Android fragmentation hasn’t gotten much better in recent years, but Google remains hopeful that its latest solution will actually cause a shift in the long-standing problem. It’s called Treble. By redesigning the framework behind Android, partners are able to take advantage of a modular-based system and easily transition between upgrades. Yet it hasn’t gained steady momentum.

You won’t hear about any worrying out in Mountain View, though. In a blog post, Google detailed how Treble has already made gains. Specific numbers are unknown, but there’s one milestone that Google expects it’ll achieve by the end of the year.

There will be more Android devices on Pie at the end of 2018 than there were Android devices on Oreo at the end of 2017, according to Google. It’s a bold prediction, no doubt. But it’s also one that everyone should be very hesitant to embrace. The older versions of Android are still alive and Pie is so far behind.

It’s particularly surprising for Google to make this prediction as Pie, which was released in late August, hasn’t even landed on 0.1% of all Android devices. Few companies are offering it on their products, and Pixel phones don’t sell in high volume.

Here’s a look at the Android distribution numbers from October 2018:

The other problem? Partners are continuing to ship new Android devices with Oreo. For whatever reason, they’re ignoring Pie and will likely do so until mid-to-late 2019. It’s the same cycle where the last version of Android finally gets a boost when a new version launches.

Oreo closed out last year by running on somewhere between 0.5% and 0.7% of the world’s Android devices. With less than two months left in 2018, it’s hard to imagine that Pie will leap ahead and meet Google’s expectation.

The good news is that Treble will be required on all Android devices with Pie or later. Does that mean software updates will arrive in a timely manner no matter the brand? Nope, but it should encourage companies to show more of an interest as the process isn’t as difficult and time-consuming.