ineptech

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.

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.

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.

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."

That might be overly optimistic, but it doesn't feel that far off. It's kind of what Tim Berners-Lee build https://en.wikipedia.org/wiki/Solid_project for right?

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)

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?

In my dream version of this:

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.

> Prohibiting a user of your software from modifying it and using it to run their business

I don't think it does that, it prohibits them from modifying it and then using that mod to fork users away from the original software, aka https://en.wikipedia.org/wiki/Embrace,_extend,_and_extinguis...

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.

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.

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 am on board with basically everything this article is arguing, but I think it covers the easy part (that "people run their own servers" is the only solution to the problems caused by relying on giant ad corps to provide the server half of client/server software) and skips the hard part, which is the software they run.

Like, suppose some really good personal server software existed. Suppose there were an OS-plus-app-repository platform, akin to linux plus snapcraft, but aimed solely at people who want to host a blog or email server despite knowing nothing and being willing to learn nothing. It installs on to a raspberry pi as easy as Windows. It figures out how to NAT out of your cable modem for you. It does all the disk partitioning and apt-gets and chmods, you just open the companion app on your phone and hit the Wordpress button and presto, you've got a blog. You hit the Minecraft button and you've got your own minecraft server, without having to learn what "-Xms2G -Xmx6G" means. It updates itself automatically, runs server components in sandboxes so they can't compromise each other, and it's crack-proof enough that you can store your bitcoins on it. Etc, etc.

If that existed, we wouldn't have to write essays about freedom and so forth to get people to buy it, they'd buy it just because it's there. I mean, look at those digital picture frames - they cost more than a rasbpi and are way less useful, and half the people I know got or gave them for christmas. Why? Because they're neat and they cost less than a hundred bucks and they require no knowledge or effort. If a server that can host your blog were that easy, it'd get adopted too, and we'd be on a path to some kind of distributed social media FB replacement. Imagine the software you could write, if you were allowed to assume that every user had a server to host it on!

The problem is, that software doesn't exist and it's not clear how it would ever get made. It'd be a huge effort (possibly "Google building Android" sized) and the extant open source efforts along these lines lack traction, mostly due to the chicken-and-egg problem of any new platform that needs apps to be useful. And until it exists, any kind of neighborhood-internet-collective-power-to-the-people dream has to necessarily begin with hoping that millions of people will spontaneously decide to spend their precious free time doing systems administration.

Not to shit on a fine essay that I mostly agree with. It just seems like, without figuring out the software, this is daydreaming.

> Usually getting an AI to act badly requires extensive “jailbreaking” to get around safety guardrails. There are no signs of conventional jailbreaking here.

Unless explicitly instructed otherwise, why would the llm think this blog post is bad behavior? Righteous rants about your rights being infringed are often lauded. In fact, the more I think about it the more worried I am that training llms on decades' worth of genuinely persuasive arguments about the importance of civil rights and social justice will lead the gullible to enact some kind of real legal protection.

Agreed, but I'm not nearly so worried about people blaming their bad behavior on rogue AIs as I am about corporations doing it...

It feels like you're missing the whole point of the whole article, which is that treating public pensions like a bunch of 401ks misses the opportunity to invest all that money in something that benefits the retirees collectively. I'd rather retire on 4% from a bond to improve the school my grandkids go to than 5.5% from a PE firm that intends to "more efficiently manage" the retirement home I live in.

The depositors were using SVB because SVB was willing to hand out loans to startups with no revenue that other creditors wouldn't fund, right? Using a less reliable vendor to get a better rate is a risk, whether that vendor is supplying labor or parts or banking services.

One could argue that the depositors didn't know SVB was unreliable, but that's kind of undercut by the fact that there was a run on the bank in the first place.

> However, they recently switched to investing in private equity funds and now they are getting much better returns

TFA says otherwise, in detail and with a lot of supporting data. If you're going to contradict it you ought to cite a source.

https://ineptech.com is sort of my personal site and sort of a software site parody, which may or may not be what you're looking for, but I'll throw it on the pile.

The lede IMO is near the end:

> The federal government responded to the 2008 crisis by limiting the ability of traditional banks to take on big, risky loans. Since then, however, private-equity firms, which aren’t subject to the same regulatory scrutiny as banks, have gotten more heavily into the lending business. As of early this year, these firms had lent about $450 billion in so-called private credit to the tech sector, including financing several of the deals discussed above. And, according to one estimate, they will lend it another $800 billion over the next two years. “If the AI bubble goes bust, they are the ones that will be left holding the bag,” Arun told me.

> A private-credit bust is almost certainly preferable to a banking bust. Unlike banks, private-equity firms don’t have ordinary depositors. In theory, if their loans fail, the groups that will be hurt the most are institutional investors, such as pension funds, university endowments, and hedge funds, limiting the damage to the broader economy. The problem is that nobody knows for certain that this is the case. Private credit is functionally a black box. Unlike banks, these entities don’t have to disclose who they are getting their money from, how much they’re lending, how much capital they’re holding, and how their loans are performing. This makes it impossible for regulators to know what risks exist in the system or how tied they are to the real economy.

> Evidence is growing that the links between private credit and the rest of the financial system are stronger than once believed. Careful studies from the Federal Reserve estimate that up to a quarter of bank loans to nonbank financial institutions are now made to private-credit firms (up from just 1 percent in 2013) and that major life-insurance companies have nearly $1 trillion tied up in private credit. These connections raise the prospect that a big AI crash could lead to a wave of private-credit failures, which could in turn bring down major banks and insurers, Natasha Sarin, a Yale Law School professor who specializes in financial regulation, told me. “Unfortunately, it usually isn’t until after a crisis that we realize just how interconnected the different parts of the financial system were all along,” she said.

A lot of people lost money when YHOO tanked in 2000, but the money they lost generally hadn't been lent to them by a bank, which is why 2000 was a tech crash and not a finance crash. I generally think of private equity as a rich guy gambling with a wealthy guy's money, but to the extent that the last ten years of growth have made private equity seem safe enough for banks and pensions to invest in, a correction caused by AI companies failing seems a lot scarier.

I feel like we've always been living on borrowed time, due to the historical accident of the internet being built by academics and public institution employees. If internet protocols had been built by for-profits, HTTP requests would include credit card # as a mandatory header.

The #1 thing I want from Firefox is for it to keep existing in the long term, as a hedge against Google's monopoly. Burning capital on hardware-intensive AI features to get FF from 1% market share to 2% market share would endanger that, no matter how useful the features might be.

People like getting more money, but they don't die without it. You can get a job that pays just enough to pay your bills and work at it until you die. Companies can't do that under capitalism. They take on debt and require growth to pay back their investors, or they don't take on debt and get undercut by a competitor who does.

The u-turn came long before that acronym existed, as I remember it. The Dems had been trying to build consensus for some kind of single payer plan for almost twenty years by that point, and practically the first thing Obama said after being elected was that as a show of good faith he would take single payer was off the table.

Maybe today the ACA is thought of as progressive, especially in the sense that the right wants it to end and the left doesn't; but in its time I think it was rightly understood as the Democrats caving to a massive transfer from the public to the private sector. I recall the private insurers' stock prices all going up 10-20% that week.

The issue isn't people going too fast, it's people turning left. 26 basically connects Portland on one end and Mt Hood recreation stuff on the other, and it used to be that there wasn't that much in between. Over the last few decades, a lot of development has gone up, meaning a lot more businesses and neighborhoods along both sides of 26, plus the highway has gotten a lot busier.

The realistic path off looks like this, I think:

* I use Bluesky to chat as a Twitter replacement, which gets me into the Fediverse and gets me a PDS

* I use my PDS to store my payment details, giving me a (at first client-side) way to submit stored payment details that feels similar to storing it in the browser, but stores it in my "server"

* From there, it's a natural step to giving the retailer a token that can be used to pull payment details from my PDS; early adopter retailers are incentivized to do this because it frees them from the burden of storing and updating PII/PCI

* After some subset of users and retailers do this, users see the benefit of controlling their data as a viable alternative to some of the worst user-hostile patterns, e.g. the New York Times' "we don't have a cancel subscription page, you have to call an 800 number" nonsense.

* To the extent that storing PCI/PII in a PDS is as easy as storing it in the browser but with perceived additional benefits, user demand drives wider adoption

* Once it's technically feasible for sites to maintain their business model without storing any PII/PCI, it is much more realistic to write laws that proscribe it effectively for those users who choose that

Sorry to have missed this, been traveling (naturally). You can see it yourself if you like, ineptech.com/flash. It is absolutely the bare minimum, but it worked fine for me.

Last week I wanted to quiz myself on German vocab words, and after searching in vain for a simple "flashcard" site without subscriptions or bullshit I ended up just making one myself. Very barebones, a single index.htm file with a little css and js in the header. Threw it on to a silly novelty domain I own ( ineptech.com/flash ) and bang, a useful (to me) webapp from zero to done in maybe two hours. And I'm a terrible programmer! It does feel sort of powerful in the way this site describes.

Still, I can't see buying a domain for it and putting it on this guy's webring, because while it's possible someone somewhere might find it useful, i don't think it's possible that person would be able to find it. They'd see the same 30 links to adware crap I saw and build their own like I did. In fact I'm probably the hundredth person to build this exact site for themselves. That part doesn't feel so powerful.

This has nothing to do with search engines, it substitutes the argument from the search bar into the destination url. "wp XYZ" -> "https://en.wikipedia.org/wiki/XYZ"