eximius

The reasons against doing this is obvious. Is there any coherent description of why someone would think it is a good idea? And I don't mean, like, delusional reasons, but an actually rational argument for "here is where it might work out and we need to try and measure to find out".

Lot's of talk about how LLMs are bad at optimizing. And they are. They are still fundamentally semi-learned stochastic parrots. But LLMs are related to the Infinite Monkey Theorem - we've given monkeys typewriters, but pretrained the monkeys to be better than random.

They don't have to be good if they're cheap, directed, and sufficiently random to hit on something good within your budget.

Are LLMs better than our top mathematicians? No way! But they're able to spitball and falsify in parallel way faster. Eventually their directionally random output will hit something novel and interesting.

This is only a reasonable stance at the very surface level.

1. "You either work with what we use" - so whatever organization you represent isn't capable of evaluating and shifting to more secure technologies?

2. "it is mostly unacceptable for an enterprise product to have opinionated decisions about what authentication it works with" - you think companies that care about security should not care about integrating with flawed protocols?

A potential customer making bad choices does not obligate a business to make bad choices for their business.

I think the main thing I'm waiting for in the atproto space, for _most_ applications, is private data solutions. And the recent RFC on private data spaces (I forget the terminology) is a start, but has a ways to go.

* You're just describing arranging the test setup, which is independent of storage medium.

* Is your repository interface untyped?

* Depends on the fidelity of your fake, but generally I find FKs to be an antipattern these days.

* What are you checking primary keys FOR? Uniqueness is easy and I avoid more complex constraints and triggers.

* ... I'm beginning to suspect we have a difference in terminology. I'm not saying a mock. A typical DB fake would be array or hash table backed in memory. So an insert is "real" and an update or delete would be too.

* Well, sure, but I can get 95% fidelity for 1% of the resources.

I've yet to be convinced all of this effort is worth it, compared to the ease of using a repository pattern and just using a fake for tests.

And, like, I don't think we shouldn't be doing these efforts, I guess, as they may still pay technical advancement dividends down the road or help with cheaper, faster QA envs, all-in-one e2e envs, etc... but for the unit test and service test layers, those bottom several layers of your testing pyramid, fakes for your repository interfaces is so much easier and orders of magnitude cheaper.

While there is _some_ truth to what they are saying, it is an unnecessarily hostile and nihilistic take, intentionally eschewing nuance.

Often hot takes like this can serve a good purpose. I'm not sure that this does.

Alright, we're halfway through the Pantheon scene where David hacks the data center with the satellite phone.

Though, the phone moving with the vibration was always the more believable half. The ultrasonic pressing of the keyboard is mostly sound science but questionable from an engineering standpoint.

Hmm if a particle is a quantized packet of a field, then if multiple quantizations are possible in a field, then it's possible for more particles than fields?

Is that not kinda the point?

People reach for the things you mentioned wayyyyyy before they should.

Just because you want something queue shaped or search, doesn't mean you should reach for the big, specialized, expensive technology, when Postgres can already support it in your existing infrastructure up until some significant scale you often won't surpass.

Holy fuck, not everything in this life needs to be profit maximized.

I made Janus which is kinda like a less bloated Chezmoi

https://github.com/logannc/janus

1. This is kinda a dick move.

2. Regarding the title... you can definitely prompt them to be dumber, clearly. We know performance can be improved via prompts, from "baseline" performance. So this is a weird title.

If you ever suddenly get IPv6, it may become globally reputable without you realizing.

1) They reached out to me. 2) My answer was applicable and accurate, but insufficiently _unique_ for them.

Earlier this year I was told I failed an interview because when asked why I wanted to join a company, my answer "could apply to other companies in the same stage of life." They apparently required me to be _uniquely_ interested in their company. There were other oddities about their interview process.

Some interviewers just want to feel special.

I recommend installing Omarchy and playing with it. Read through the Omarchy repo to see how the things you liked and disliked were set up.

Then install arch or Cachy or some other Arch based distro and just use the parts you liked.

I am completely disinterested in the author's concerns because Omarchy was useful to me in demonstrating how to compose pieces into a useful whole in a relatively minimalist distribution like Arch. It is much easier to learn with good examples than just only reading DIY reference docs. (And they are good examples even if you dislike the targets of the apps installed - I agree on those points, but they're just not that important)

So, capabilities/type systems. Building code architecture guardrails steep enough the AI won't jump the fence/take shortcuts.

On both points, sure, trivially, if I do not value property or the rule of law, I will not care about destroying it. But obviously we're talking about the moral value of the implications more than the direct value of the camera. Do keep up.

On the second point specifically, that's actually much more interesting! If one values the rule of law, then you would actually want your laws to be morally aligned! Otherwise, the conflict of the law with morality _devalues_ the rule of law. Valuing rule of law does not imply only some sort of legalistic value of laws unto themselves, but of the value of a society with good laws, enforced well. This incentivizes analysis and evolution _of the law_ and, to some degree, forcing conflict to bring about those changes.

It is against the law, but I would wager it is morally coherent to smash them.

Presumably because you just put the styles in the component.

Those are fairly indistinguishable. It's when they start removing letters from words to save... debug symbol bytes or something? That's when c-style naming annoys me.

Nitor - a discord clone with a shared/federated Identity layer, but self-hosted "servers/guilds". Trust model is to trust the guild server (e.g., so private channels work as one would expect with moderation capabilities), but to enable E2E DMs and friend/presence systems via guild servers as relays. Rust, Iced, Iroh.

Glyphcraft - a Minecraft mod (imagine if Thaumcraft, Ars Nouveau, and Hex Casting were smashed together)

Looks like almost as good as JJ but with VC money.

Fuck no! Of course he can't be trusted. We know that. Nobody questions that. We know that about most of the "elites" running the show.

We're just in this shitty pit of despair where people are desperate. It's difficult to campaign for good when you're struggling and capital can jerk people around.

People pursue good for the sake of good at cost to themselves when times are very good or times are very, very bad.

Right now times are only merely very bad.

Mastodon, Bluesky, etc are neat - both in what they're trying to be and their technology. But ultimately these days I reject them in favor of more local socialization (again, not geographically). What this looks like is a constellation of private (or pseudo private) discord communities. If I make friends in one, I often get invited to another. I recognize the merit in broader social forums like Mastodon, but it is not worth the drawbacks to me.

As an aside, I'm not happy with Discord as a platform so I'm working on my own clone with some common identity stuff but with community servers run independently. That is, there are some "federated" identity providers so community servers can agree on identity across servers, then each community server runs its own thing. The trust model is based on the community server - private channels in a community server are not E2E encrypted, you must trust the server. But DMs and DM groups are E2E encrypted and use mutual community servers as relays (with a special class of relay server for people who want to DM but don't have an actual mutual server). I'm having fun with it. Now if only I could figure out why my video has such high latency (even locally!).

Let's be real, LinkedIn is full of LinkedIn Lunatics but pretty much all mainstream social media is pretty shit. They're just different flavors of shit. LinkedIn: bad. Facebook: bad. Twitter: I literally think it contributed to the collapse of discourse and rise of shallow thought / rejection of expertise. I'm not going to list more because the theme is, you guessed it, they're bad.

Google+ had promise in that the many problems of the other platforms could be curtailed with tooling to make your social experience effectively local (not necessarily geographically).

Yea, and it's a protocol problem more than anything. It would not be difficult to bolt on some kind of permissioning system to a PDS, but nothing else in the system would know how to handle that.

Eh, depending on how you're running agents, I'd be more worried about installing packages from AUR or other package ecosystems.

We've seen an increase in hijacked packages installing malware. Folks generally expect well known software to be safe to install. I trust that the claude code harness is safe and I'm reviewing all of the non-trivial commands it's running. So I think my claude usage is actually safer than my AUR installs.

Granted, if you're bypassing permissions and running dangerously, then... yea, you are basically just giving a keyboard to an idiot savant with the tendency to hallucinate.

That's kinda what I mean by the "very useful" part of my description. For all the flaws of the LLMs of today, in some ways context management - nominally a weakness - can be leveraged as a tool.