SchemaLoad

Security cameras have been internet connected for a lot longer than we had this kind of system where they all have computer vision running and tagging everything in a massive database.

They were always internet connected and mass surveillance. It's the addition of search that changed things.

In my experience it's always the permissions system where messy real world business has to be translated in to SQL that doesn't lock up the system when some program admin added to 200 teams tries to load the index page.

"What can this user see" has always been a very complex question in every crud app I've worked on.

This is what the managers at my company think. But it ends up generating a wall of slop. I always end up scrolling to the bottom and looking for the original customer email or product request that was used to generate the ticket above.

Most vibeslop these days is superficially good in that it has unit tests, compiles, passes the style guide, but it's bad on a macro level in that it makes the system more complicated, ignores the wider design of the app, implements what the prompter asked for and not what the app actually needed, etc.

I'm seeing a bad pattern of vibeslop being used to solve the immediate complaint of the user without stepping back and reconsidering things from a product perspective. Just add another conditional statement to make this very specific scenario the user complained about work the way they want.

A single bad dev with AI can overwhelm multiple good devs trying to hold back the tidal wave of slop.

I feel like we should be doing something like requiring paying overtime for over 30 hours a week. The complication here is countries are competing with each other as well as internally. That said, most jobs are local, so cheap construction workers or child care workers overseas don't compete with local ones.

This combined with doing something about housing prices seems like it could work.

This is still less safe and less convenient than sandboxing. I want them to be able to run whatever command they want inside the sandbox even if it's traditionally dangerous (overwrites files, etc) because it can't touch anything that matters and I can revert any changes to the code myself.

If anything the only thing you'd have to have reviewed is sandbox escape attempts.

I think the simplest answer that probably accounts for most of it is just that employees all have overhead, 20 people working full time is more efficient than 40 people working half time. Less HR work, less moving context between people, less office space, less laptops, etc.

So I can see how naturally the system has prioritised this way of working, but despite the efficiency I think we need to do something to normalise 20-30 hours as full time. Give people less iphones and more hours in their life.

These models generate the most unreadable bash commands I've ever seen. Utilising every single option available and piping the result through multiple layers of regex and junk.

The mental load of having to actually review these commands beyond the most surface level glance is too much.

Sandboxing seems like the obvious solution. These tools prompt me to review an absurd number of very complex bash commands where there is realistically no way to carefully review them. I just want to put the thing in a box with the code and let it run any number of commands on the code without prompting.

The dream society would be where we all just work less hours. Everyone can get a job but doesn't have to burn every workable hour of their life.

Even with the latest models today, the hallucination rate is absurdly high on anything deeper than surface level knowledge or something that can be directly scraped from reddit.

And you notice when it's a topic you know well or something like software where you can immediately tell the options it's giving you don't exist on the page. Leading to the amusing statement "LLMs are bad at what I do but great at everything else".

Active investing was the source of the problem here. Retail investors were buying leveraged bets on Samsung and SK Hynix which has wiped out their investment now it's crashed.

Had they passively invested in indexes over a long time they would be up still.

It's already burst in South Korea when China showed they can make memory chips too. And it would have burst by now in the US if OpenAI and Anthropic were public.

I wonder if the realised the entire website has become the most unbearable AI slop imaginable and if they don't do anything about it they won't have any real humans reading posts, just agents trying to advertise to each other.

Having out of date public book boxes isn't really that big of a deal, at least compared to the massive amounts of broken entries, duplicates, and incorrectly licensed data that would end up from free and unrestricted bulk imports.

As it is today you can largely expect OSM POI data to be incomplete or out of date unless you have someone particularly keen on keeping your local area up to date. OSM is best used as a background map layer for your own custom applications, like embedded maps in apps showing your own data overlayed. Rather than an alternative to the google maps app.

>A lot of words, but essentially says nothing useful.

This is essentially the state of all blog posts now. As well as much code. Lots of content, not a lot of value.

Triggering an AI crash which the whole US economy is dependent on.

Whatever Elon Musk, Jeff Bezos and Peter Thiel are calling for, the public instinctually wants the opposite.

They also just work better and aren't substantially priced different to the equivalent android option.

Personally I would much rather Google delete my files than hand them to an attacker. I can restore them from a backup. Tech companies need to do better at encouraging users to keep their own local backup alongside the cloud one.

If you are a delivery company or such paying for mapbox you can build your own solution and get better quality than the open source general purpose apps.

OSM is actually perfectly sufficient if you just want street and address data. This stuff gets pulled in from official government sources so it's all present and up to date.

Where it's lacking is business info since this has to be entered manually and changes much more often than streets or buildings change.

This is sadly not a rare type of person. I'm worried my parents fit this description, they drive everywhere and work an office job. I'd guess on average they get 0 minutes of exercise a day.

I think people get this image in their head that someone who doesn't exercise ever is this comically fat unemployed person when in reality it's the average office worker who isn't fitness minded. A good chunk of HN users wouldn't be getting 15 minutes of exercise a day.

What if we could just lay in a pod 24/7 taking peptides and nutrient supplements while having an endless stream of instagram reals beamed in to our eyeballs on our Meta Raybans. That way we would never have to do pesky things like go outside or move.

A lot of people who use a car to get around will spend most days doing literally no actual exercise. For someone who lives in a more walkable area, 22 minutes of exercise is just living live normally without actively "exercising".

I avoid sleep on my desktop because it often wakes up in a weird broken state, particularly with audio glitching. I avoid sleep on my Steamdeck because the battery slowly drains in sleep.

On my Macbook I always use sleep since there are no downsides.

On Windows and Linux it is better to just turn it off because sleep is broken on most hardware. But on Mac there is next to no advantage to doing that. MacOS sleep is also more of a hybrid between sleep and hibernate where it will after enough time offload the ram to the SSD and fully power down so the machine can be in "sleep" without any power drain.

Agreed, I'm waiting for the time when 48GB+ ram is just the standard that computers come with rather than being the absolute top tier option. It just doesn't make sense to spend extra on a local AI computer right now when the same money would last for a decade of API pricing.

What I have been doing seems a bit different to what's described, but I always make sure to define how to know the task is done so the agent doesn't quit early. Usually this means telling it to to run the tests and type checks to ensure it runs without errors.

Otherwise they often do a first pass looks good enough but it doesn't actually work.