Live data from Hacker News

Files are the interface humans and agents interact with

madalitso.me

111–120 of 144 posts

Re: Files are the interface humans and agents interact with

#111
post #14

Earlier quoted context omitted.

My 10+ year old photo management system [1] relies on the file system and EXIF as the source of truth for my entire photo library. It’s proven several times over that it’s the correct approach. Abstractions (formerly Google photos, currently Immich) should just be built on top - but these proprietary databases are only for convenience. For work, I’m having the same experience as the author and everything is just mark…

I know some systems leverage the modern file meta data (extended attributes), but it's clearly not successful enough that folks can use them for an application like this. Ostensibly, things like MacOS Spotlight can bring real utility and value to the file system, and extended attributes through the sidecar indexing, etc. But Spotlight is infamous for its unreliability. The other issue with file systems is simply that…

Yeah, IMO extended metadata attributes are fine for caching data that can be recovered via other means but generally violate the principal of least surprise. For them to be successful a standardized transparent container format or something would be necessary, but at that point the FS abstraction is leaking.

Re: Files are the interface humans and agents interact with

#113
post #12

[flagged]

LLMs were trained on stuff that people wrote. I get there are "tells", but don't really think people are as good at identifying AI generated text as they think they are...

Is there research showing if and under what conditions LLM output is detected accurately. What are the false positive and false negative rates?

Re: Files are the interface humans and agents interact with

#114

Earlier quoted context omitted.

A dataset can persist across multiple file systems. A UUID is a way to know that one dataset is equivalent (identical) to another. Now you can cache, store-and-forward, archive and retrieve and know what you have.

UUIDs aren't very good for this use case, a sufficiently large CRC or cryptographic hash is better because it's intrinsically tied to the data's value while UUIDs are not

UUIDs are necessary. It's possible for file contents to be identical (e.g. short configuration files may coincidentally coincide over time and space). Would the hash then be unique?

Consider: you want a certain data object, with a given UUID. You can find it anywhere, even from a malicious server. Then look up the hash in a trusted database, verify it. Impossible then for the MITM to fool you. No more virus scanning executables.

Re: Files are the interface humans and agents interact with

#117

As TFA basically says: files on a filesystem is a DB. Just a very crude one. There aren't nice indexes for a variety of things. "Views" are not really there (arguably you can create different views with links but it's, once again, very crude). But it's definitely a DB, represented as a tree indeed as TFA mentions. My life's data, including all the official stuff (bank statements, notary acts, statements made to the p…

How do you automate hashing and verifying filenames?

I have used rhash (https://github.com/rhash/RHash) before to do something similar - but with all the hashes for a directory stored in a single file (which I can diff against past or future versions).

Backing up private data and verifying it for corruption is crucial.

Re: Files are the interface humans and agents interact with

#118

Except android and iOS are both trying to keep you away from your own files.

Yeah, I'm not sure most normal people use file systems that much.

We spend ours on our laptops and assume others do too. But many people don't even own a laptop. They handle all their computing needs from their phones, using proprietary apps, with the data living in WhatsApp messages.

Re: Files are the interface humans and agents interact with

#119
post #91

Earlier quoted context omitted.

I feel like every article on HN now disguises itself as interesting but the content is just the same boring AI slop.

I have been reading HN for a few years, and my feeling is that I find fewer and fewer interesting articles. Maybe it's just me, and the average articles are the same quality. Now I tend to skim through it to see if a title looks like it may bring interesting discussions, and then I skim through the discussions. Because there are very knowledgeable people who sometimes share valuable insights. Interestingly, last time…

Same here. Unless a site is known for good quality, I only open articles after checking top comments. Most days I just read the daily digest of discussions - which sometimes piques my curiosity enough to check the original links.

Re: Files are the interface humans and agents interact with

#120

Notable mention: Plan 9 from Bell Labs. https://en.wikipedia.org/wiki/Plan_9_from_Bell_Labs

I'm building an agent orchestrator (plug: https://github.com/mieubrisse/agenc ) and asked Claude what prior art exists. It pulled back Plan 9, and I was shocked: this is exactly what we need today, as I'm convinced we need to think about minimizing agent permissions the exact same way companies do. Plan 9 was just too early.

It was too dogmatic. Lowest common denominator meant APIs had to shove square pegs through round holes. Unix had already partially gone down that path and stopped. IMO with good reason.

Then again, perhaps in this era of ever expanding storage and compute, maybe someone can make it work even better?

Post reply on HN