist of music they don’t own and then listens to it, they’ve listened on-demand. Under RIAA agreements, that’s unauthorized.

This means all those mixtapes users toiled over are only for the good of the community. The only reward they get for building them is possible popularity on a niche music site. Not much of a motivator. This is one reason why Songza killed its user-generated playlist model and moved to expert-curated playlists.

Playground.fm is new music startup that tries to solve the on-demand problem with smarter robot-generated playlists. I wish them the best of luck because, as history has shown, music is the most difficult category of startup to build. (If it doesn’t work out, have you guys heard of enterprise? It’s cool again!)

Started by Stanford grad and music composer Vivek Agrawal, the app quietly launched earlier this year and has been iterating since. Playground.fm’s algorithms take the music you listen to on Spotify, Rdio, your iTunes, etc, and distills it into playlists, taking the work out of creating a playlist, particularly one you can’t even listen to. It’s basically a better, smarter Pandora that you don’t have to train, and it’s mobile-first. Eventually the company might do a subset of curated playlists a la Songza. The idea is to make finding the right playlist one touch away, and to integrate with the music you already own.

“When I’m on the phone, the action point is, I’m about to go on a run, I need to find something in the next 10 seconds,” said Agrawal. It may be more of a feature than a ground-breaking revolution, but its technology is an improvement on the current feast of radio-playlist hybrids.

Meanwhile, the music industry is in the middle of a messy blogging heads back and forth right now over whether unprofitable and barely profitable Spotify and Pandora pay artists enough. (Note: I welcome any thoughts on this topic. Email me.) Congrats to those two — they’re finally big and meaningful enough to look like the big bad enemy. Meanwhile the startups trying to offer something demanding music fans actually want can barely afford to make it work.