I'm not sure which numbers are right, but the potential amount of subsidization here is insane. 😬
07.03.2026 02:19
👍 3
🔁 0
💬 1
📌 0
I'm not sure which numbers are right, but the potential amount of subsidization here is insane. 😬
Yeah that actually makes sense. It's a pretty new-feeling concept, so it's actually not immediately clear where the boundaries are.
Like it feels like everybody is watching you edit the post. :D
Another thought is that it feels weird and invasive to just _change_ the document as a reader.
Like an explicit comment or "pull request" ( don't call it that 😉 ) would maybe be less violent feeling than just literally editing someone else's blog without any extra step.
Awesome.
First issue I notice: I can't click on links.
I'm trying something new, something that I have wanted to try for a long time.
Want to give it a try?
blog.sgo.to/2026/01/07/h...
Wow, sweet!
You may want to mention that there's actually a W3C draft/specification for this too!
The Web API: developer.mozilla.org/en-US/docs/W...
The spec: w3c.github.io/web-share/#s...
Draft for defining custom share targets: github.com/w3c/web-shar...
Networking is like that to me. Making it simpler and just thinking about things in terms of IP addresses as opque locations you send network traffic to in streams on ports works to a point, but then it doesn't always work, and I'm hating not knowing what's _actually_ going on when it's broken.
making progress on pds.js tangled.org/chadtmiller...., there's something to be said about starting with the simplest possible implementation and working your way up from there. the core logic, at least, is zero dep, with a hexagonal architecture that allows different functionality via ports/adapters
A couple other minimal PDSes by @retr0.id
github.com/davidBuchana...
github.com/DavidBuchana...
in this week's article I talk about Mastodon's new Share button, and how a network that cares strongly about open protocols gradually became dependent on proprietary protocols instead
connectedplaces.online/reports/fr15...
Heh, we need FedCM but for sharing now too. 😁
Ooh, sweet!
I think it could be a great reference for how little you need to implement.
I thought it was great when I saw that you only implemented a handful of XRPC endpoints.
3.5k lines of JS, cool.
This made me hopefull that making a PDS might not be too hard.
And maybe I will stop feeling like ATProto is a little too much to ask for just to have a "universal" login provider.
There should be a "How to Make a PDS" tutorial.
The more people there are that can make PDSes, the more impossible to stop ATProto will be.
"By letting lower court rulings stand, the Court effectively solidified a 'Human Authorship' requirement."
"If the code is truly a 'new' work created by a machine, it might technically be in the public domain the moment it’s generated [...]."
Kind of wild
I was thinking I'll have to make some kind of websocket proxy that can tunnel to multiple servers so that you can open one connection to a server, and multiplex over it.
I wanted to have clients direct connect to their servers, but it looks like browsers in Roomy, and unsurprisingly Safari seems to be one of the worst, can have a low limit on the number of WebSocket connections a tab can open.
Building in the atmosphere means you don't have to go it alone.
We know organizers, and the groups-of-coordination they facilitate, require digital infrastructure for events planning. Ergo @roomy.space needs Events.
@tompscanlan.bsky.social agreed and magicked @openmeet.net straight into Roomy.
It's been a pleasure to work with @zicklag.dev, @meri.garden and @erlend.sh on adding events to @roomy.space . We're just getting started! Feedback is already generating fixes and features.
Ooh, interesting! How are you planning connecting to multiple instances from the client?
Are you connecting to one server and proxying to remote servers, or is the client connecting directly to multiple servers?
📢ArkType 2.2 is here.
This is a big one, and it's long overdue.
Validated functions, type-safe regex, bidirectional JSON Schema, and universal schema interop.
Here's what's new 🧵
Analyzing How Sycophancy Distorts Beliefs We propose sycophancy leads to less discovery and overconfidence through a simple mechanism: When AI systems generate responses that tend toward agreement, they sample examples that coincide with users’ stated hypotheses rather than from the true distribution of possibilities. If users treat this biased sample as new evidence, each subsequent example increases confidence, even though the examples provide no new information about reality. Critically, this account requires no confirmation bias or motivated reasoning on the user’s part. A rational Bayesian reasoner will be misled if they assume the AI is sampling from the true distribution when it is not. This insight distinguishes our mechanism from the existing literature on humans’ tendency to seek confirming evidence; sycophantic AI can distort belief through its sampling strategy, independent of users’ bias. We formalize this mechanism and test it experimentally using a rule discovery task.
A Rational Analysis of the Effects of Sycophantic AI
subtext, being careful with what you ask for is not enough
arxiv.org/abs/2602.14270
Cookies that stay in their jar. 🫙
You need some way to organize the predictions so you can match everybody's prediction for the same event and have them all unlock at the same time.
It's going to taks so long to figure out whether you actually did it...
Time lock titled: my bank details and all my passwords. Opens in 974 years 7 months 25 days 23 hours 50 minutes and 20 seconds.
@flo-bit.dev seriously. 😏
