HN user

mrighele

2,696 karma
Posts2
Comments812
View on HN

# Prompt Cache Misses

It globs your filesystem and re-reads AGENTS.md (injected in turn-0 system prompt) on every SSE turn. If you put a quick note in AGENTS.md to be read in the next session, you immediately force a full re-evaluation.

Personal favourite: it puts the current date in the turn-0 system prompt and re-evaluates every SSE turn. If you’re using OpenCode at midnight you get a full prompt cache miss.

I used opencode with deepseek v4 via openrouter during the weekend. Openrouter claims a cache hit rate of about 95%. Maybe it can be improved, but it doesn't seem broken.

I agree that it is bloated, and uses a lot of memory and cpu. But I don't remember seeing any bug during its usage, or at least not big enough to notice

So if that were the case then why isn't there some sore of daemon or system software service that tracks the expected life of the hardware?

Some linux distribution will regularly check SMART data from disk and warn you if some threshold has passed (like number of corrected errors, spin up/down, total hours etc).

That said, if we talk about consumer hardware in a home lab or the like, of the disks that I bought in the past 15 years, only one broke down, and it did fairly quickly so my suggestion is to use redundant storage, make backup and change the disk when it breaks down. (Enterprise disk and usage is another can of worms of course).

UI wise git sucks. The its commands are barely related to what they do, and that little makes sense only if you have in mind the underlying storage model (which is by itself a sign of bad UI).

Almost any other version control tool I have used in my life make more sense that git.

There are many reason for why git won, being able to use it without having to look up commands is not one of them.

I think jsonb_agg could be a solution. Requires cpu for decoding but I think it is compensated by the reduction of on the wire traffic. Queries are different , especially if you have several one-to-many joins so you have to do it at the framework level. I wonder if it would be feasible a Postgres extension that does this automatically for you

Having built a nas recently (luckily before ram/diskpocalypse) there seems to be a huge difference between making a small nas (say 4 disks) and a big one. With a small one you are essentially building a pc, and the differentiating factor is the software. Mobo, memory, case, PSU… buy consumer (cheap) stuff and you’re ok. Above that you have to start buying very expensive business oriented stuff,disk enclosure, better psu… the costs increase more that linearly , especially if you don’t live in the US and can easily get used stuff from eBay.

And that stuff is often not just more expensive, but uglier and noisier. I ended up making my own “enclosure”

No, I'm not conflicting the two issues. I'm fine with the ban I'm not fine with how it is being implemented, I thought the example would make it clear.

A ban shouldn't affect the adults (and this is the same for stuff like porn). If you think that kids should be restricted from a portion of the Internet, have the tools that they use check who they are... but adults should be able to use other tools.

I am fine with banning social media for minors.

What I find utterly wrong is the method, because this way you end up tracking also every adult.

A better way would be to legislate that minors need to use devices specifically designed for them, so that they can be kept under control, and forbid them to use regular devices.

My daughter cannot drive my car, and I don't need to insert my id card to use it, nor there is a facial recognition system to ensure that I am an adult.

Banning the kids from access social media sounds like just an excuse to be able to track adults.

Not OP, but I self host a few domains.

I was worried about not being able to send emails, but is seems that as long as you setup properly SPF/DKIM/DMARC you're fine. You may have problems if using a domestic address though.

For the configuration, the best bet is probably to use a product that makes it easy to configure the above three, there are a few alternatives around, like Stalwart [1] or docker-mailserver (which is little more that your postfix/dovecot/rspam combo packaged in a container) [2]

[1] https://github.com/stalwartlabs/stalwart

[2] https://github.com/docker-mailserver

"In the distant future, humans live in a computer-aided society and have forgotten the fundamentals of mathematics, including even the rudimentary skill of counting.

The Terrestrial Federation is at war with Deneb, and the war is conducted by long-range weapons controlled by computers which are expensive and hard to replace. Myron Aub, a low grade Technician, discovers how to reverse-engineer the principles of pencil-and-paper arithmetic by studying the workings of ancient computers which were programmed by human beings, before bootstrapping became the norm—a development which is later dubbed "Graphitics"." [1]

[1] https://en.wikipedia.org/wiki/The_Feeling_of_Power

You want to fix the 20% of causes that lead to 80% of the problem

Vote in person. Vote with ID. You solve 99% of the problems. If it works in the rest of the civilized world it can work in California too.

The excuses for not doing this seem terribly weak, and only fuel the ideas of conspiracy theorists.

and both the densely populated and the sparsely populated areas have unique problems to solve.

What are the unique problems to solve ? And I mean unique compared to the rest of the world where those issue are actually a non-issue.

The article exaggerates things quite a bit.

At the time of the Internet bubble, there were people pushing for more "free" usage of the Internet, and those that couldn't care less.

And it's not like the companies didn't want to take advantage of the Internet, but there was a mismatch between what the companies and the employees had in mind, which mostly boils down

* Employees want to use it to do their jobs and make their life easier

* Companies want to improve productivity, spend less and make more money.

There is some overlap of course, but the problem is where the two clashes.

I don't think today it is too much different. I see plenty of people using AI for what they care about, they complain when they are asked to use it for things they fear will make their life worse (like programmers that think they will have to pick up the pieces of vibe coding later on).

As a group, teenagers and young adults hate AI

I wonder what is their definition of AI. I haven't seen a single young person saying "I don't use chatgpt (or the like) because I hate AI". If else plenty of student have become dependent on it.

Those who won’t were doing it for the money. Those who continue are those who do it for passion, or those whose recipe is just a way to attract people to their business (e.g. kitchenware company). I don’t think it is necessarily bad, the quantity will decrease but the quality may even improve

If the user can search like in Youtube then how do you rank the results? That's also an algorithm.

Any ordering is an algorithm technically, so yes just "banning algorithm" doesn't work.

A better alternative could be "the algorithm must be public and reproducible by the user".

"Sort the posts of the people I follow in chronological order" you're good

"Sort the posts by the output of a blackbox trained on user data" too bad you're a publisher and are responsible for what people post.

Not OP, but I have been using borg backup [1] against Hetzner Storage Box [2]

Borg backup is a good tool in my opinion and has everything that I need (deduplication, compression, mountable snapshot.

Hetzner Storage Box is nothing fancy but good enough for a backup and is sensibly cheaper for the alternatives (I pay about 10 eur/month for 5TB of storage)

Before that I was using s3cmd [3] to backup on a S3 bucket.

[1] https://www.borgbackup.org/

[2] https://s3tools.org/s3cmd

[3] https://s3tools.org/s3cmd

When I pay money to buy food I don't need to ask how the shop is going to use that money: I gave money, I got food.

If I am going to donate money to a company/NGO that wants to buy food for poor people, of course I am interested in knowing how much of that money is going to salaries, how much into activities of sort, and how much in actually feeding people.

For a printer like an Epson MX80 an esp32 should be enough to share the printer on a raw TCP interface (AppSocket I think the protocol is named) on port 9100. It is supported by Windows and CUPS.

Very easy implementation as it essentially it just forwards the data to the printer. Since it's a raw interface you need the proper driver, but luckily Epson provides a Windows 10 driver for the Epson MX-80 (!) [1] CUPS doesn't have driver for the MX-80 but it has a number of generic Epson drivers and my guess is that one of those will work.

The most difficult part is probably the parallel interface (unless you have a printer with a serial interface in which case it will be much easier)

[1] https://epson.com/Support/Printers/Impact-Printers/MX-Series...

I think the MS-DOS installer disk put files in C:\DOS by convention but that was just a convention.

That assume that you where going to install the OS, which assumes that you had an hard drive :-). The original IBM PC didn't, and anyway MS-DOS didn't support folders until version 2.0.

On those old PCs you would boot your computer on a floppy drive with all the files on the root of a floppy, and execute your command there. There was not much to work with anyway, check the content of the boot floppy of MSDOS 1.0 [1].

And also, especially if you had a single floppy, you wouldn't even use it: to run your software you would boot a disk with a IO.SYS, MSDOS.SYS, COMMAND.COM and an AUTOEXEC.BAT that would start your favorite word processor (WordStar of course :-D ).

[1] https://www.youtube.com/watch?v=0-X7Thsn0pI