Hi everyone! Welcome to the last issue of OpenML.fyi.

TLDR: I'm not going away, but I'm trying to switch my learning mode from writing to doing. That doesn't mean I can shut up, so follow the blog and Bluesky/Mastodon, but this particular journey is done.

You can be forgiven for having thought that it was already done, since I haven't emailed in roughly forever. But I haven't gone anywhere, and the pace of change at the intersections of open and ML have only accelerated.

If anything, I've become more engaged–which is ironically why I need to write this.

In case you forgot: what was this all about anyway?

At the beginning of the newsletter, the question was: what does "openish" mean? Three years later, I still don't have a clean answer, but no one else really is either.

On the licensing front, OSI has offered an answer and LF another. "open weights" are a practical reality even if it doesn't feel very satisfactory to anyone except as a marketing gimmick. (If anything the current argument is about whether licensing is even real right now.)

Outside of licensing, and much more importantly, the situation is even more unsettled. I wrote early on about all the many, many non-licensing factors I thought about when I thought about open. All of them are still very live topics. They're also still very messy topics. Power and autonomy have always been messy topics, of course, so of course I didn't expect writing to answer them–but I didn't expect the problem to get messier over time, with the good outcomes possibly even better than I realized and the bad outcomes very bad indeed.

Why wrap it up now?

So if this all still matters, why wrap the newsletter up?

I started realizing I needed to call it a day about a year and a half ago. OSI was working to close out their definition of "open" for AI, and Meta was pushing hard to claim the word for themselves. It was the biggest definitional fight around "open" in 20+ years–exactly the kind of thing younger me would have thrown myself into.

I didn't really engage.

Part of that was practical: I was in the middle of selling Tidelift and simply didn't have the bandwidth to truly sink my teeth into it.

But I also didn't feel comfortable that I had a useful answer. Since before I went to law school (but not always!) my mechanism for change has been to write: to express myself, to help others understand, and to give advice where I could. But around open in AI, writing hadn't clarified that for me. The questions got ever sharper but the answers did not.

Why no answers: two hypotheses

There's two reasons why I think I've been struggling to reach clarity.

The first is that we're in the middle of a paradigm shift. I've said from the beginning that we're in the middle of a paradigm shift, so this shouldn't be surprising! And yet... I was not ready for how destabilizing it would be. Surely, having seen it coming, I would be ready!

Nope. Even things that a few years ago I would have taken as utter gospel, like "Wikipedia is the most important thing humanity has ever built together", are somewhat in question: Wikipedia may soon be important in the same way newspapers are, which is to say both critically important and essentially collapsed. Ditto, I've always taken human collaboration around the commons as something that happens very naturally. And... right now that is actually hard in some domains, not just because of slop but also because it's so seductively easy to be a loner. (Says, uh, a loner with a newsletter.)

Which leads to the second hypothesis: my core views were formed by being active in communities — IRC, mailing lists, arguing in Bugzilla. Getting my hands dirty is what made me a decent thinker and more importantly a good human. (Turns out the praxis people were right!) And the newsletter... was great but was not praxis.

Still important, more important than we could have known

As Ethan wrote a while back, the stakes of the current fight could not possibly be higher, because the winners could literally own the lens we see our entire world through. And in ways I could not have foreseen when the newsletter launched, that fight needs all of us.

So why give up on the newsletter, if these ideas still matter? Newslettering is one form of critical discourse; it isn't a bad thing. But the discourse that actually shapes norms isn't happening here, it is happening in the places where people shape the facts on the ground--in other words, where people are doing and building things.

To put it another way: I think I can probably do more for open by making some open things really good, and letting that speak for itself. Linus didn't choose GPL because of impassioned arguments from RMS, he did it because gcc and emacs were impassioned objects that made RMS's case for him.

And there are so many needs for good humans Doing The Work right now. So that's what I'm going to try to do.

So what's next?

I am of course going to keep working with Creative Commons, OpenET, and other groups doing the important work on freeing knowledge, particularly climate-related knowledge. If you have problems in those spaces, drop me a note!

I'm also getting my hands dirty as a Wikipedian again. So far that's been in digging up some bad news but I'm eager also to start experimenting. It's critical for humanity that we have not just Wikipedia as it is today, but a profusion of anti-Grokipedias: sources of knowledge that seek truth, while being aware that this quest is fraught and hard and fractal and never-ending. That is aided by machines, but ultimately deeply human.

And of course the day job at Sonar continues–I think there's very much a role for mixing "traditional" tools into all of our processes. (Whether you want to call that "making them robust" or "keeping an eye on them" is up to you.)

Thanks!

Thank you to every single person who thanked me for any piece here, or eagerly subscribed--you gave the energy to get past issues out, and you're the ones I'm writing this last issue for as well.

... and see you on the interwebs

Like I said, I'm not going too far–lu.is (feed, yes, feed) may become even more active, and I'm very much (too much) on Bluesky these days. Have a group chat? Can't hurt to invite me–I probably won't say yes to all of them but eager to get down on that level with folks too.