The decentralized nature of the internet is going to become very obviously important again.
10.03.2026 14:04
👍 18
🔁 3
💬 4
📌 0
The decentralized nature of the internet is going to become very obviously important again.
This puts me at 10 people released. I filed my 1st habeas petition 29 days ago. I am an incredibly small-time, nobody lawyer who knew nothing about immigration law the day I submitted that case. I barely know any more today. I had never sued the federal government and now I've beaten them 10 times.
Congrats! One thing I'd like to see: post more. A CEO needs to use their product a lot to have a feel for what the vibes really are and what works and what doesn't and what's important.
I mean, there is Plague Inc.
Crossing over Technology sector, worker flows, '000 12-month moving average From Europe to US From US to Europe TTT 2023 Source: Revelio Labs 24 25 3 2.0 1.8 1.6 1.4 1.2
More tech workers now moving from the US to Europe than in the other direction.
Wow. Hardware?
😢
Lol, .at might be the new .io and really stimulate Austria's economy 🤣
It's just missing that usual happy background music that tech product demos always have!
Tell me about it 🔥 🦅
why atproto?
this is why
Sadly, I come to the same conclusion every time I deal with a knockoff brand. Either the bricks don't snap together easily enough, don't stick well enough, or stick too hard.
💯
Except for the user data that doesn't make sense to store in a PDS. Ephemeral data, private data (but will be possible once permissioned data is figured out), or data you don't want to allow a user to change (think online game scores).
I think that's a great idea! Just didn't expect it in this case.
Any opinion on this new t3.codes thing by @t3.gg? I guess in addition to Codex, it'll support Claude Code, opencode, and other scaffolds in the backend, but manage them in the background.
Hah. This made me notice for the first time that chat messages on @stream.place are not ephemeral, but stored in my PDS. Good to know 😬
Oh wow! Damnit, Nvidia. Ok thanks!
Two Amazon product pages open side-by-side. * ASUS Ascent GX10 for 3.299,99€ * GMKtec EVO-X2 AI Mini-PC AMD Ryzen AI Max+ 395 for 2.799,96€
Ah interesting, thanks. Hadn't seen the GX10 yet. It's all a bit more in Germany, but still roughly true. Here's what Amazon shows me for the two. So maybe still could make sense to go the Nvidia route if it works better as well.
For Prometheus 3.10.0, I built a new PromQL feature: you can now specify default values for missing series in binary operations!
Thanks to @dash0.com for inspiring and sponsoring the development of this feature.
Here's how it works and what you have to watch out for:
promlabs.com/blog/2026/03...
Looking at www.reddit.com/r/LocalLLaMA..., DGX Spark is supposed to be faster for video generation, but Ryzen seems good for regular LLMs. I only care about using text-based models, am on Linux and happy to experiment a bit, so it sounds like Ryzen is a good choice.
Comparing
- DGX Spark (128GB) @ $4,699
- Ryzen AI Max+ 395 (128GB / 96GB VRAM-allocatable) @ e.g. $2,699 (Beelink GTR9 Pro)
...wouldn't I even get more bang for the buck with the Ryzen, or are there big downsides like speed or other noticeable things? Also happy to give AMD a try :)
Reminder that northern canada is extremely desolate but also filled with genuinely absurd fantasy terrain
I'm always happy when the beats start
The sun is woke
I pray for DHH for this reason 🤣
Don't get me wrong, I use LLMs all the time, but going from natural language to code feels like a different type of abstraction than going from high-level code to assembly, due to the high degree of indeterminism (literally random numbers involved) and the lack of assurances about the final output.
Looooool the Bari Weiss ending
I think it's fair to be critical of LLMs, but I can't remember the last time I saw Bluesky breaking down, and I use it daily. Maybe it's different for you.
Beyond that, I just feel like you are confirming @knotbin.com's point of definitely blaming LLMs without providing any corroborating details.
