jaggederest

https://www.youtube.com/@MoritzKlein0 has revolutionized my understanding of audio signal processing in software. Understanding what the voltages do is absolutely amazing.

LLMs, natively, are better at programming than nearly every professional software engineer, full stop. There are a lot of meta tasks (not programming, software engineering) that they are worse at, but the actual programming part, they've transcended human potential.

You have to treat them like an ignorant junior engineer in nearly every respect, but trying to match with them at programming per se is a fool's errand.

The reason to read the code is not because you can catch their errors in programming. It's because they may have missed the forest for the trees at the operational or strategic level, and so that you can learn from them to understand better. It's like trying to beat the engine in chess now.

> (sure maybe there are some psychological effects that can occur between the players, but no strategical impact by the actions of one player on the other).

You'll see a golfer perform at least a few strokes different depending on their partner, and at a pro level that's the difference between "in the money earning an extravagant living" and "washed out on day one, thinking about taking a different job"

Not only that, but a lot of these dams are not even in use any more, so doing additional work doesn't make much sense. Remove the dam, remove the problem, instead of duct tape solutions.

Then you remove them and move on with your life? We're still software developers at the end of the day, my delete key works better than ever.

Also maybe I'm a curmudgeon but I generally at least mark-one-eyeball all the tests it's generating and tell it when it's being extra dumb.

Weirdly enough I've found the opposite, using too smart a review model gets you overfit instead of benefit. The coding side can be more complex in a way that benefits from fable/astra, but the review side sometimes not so much.

FWIW gpt-6.1-sol and opus 5.5 are much, much less prone to this, if you include some coaching around complexity, YAGNI, that kind of thing

USFS uses it themselves so I don't know that they have much in the way of complaining to do. It's regarded as the least toxic mosquito management option, since it's quite specific and biological instead of pesticide.

https://www.epa.gov/mosquitocontrol/bti-mosquito-control

I'm slightly less pro-social in motivation because the bastards bite me preferentially and I'm fairly allergic. The pathogen-transmitting mosquito species, as far as I am concerned, could be magically extinct tomorrow and the world would be better off for it. There are lots of mosquitoes that don't drink blood, let them have the niche.

They are dropped from planes by the thousands or millions in Florida and other places, for what that's worth. Between the actual pesticide fogging (back in the day it was DDT) and bacterial controls, they're a big part of the reason that Florida and other southern states don't have malaria any more.

To be fair, my building inspector gave me a sheet entitled "what you can do without consulting the planning department" with like 8 pages and example deck/shed plans in it already.

But yeah I don't mess with wetlands.

If you throw some mosquito dunks with BT into any given water body small enough it doesn't have a name, nobody is going to come after you. I think it's probably a pretty prosocial thing to do, unless I'm missing something

Prior to air conditioning the venerable front porch was the socialization locale of choice. A lot of houses in my city have a massive one and it's kinda sad to me to not see many people sitting on them virtually ever

You would think so, but insurance has infected everything.

I often try to go cash pay, since in many cases at this point it's actually cheaper, absurdly enough, and get told that I can't even get an appointment that way. I actually got kicked off primary care at a regional (fairly large) practice because they couldn't take my insurance, would not bill out of network, didn't service uninsured people (!!), and would not take cash, credit, barter or trade.

Sadly the teaching hospital near me is both not especially competent and essentially the only teaching hospital in the state, plus even more bureaucratic about access than the other hospitals

Once you're at that point, just go to the top of the list for whatever specialty and fly out of state, or even to a different country. US healthcare is so absurdly expensive that flights and hotels kinda have to be cheaper than the care no matter how crazy you go.

> NDs have a hard time understanding the shared context of their audience, so they splay out all the detail to compensate

I would say this is fascinatingly true of AI as well, and probably one of the reasons I (and maybe other ND folks?) work with it so well, cringey slop word choice aside.

I was joking a few years ago that if the AI was trained on the internet, the corpus included massive overrepresentation with ND style writing.

Damn near a decade or more. We're a little under 4 years into the chatbot era.

This is something I've been noodling at, splitting a monorepo into a bunch of component projects. I think, unless they have access to something that is wildly outside the scope and scale of current API models, the downside of having a monorepo is the context pollution, and there's no way to escape that.

I'm not fully sold either way yet but I think splitting might actually be critical for moderating the blast radius of agent intervention.

The guidelines already cover this:

> Please don't use HN primarily for promotion. It's ok to post your own stuff part of the time, but the primary use of the site should be for curiosity.

https://news.ycombinator.com/newsguidelines.html

I assume that users who primarily/only post their own stuff and get flagged a lot reap the consequences, there's plenty of spam filtering on here thank goodness.

I've never frowned while paying a legal bill. They do good work, provide useful perspective, avoid problems before they start... reminds me of the better engineers I know, but for contracts and compliance.

Unfortunately did not even come close to capturing the magic of the original or TASpring / BAR though, sadly. I wanted to like it but it suffered from a lot of sequelitis.

I had a similarly useful analogy of the legal system emerge, in a similar way. I think there's a great analogy between policy in the legal system and policy in software engineering, and they have some awfully good (and very, very historical) ways to think about e.g. amending, repealing, and adjudicating things based on those policies.

I think definitionally the model can't know about the harness, since the model is at least a few months out of date. So every model must be behind it's harness or something really weird is going on.

Makes me wonder about doing some archaeology and trying out really old harnesses on modern models...

Here's a couple examples, I've seen others as well:

https://github.com/seb3773/ntfs-repair-rfc

https://github.com/Kotivskyi/screenshot-tool

I think it was more popular around the beginning of the year to mid-year

I expect by that point we'll have local models that can do a decent job, I would guess give it a decade and we'll be running custom accelerators that are smarter than current frontier models.

In the same way that only supercomputers used to have multiple processors and caches but it's now standard.

I've had some of the same ideas, I would love a full macro system on a dependently typed language. My stab at it earlier in January was not quite the right thing, but I do think this is the way. Weirdly enough concision is even more important in LLM context than in regular human context.

Smalltalk, in image based systems, absolutely. Erlang not as much, it's more of a "crash the actor and try again" than "rewind, pause, and handle errors in-system"

Same, I have scads of them (~8) and have never had connection issues with any of them. I made sure to only pull the battery tab and turn them on for the first time in their target location, maybe that makes a difference, and my zigbee network is quite dense, so there's no lack of mesh for them to connect to.

I'm 2 years in, I think, and I have never noticed any issues.

To be fair all of those islands are awesome and mega pedant time

> is Venice an island or is every block in Venice an island? Is Venice an archipelago?

Fun fact: Venice is, in fact, an island, and every block in venice is an island, and venice is an archipelago: https://maps.app.goo.gl/TfbkNvrLU55E9n9KA

The other aspect I think people are missing is that, for the right person, the choice isn't between a white collar office job and trades, it's between struggling in retail and service work with a (possibly bad) degree and no upward mobility, where your only escape is into either management or retraining.

For those people, trades are absolutely a better choice - it's not like becoming an HVAC dogs body prevents you from later going to school if you decide it's not for you.