Timeline for pleroma list by gnusocialjp
gnusocialjp-
Embed this notice
Roy Tam (roytam1@miniwa.moe)'s status on Tuesday, 19-Aug-2025 01:45:42 JST 
Roy Tam
Cut a little piece of kapton tape and try to mask pin 6&8 to see if this will work or not, but I don't have time for this right now.
Embed this notice
-
Embed this notice
Roy Tam (roytam1@miniwa.moe)'s status on Tuesday, 19-Aug-2025 01:43:35 JST
Roy Tam
Cut a little piece of kapton tape and try to mask pin 6&8 to see if this will work or not, but I don't have time for this right now.
Embed this notice
Embed this notice
Embed this notice
Embed this notice
-
Embed this notice
Roy Tam (roytam1@miniwa.moe)'s status on Monday, 18-Aug-2025 22:26:25 JST
Roy Tam
@mawei 能讓人每話給好評的番組不多，這部是其中之一。 -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 22:04:10 JST 
翠星石
@fwaggle @alison Hotmail was originally nonfreeBSD, which made working around bugs easy, as the maintainers could just quickly restart the server.
Then microsoft ordered that Hotmail be moved onto windows and that made debugging suck due to windows brokenness and how rebooting windows took like 15 minutes. -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 22:04:09 JST
翠星石
@mjg59 Secure boot is a plot by microsoft to kill GNU - they do not want people to run the GNU GRUB OS. -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 21:59:00 JST
翠星石
@veinglory Yes, people are obsessively trying to shove LLM's into everything - even though practically the only task those can perform is a convincing chatbot (that is wrong most of the time, but confidently written).
There are also LLM's that are used to copy software, except with combinations of the input and the copyright information stripped off (most software is poorly written and the result is always bad code outputted).
Unrelated to LLMs, there are machine learning tools that are actually useful.
One example is a tool that can flag possible cancer in scans to assist with cancer diagnosis - which does not replace the clinician - rather supplements them.
Such tool was exclusively trained on medical scans with accurate prognoses and was carefully done - it's very easy to instead train on some irrelevant detail - for example a tank spotting ML program that turned out to be a daylight detector. -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 21:51:49 JST
翠星石
@veinglory It appears LLM software is a "VM" of sorts, with the models running as proprietary blobs.
The "VM" can be free software, but that doesn't magically make the binary blobs that get executed free. -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 21:41:57 JST
翠星石
@newt @redstarfish @mangeurdenuage Every windows users has been forced to accept an agreement to not share copies of windows.
microsoft could and would force every windows used in every country to sign an NDA. -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 21:29:12 JST
翠星石
@veinglory >i want to understand how it works. I'm looking for resources for that.
That's the thing - nobody knows how it actually works.
>Also of the source code is availabl
The source code is not available - all you get from training is binary blobs - understanding the operation of such binary blobs is extremely difficult reverse engineering. -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 21:26:50 JST
翠星石
@N33R I don't see how it's a horror that messages are routed through the Church of Emacs - as that's plenty reliable. -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 21:23:32 JST
翠星石
@deadheat @libreoffice @takao -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 21:21:34 JST
翠星石
@veinglory It came to me in a dream.
I looked at LLM models and realized that those are inscrutable binary blobs. -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 21:17:57 JST
翠星石
@veinglory You can continue training the model and hope its gets better instead of worse - if you want to make a specific change - you would need to reverse engineer the object code into the source code - but that may not be possible. -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 21:18:01 JST
翠星石
@veinglory GNU/JIHAD AGAINST "OPEN SOURCE"!!!!!!!
With LLMs, the training result is object code and not source code, therefore it's proprietary software, even if it's under a free license. -
Embed this notice
翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 18-Aug-2025 21:11:26 JST
翠星石
@CaptainButthurt @BasedLunatic @zer0unplanned You have: 40 rods/hogshead
You want: litres/100km
reciprocal conversion
* 118548.15
All GNU social JP content and data are available under the 
