kortex

The younger devs have largely been the ones showing us old farts (eg millennials) how do to the really sophisticated stuff with claude - custom skills, plugins, tooling like openspec, things that have had massive benefits over stock claude.

We are nowhere near the ceiling in terms of process either.

Re: flat rate going to by-usage, I believe this is largely a long tail problem. You have a small number of power users that capitalize on the flat rate to use the service orders of magnitude more than the average user.

I think my company is a microcosm of the current state. The non-engineering side (HR, correspondence, marketing) are on the "forced adoption" side, giving out gift cards to folks using Glean the most.

In engineering, we can't raise token budgets fast enough. Devs are "routing around damage" when they hit caps, going from claude to opencode to copilot. Productivity is up (roughly) 100-300% in terms of story points and 75-200% in lines of code. And defect rate is down [0], more bugs are caught in review before QA or prod. Our teams are just starting to figure out our new workflows too, for design -> spec -> code -> review, it'll only get better as we refine the process.

It's looking like software industries will reap massive benefits, while most others which have some error tolerance will only see modest gains. It's unclear how it will impact high accuracy fields like legal (it might even be net negative).

Also which is it - a useless technology that has to be force-fed because it sucks, or a economy-shaking game changer that will put folks out of jobs en masse? Those seem like a contradiction.

0 - i think process here is extremely important. I think it would be very easy to create an unmaintainable slopocaplypse. We have an informal platform team of about three (including myself) that have been affectionately and informally dubbed the Tech Priests of Mechanicus Adeptus (warhammer reference) that ensure the prompts/skills and associated tooling are optimal, that code standards are enforced, and that solutions are converging at the system-wide level.

zionism != judaism

That's literally what OpenSpec does (https://openspec.dev/). It's quite nice. I've only exceptionally rarely seen claude do something wrong based on spec docs when it's fully spec'd out. More often it's because something wasn't nailed down and claude was forced to make assumptions.

The downside is the ospx markdown specs sometimes end up too granular, focusing on the wrong or less important details, so reading the specs feels like a slog.

Also at times aspects of the english language spec end up way more verbose than just giving a code example would be.

I agree with voidfunc. A lot of what police do could be offloaded to other occupations. A lot of needless deaths could be prevented if there were more rungs on the escalation ladder between "do nothing" and "folks with guns show up". Like the same vibe as firefighters and EMS but just like for mild social disruption.

You might wanna check out https://copier.readthedocs.io/en/stable/

Dunno if it's your exact use case but it's been amazing for keeping a polyrepo microservice ecosystem in sync.

I'm not in front of my terminal, but I'm almost certain there is a way to do this. And if not, it would not be hard to add.

I can't really take the article fully seriously when they are like "uv cant do this. Well actually it can but you gotta use an extra flag." It reads rather PEBKAC.

Thought provoking, but i dont understand what making an incompatible protocol would do. As soon as you have a machine-processable protocol (kind of a tautology), you can build bridge layers.

Sounds like an XY problem. What op wants is high barrier of entry and high curation. That exists already, but by virtue of high barrier of entry, they are hard to find :p. Discords, matrix, webrings, weird little websites.

This isn't an academic journal or general purpose news article though. It's a small blog about a subculture. It'd be like spelling out GPU on a blog for building custom PCs, or BDSM on a kink blog.

Spelling it out when 99.9% of your audience doesn't need it actually is the opposite of in-group signaling, it makes it feel like it's aimed at a wider audience, when it's not.

https://en.wikipedia.org/wiki/Cooperative_principle

NASA just dropped 12,000 photos from the Artemis II mission. But which ones are the best ones?

I know Jeff Geerling linked the site a few days ago, but Hank just unveiled the vote feature in response to the big photo drop to sift through.

This should be immediately intuitive to anyone who spends more than 5 minutes looking at firewall traffic of something public. 99.9% of the bots' requests aren't doing sophisticated penetration attacks, they are blasting all the low hanging fruit: the common ports, the common wordpress endpoints, the common bobby tables style sql injections and xss attacks.

Eh, the problem with that reasoning is one of extreme degree. The "obscurity metric" would be the surprisal associated with discovering the critical piece of info. Using a random port confers brute force resistance of 2^16. At 1ms that's about a minute. Brute forcing at the same rate a 128 bit key takes like 10^28 years.

It's like hiding your key under the mat, vs hanging on a tree limb of a specific tree only you know the gps coordinate of. Both are "obscure". Huge difference in difficulty.

And AI agents reliably-ish do tasks billions of times a day that humans struggle with, namely regurgitating information at incredible rates across wide breadths of topics. I see it as merely a matter of degree, not category.

How do you measure "deeper understanding" in humans? You usually do it by asking them to show their work, show how the dots connect. Reasoning models are getting there, and when they do, I'm sure the goalposts will move yet again.

We don't have a way of measuring "cognitive awareness" though. We have a way of measuring electrical impulses, and how they behave in response to various treatments (eg anaesthetics or magnetic fields), but we can't objectively measure whether the system is aware at all.

We can measure electrical spikes, and we can ask the system to reply what it experiences when various spikes occur. Guess what: we can do that with ANNs now too.

It'd be one thing if this were all a philosophical discussion, but in this thread so many folks are making very firm statements about the nature of reality we have no means to back up.

Ya, I seem to largely agree with your comments on this article. I was replying to brookst, did you mean to reply on a differnt thread?

Is someone tripped out on mushrooms experience ego death and total disruption of sense of self still conscious? They may even contend they are more conscious than normal life, what with all the communing with the universe and whatnot.

Trees react to the world around them in many ways.

Humans can't reliably subitize more than five-ish objects, while chimps can actually do this task better than us. That's our "cant count the R's in strawberry" (which flagship models can reliably do now, general letter counting).

https://en.wikipedia.org/wiki/Subitizing

They stopped being autocomplete years ago with RLHF

How do you know the sensation of a red photon hitting a cone cell, transduced to the optic nerve through ion junctions and processed by pyramidal neurons, is any more or less real than the excitation of electrons in a doped silicon junction activating the latent space of the "red" thought vector? Cause we are made of meat?

I'll bite that bullet. In fact I contend the idea that "humans and maybe some animals are conscious, but other things are not" is the special pleading stand. Why are the oscillating fundamental fields over here (brains) special, but the oscillations over there (computers, oceans, rocks) not? If they are, where do you draw the line? It smacks of "babies dont feel pain" (widely believed until the 80s! the 1980s!) sort of reasoning.

https://en.wikipedia.org/wiki/Panpsychism

They need to make an example out of these companies. If your whole business model is built around handling sensitive data, and you are caught shipping off that data to brokers, you should be liquidated or at least fined to within an inch of bankruptcy, as basically all of your profits are a sham.

I actually dont think it's that obvious at all (unless you are a senior engineer). It's like the classic joke:

A QA engineer walks into a bar and orders a beer. She orders 2 beers.

She orders 0 beers.

She orders -1 beers.

She orders a lizard.

She orders a NULLPTR.

She tries to leave without paying.

Satisfied, she declares the bar ready for business. The first customer comes in an orders a beer. They finish their drink, and then ask where the bathroom is.

The bar explodes.

It's usually not obvious when starting to write an API just how malformed the data could be. It's kind of a subconscious bias to sort of assume that the input is going to be well-formed, or at least malformed in predictable ways.

I think the cure for this is another "law"/maxim: "Parse, don't validate." The first step in handling external input is try to squeeze it into as strict of a structure with as many invariants as possible, and failing to do so, return an error.

It's not about perfection, but it is predictable.

I go by a philosophy that Liskov Substitution is reeeally about referential transparency. I don't care about parent/child classes, I care about interfaces and implementations, and structural subtyping. Fix that, and it's great.

Carbon + helium fusion is rather favorable, vs carbon production by the triple alpha process (3He), so it's just reaction kinetics essentially.

I don't think it's a "weird fetish." It's just most of the things privacy advocates have been warning about - PRISM, warrantless metadata requests, tech companies handing over data - are all largely invisible.

A camera pointing at your child's playground or gymnastics class is much more salient.

Super fun! But the text does get in the way mid-combo.

Complete hearsay, but I struck up a convo with someone who had spent a few hours drinking around a campfire with him and a few others at burning man, prior to GPT3's popularity. Apparently he was utterly convinced in his pivotal role to shepherd in a new era with AI, to the point where it got really messianic and culty. He didnt recall much else other than just being really weirded out by the dude.

Not an EE, but if transformers are at all like smaller scale power supplies, the issue with using multiple smaller components is it works right up until it doesn't. If you lose one or it gets overloaded, it puts more strain the rest, increasing the failure risk. Then another one pops, and the load shifts to the smaller pool, in a cascade failure.

To an extent, you can do this, as long as you have systems in place to shed load and prevent the components from failing in quick succession by circuit breaking.

Also i believe transformers are much more graceful handling overcurrent than silicon. But everything has its limits.

for one, it's one tool, that does the job of all three.

it just works. i'm not sure how else to describe it other than less faffing about. it just does the right thing, every time. there's a tiny learning curve (mostly unlearning bad or redundant habits), but once you know how to wield it, it's a one stop shop.

and as mentioned, it's crazy fast.

Usually lack of knowledge that such a thing exists, or just plain ol' momentum. Changing something long in production at established companies, even if there is a tangible benefit, can be a real challenge.