My side projects have spawned side projects which have spawned side projects.
If I built a super limited PDS that basically just handled auth and writing some lexicons would people be interested? Use case is services that want to offer "sign up" but don't want to administer a full pds
Sorry, I was trying to be terse. What I'm suggesting is that you could publish a public record to your PDS / relay that would give the address for a private data space with the call info. It's the same as your current setup but with most data in a private space.
I think you'd do this with 2 messages - 1 would be a permissions data one that only people you're calling could read. The other would be the call signal itself which could be broadcast.
If your IP address wasn't sensitive you could also just keep the broadcast and then handshake over webrtc.
Haven't you heard that that bluesky is dying? Best for them to stay away from the sinking ship, they already have enough on their plates.
A black and white Newfoundland Dog in a canoe, wearing a lifeguard life jacket vest on a pond. The background is out of focus but looks like a wooded area.
I heard we needed more dogs to get the likes up.
This one's older (8 years ago) and Amalfi's an old lady now, but she's still the best Newf.
I went through a process... It started with frustration that I would need to update my custom knot for it to work with @tangled.org it ended with me deciding to write a mobile app that is a PDS. Hopefully my Claude rate limit forces me to come to my senses.
I hope so
Currently contemplating building my own PDS... I think I need to go re-read the private data spec to see if it's a noble endeavor or a foolhearty one.
If I'm right there's basically one primitive missing from the PDS and I want to test out my theory.
Right now I just wrote my own knot that's backed by s3. But I'm thinking that there may just be an approach of either using a mix of records for commit metadata and blobs for packs or I may just have one custom server doing both.
I appreciate that you need a solution for private data right now but this is a pretty big turn off.
I run my own PDS right now, in order to use your product I would need to run a second PDS.
I'm already trying to find a way to have my PDS double as a knot.
Ultimately I think this is going to be part standard schema, part app view innovation. The primitives are in place we just need to build on it.
I think the next big thing for at proto will be reputation - probably more than just ratings
- Has to live on your PDS so you own it and it's portable.
- Cannot be something where you can delete or hide records of your mistakes.
- Cannot be something where people can review bomb you easily.
You inspired me!
I've been thinking about how we can do "verification" of at protocol records which I think is a critical piece to payments, especially when disputes come up. So here's this:
bsky.app/profile/chri...
Conceptually the more people who run the witness service and write witnessMe records to their own PDS we can build a distributed web of trust that a record existed in the past.
I'm running these 2 services here: github.com/chris-pardy/...
1 sits beside my pds and watches for "witnessMe" records from everyone and writes corresponding "witness" records to my pds.
The other service indexes those witness records and lets you look them up by the original record at:uri
It has been witnessed! - This is just raw JSON: witness.avaast.app/xrpc/app.ava...
WITNESS ME! As I turn a silly idea for at proto infrastructure into another side project.
I think that makes sense - if we had a "venmo on the at proto" it would solve payments for everyone.
Just thinking that (for now) the only option for payment is to build something on an app view for that.
@andyschwab.link have you thought about doing first party payments in the future? I'm not sure what cut Patreon and others generally take but I'd rather you get that percentage to support at.fund
I'm pretty sure all the research on human innovation points to us also basing our new ideas on prior art.
But even if your example was correct it's not how anyone serious is using these tools so it's a moot point.
I've been doing this job a while and needed novel algorithms like this never.
If anyone wants to do the VC backed thing, I won't, so here it is:
Drink from the firehose, build a database of profile similarity and record reach.
Sell access to the database to corporate social media teams.
Use LLMs extensively.
Consider frequent showers to wash off the Ick.
The fact that this is a community of people who feel the same as me, that they should invest their time and money to make something interesting for all of us is what I'm finding infectious about being here.
I have several "in progress" at proto app ideas.
- The one I'll use occasionally
- The one that will make some money
- The one I'll need VC backing for and will make me feel soulless
- The one I think will bring joy and cost me the most to operate.
Guess which one I'm going to spend time on?
I think there's a difference between writing where the written work is the craft vs software where the behavior is. We know the code itself isn't the craft because we have linter rules and unit tests, these tell us how to write and if it works. Code as craft would be good regardless of if it worked.
I think as a community we can learn from the Arts and Crafts movement. Don't eschew tools but embrace them in a way that benefits your craft, not the production of cheap low quality goods.
In part inspired by the @attie.ai discourse my thoughts on the need for Digital Joiners - the artisans who employ sophisticated tools to produce something that is still hand crafted.
It's time for a Digital Arts and Crafts movement building in public and showing the signature of the maker.
🫂
If you can solve the timestamp then you can generalize to everything, the issue is just observing. For instance a "verified" restaurant review just needs a witness that you were at the restaurant.
There's basically 2 approaches to trust.
1) Trusted third party - similar to how verification works on bluesky.
2) Trusted Network - basically your friends vouching for someone.
I would love to see some more exploration into building the network approach on atproto.
