I think it originated in 1971, when OR governor Tom McCall famously said, "Come visit us again and again. This is a state of excitement. But for heaven's sake, don't come here to live." It wasn't meant as a joke, he was speaking in earnest. One of his ongoing themes was opposing growth for growth's sake and he helped pass the strict land-use policies that largely still exist. It has become a joke though, you will see "Welcome to Oregon, now go home" bumperstickers and such. And a quick way to get dirty looks at a party is to announce that you just moved here from California and can't believe how cheap the houses are.
Errant thought - I kind of hope that the actual social network for developers of the not so distant future looks more like a bunch of us separately building bespoke applications to serve files to each other.
> Dave: "Bot, add a folder to my personal data store with the pictures I took this weekend and share it with everyone in my poker group"
> Sally: "I see that I've gotten an email from Dave with a magic link conveying access to something. Bot, use this link to negotiate a key exchange with whatever weird app Dave cobbled together (ask his bot if you need help) and then log into my VPS and add a tab to my personal webapp to display the content he has shared."
Think of it this way - why didn't we all just replace Facebook with "host files on our personal sites and give each other access to them" ten years ago? Because we'd have to store passwords, identify each other past spoofed emails, figure out when/how to trust friends of friends, etc. In practice that means everyone adopting the same protocol, which is a chicken-and-egg problem.
That hasn't changed, but what has changed is that reading docs and writing the glue code is effectively free. The main problem with something like Solid is, if it takes even an hour to figure out how to get it going, I am effectively assigning my friends an hour of homework to be able to network with me. That's no longer true. If you like the Fediverse and I like Solid and Dave uses Nostr and Sally is stuck with some proprietary thing that came with her NAS, we can all build a network without any of us switching to, or even learning much about, each other's platforms.
It’s actually changed for the worse. Now everyone wants their slop to be the winning protocol, so the competing standards problem has become far broader
I think that a million people asking Claude to build something is going to tend to converge on one protocol, not diverge. But the point is, I don't think it matters that much if they diverge. The whole idea of a standardization is predicated on the assumption that bespoke software is expensive, and that is now not always true.
I wouldn’t have agreed before I learned software more deeply, now I especially agree in the age of better and better AI.
Put simply there are solutions that are elegant and irreducibly correct. I think we got away from this because software expertise got so fragmented by frameworks and packages but AI can cut through all of that.
I thought this is why we established standards. I truly hope the LLM singularity brings further adherence to existing standards and doesn't lead to an explosion in protocols and formats. That sounds like a nightmare! :)
I follow the same approach (finding LLMs useful but only running open models on my own hardware) but I've been limited by also not letting them into my email and being pretty restrictive about their internet access. Would you mind sharing what stack and agent harness you use, since your other comments make it clear you're highly restrictive about what kinds of tech you run at home and I assume you vetted it thoroughly, and it sounds like it's working well?
I'm wondering specifically about things like what kind of tool or restrictions you use on web browsing to avoid being flagged as a bot and how you guardrail access to your email (to avoid, say, the LLM deciding that "help me organize" includes telling people you want to reschedule). I use Qwen 3.8 26b at home and find it very useful for some tasks but when I think of using it as a general assistant rather than coding assistant, I keep thinking of times it has gone wildly wrong (e.g. it once spent 20 minutes trying to get around Supabase authentication because it couldn't figure why a certain call wasn't working)
My go to right now is jcode, and doing a lot of one-shot experiments. Like I have a session being my MPD DJ as my local spotify right now.
That said for persistent things like interfacing with homeassistant etc I am playing with zeroclaw.
I run the same model as you for coding. For more sane automation tasks i recommend doing everything possible with MPCs so you can tell it to only use existing tools. ymmv
By no means is my setup well oiled yet. Different vms and systems doing different things. As a QubesOS user jailing an agent for one job to one VM is easy, though I am working on an enclave native operating system that should make this easier for most people.
Hi, I think I might be your target audience, I currently run pi on the server in my basement, in a minimalist docker container that gives it access to my code workspace and a config dir. Give me an idea what benefit I get on top of that by using the self-hosted pi pod?
For you the main benefit would probably be portability via the mobile app. The mobile app is not released yet but you can build it yourself from the swift/kotlin source in the github repository.
Otherwise, the main benefit here is for people to be able to clearly define all of their pi and agent config at the org, user, and project level. This most directly benefits teams of people working together, but I also find it helps me as a single user working on multiple projects on multiple devices.
I have a similar project you might be interested in if you use ACP via Zed or something; I need to make a Show HN post at some point.
The gist is that it pretends to be an ACP agent to Zed, but it actually spins up a Docker container on your PC and proxies the ACP connection over websocket to the agent. It supports bind-mounting your Pi config like you're doing (as well as copying files, so changes don't propagate to your host). It also has early plugin support for your ACP connection. It's basically an nginx proxy for ACP, I just made a plugin that will automatically kill a session if it detects OpenRouter secrets in agent output, tool calls, file access or shell commands.
It doesn't currently support remote targets like your server, but it's on the list.
Another notable difference is that it creates a new container every time you connect so you get an idempotent environment, though I am also working on a persistent style. Everything supports it, I just need to make the host-side binary support finding a container to use before it tries starting a new one.
I am developing tarvis.io that allows you to self host apps without managing the servers. In addition to that it also comes with cloud dev env where you can do coding and spawn dev servers and you can host same app in your tarvis workspace as a self hosted app. Tarvis takes care of security, storage, backups and ssl domains. Each app auto gets ssl domain and only you can access your apps. Give it a try!
I am running deepseek harness as self hosted app on tarvis for myself. You can run pi as self hosted harness there too by telling the agent to set it up and then access on the go from the browser.
a) the timer would be cumulative, so that solving the early words faster gives more time for the harder words,
b) going negative wouldn't end the game, it would just turn the time red or something, with the goal being to finish all of the words with the highest time-remaining possible, rather than just to win/lose.
Might as well add my own experience since I just set up a local llm this week. I went with a 32GB card made by Intel called Arc B70, which is cheaper than a 3090 and more has ram, at the cost of a slower memory bus. edited to remove something incorrect, thanks diablod3
I went with this because a) the models I wanted to use are a little too big to fit comfortably in 24gb, plus I need room for a few additional small models for autocomplete and speech recognition, and b) I already had a cheap server to use and dual gpus would've required upgrading the mobo and power supply and probably the case as well.
It was definitely a little tricky to set up. The Intel line requires a driver package called "level zero" to support something called SYCL (Intel's version of CUDA basically, AFAICT) that was tricky to get working. I am running llama.cpp in a docker container, which also required some fiddling to get the container to see the card. You also need a kernel from the last few months.
Once I got it working though, the results are very impressive for a $1k investment. Qwen 3.6 35B at q4 quantization takes about 3/4 of the ram and delivers like 88 tokens/sec. So, if you want a decent-sized model for cheap, this is one way to go.
Whoops thanks, was going from memory. At any rate, the effect is that it's somewhat slower than the 3090, when using a model small enough to fit entirely in nvram, but can fit models the 3090 can't.
Yep, but the flip side is you can get two B70s for the price of a single 3090 (MSRP, obviously; 3090s used go for about the same as a new B70), and they're true 2 slot, so they can fit on x8/x8 consumer boards fine.
The side effect is Intel has fired their entire Arc team, including the driver team, as well have canceled all Celestial products (only low end combined GPU + IO die products will have Celestial, as its too late in production to change). All future Intel products will have Nvidia graphics tiles.
Llama.cpp is, of course, still trying to better support the existing Intel products, but its kinda hard when they might stop working any day due to driver breakage.
HN front page circa September through December of last year. Its part of the $5B investment into Nvidia, and the new CEO, Lip-Bu Tan, happily did whatever Nvidia told him to do.
Nova Lake (Series 4) is already too late in development to switch, so its shipping with Celestial, and Titan Lake (Series 5) after it will be essentially a Nova Lake refresh using a hybrid tile that is Celestial's graphics engine with Druid's media, memory, and bus controllers (since Druid development has been halted).
Serpent Lake (Series 6) will be the first that can ship with Nvidia tiles, which should launch in early 2028.
Unless something changes, there will be no DGPU Celestial products.
Unfortunately, the ink on the contract is already dry, so this is the end of Intel's GPU unit.
Serpent Lake (and its siblings) cover both consumer desktop and laptop.
Arc's team was disproportionately effected by the layoffs. A lot of the major engineers that worked there changed their Linked in details to list Nvidia, AMD, etc.
As for DC and Workstation GPUs, they also seem to be canceled.
Always a pleasure to hear capital explain to labor why taxing capital is bad, but this seems like a giant red herring. I don't want a wealth tax so I can cut my income tax, I want a wealth tax to address inequality. Our existing policies have produced a very bad bad outcome - wealth inequality exceeding that of pre-Industrial England led by a small, essentially randomly-selected group of people so wealthy that they effectively run everything who have entirely captured a corrupt government and are very close to making the situation permanent - and a wealth tax is the only policy idea I know of with any chance of changing that.
How can they ever see a dollar of profit without a rug pull, license change or hosted moat? This is a neat idea - besides just replacing github, a network of loosely-federated git servers seems like a promising base for distributed social media or chat platform someday - but it seems like the only way it can really stay open is if you're planning to stiff your investors.
It also doubles as a way to verify that someone is a real person using their real identity, which is starting to become pretty important these days. If Alice and Bob are both on this platform, the confidence Alice can have in the proposition "the Bob account is really controlled by a guy named Bob who really knows some people I know, as opposed to being AI or an overseas scammer" would be roughly proportional to the strength of the friend network connecting them. That sounds useful.
I’m not convinced that’s the case. A relatively small subset of bad actors can join the network, create new accounts on a second phone, tap (or find a way to fake that process via the API), then eventually use those accounts from bots.
It’s of course more friction, which in itself is good to avoid spam/bots, but over time all of that can very likely be automated
There's a German gay social/dating app called Romeo that has a feature where you can show which people you know personally. There's no physical validation though, so it's easy to fake.