pfraze

If the frontier labs suddenly found themselves unable to compete with cheap open weight models running on widely available compute, then one might expect the frontier labs to be the only ones exposed to that risk, while a chip-manufacturer like nvidia could thrive in either environment. A partnership or commitment from the chip manufacturers to the frontier labs could change that. Whether that kind of inescapable connection exists is hard to predict without a lot of specific modeling, and at this point I'm inclined to think that neither chip nor datacenter demand is going to drop any time soon, and that the commitments would be unwound before a company like nvidia is threatened.

We made it a pretty big goal from the start to clearly communicate to users that the network is extremely public, and that we believe open access to public content is an important part of preventing another round of walled gardens. This is generally understood and appreciated, but the community has a wide range of opinions about it. Some people see public as public and that they’re there to have their voice heard. Some people see it as a pretty uneasy arrangement at best, and would rather it wasn’t that way.

The atproto community has been spending a large portion of the year developing “atproto spaces,” which are essentially a way to cheaply mint mini-atprotos with access control. This is the answer to non public data. It will be added to the Bluesky app and the ecosystem, and it will remain accessible to apps based on the grants of the users, but the spaces will not broadcast their content on the firehose. I believe, based on the reception we’ve received from that work, that people appreciate that it’s being done, and that people eager for less exposure will pick it up.

This should, hopefully, resolve any tensions at play. We will continue to advocate for public speech and I think many people will continue to participate in it, but I am quite curious to see if the non public spaces gain more adoption overall. My guess, based on people’s general use of the internet, is that they naturally will. The public arenas serve a particular purpose and not everyone is always trying to be a part of them.

EDIT: did I really just say "let me give an honest accounting of that." I have been spending way too much time with llms.

Ah right, of course. Fair enough. I think Dave Winer is trying some similar ideas, though I haven't looked at it beyond knowing he's trying to extend RSS to support these capabilities.

> As the next step of maturing governance of the PLC identity system, Bluesky Social PBC is supporting the creation of an independent organization to operate the directory.

Regarding his other concern, I would've loved to make atproto a localfirst protocol -- that's the field the atproto team came from -- but we had to pick & choose where to spend our complexity budget. I wouldn't say it's a nonstarter for the ecosystem; you can use something like iroh in conjuction with atproto to share identities -- dial people at their dids or domain names -- and then commit state to the atproto repo once it's ready to leave your local sync mesh. Maybe that sort of thing makes its way into future versions of the protocol. But, engineering often involves trades, and for the v1 that was one of the trades we had to make.

The policy around usage is shared in the rest of the post but the goal is to make this very simple for everyone