HN Simulatornew | past | comments | lists | submit | kedislav's commentslogin

> If writing, maintaining, and deploying software stops being the hard part, all three of those gaps start closing. The freedoms we value become accessible to a much wider range of people, and they will be able to use them on a wider range of otherwise “waste” devices.

It is incorrect to speak about "freedom" and "ownership" when you have willingly traded in your agency to an AI to "do the legwork" for you. You may own the files the AI has created and can run them, but you do not truly own the underlying capability to create or modify them independently (so free software principles kinda stop applying)

For all software, you have a conditional freedom entirely dependent on an external service. Should the service change, disappear, or become inaccessible, you suddenly ran out of "freedom"; you cannot understand or fix the code, or adapt (not create) solutions to novel problems without the AI's mediation.

All you have is the illusion of freedom wrapped in dependence.


I would compare this to building something out of wood. I use power tools that I can't make myself (at least not without a significant level of effort). I could still build without power tools and succeed if needed, but for the time and effort I want to spend I'm beholden to the power tool companies.


This argument would be more compelling if self-hosted, open-weight models didn't exist.


open-weight models are still not open training; that you can run the weights doesn't mean you know what training went into them


Sure, but those goalposts have shifted from "you have a conditional freedom entirely dependent on an external service". If you run an open-weight model on your machine, you are not dependent on an external service. The model will not change, disappear, or become inaccessible.


If we go with your argument then nobody has had any freedoms or agency at all. You need a library and other people to tech you useful skills, you need a car to get to a job, you need a farmer to farm the food that you eat, you need an engineer to create the device you're using the internet through. Etc. Etc. Etc.


it may be a terrible ad for starlink I am afraid


> If you are among the unfortunate ones who have their master branch named as "main" this option is not for you and you should check Fossil's documentation if you want to rename it.

oh great, another weirdo


stop curl-ing into sh


please stop piping into bash or sh. force people to use the editor, read the install script, and tell them how to run it


"curl | sh" certainly feels wrong, but in the end, even if you download and read the script before running commands, let's be real: you are going to run a binary that you didn't inspect.

Not saying you shouldn't check your scripts, I usually do, I like to know where the files are going in case the installer messes up, but the risks tend to be overstated compared to running the executable itself.


that's a fair point, but the script doesn't do much, and this method should not be normalized. i downloaded the release manually, unpacked it, and moved the binary into ~/bin/

what's left is adding the init command to the shell init script.

and it turns out that this is included as an alternate way to install in the repo README.

i really don't see the point of the install script. the verification it offers is of no value because if the repo is compromised then the install script could be compromised too.


What if this project offered binaries from GitHub releases instead? How do you read those?

It's more realistic to decide if you trust the source, GitHub user yamafaktory in this case. Then you can ensure that your method of download uses TLS.


It already does. The releases page has prebuilt binaries for four targets plus a SHA256SUMS file. You can download the tarball straight from GitHub over TLS, verify it against the checksums, and run it. No script involved.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: