Remote
Stefan Prandl
@redezem@aus.social
mastodon 4.7.2Your local Perth Cyber security guy! CTO of Hyprfire, and computer sorcerer.
Likes: infosec, networks, people, and coffee
Pros: he/they
PGP Key: e66fd000b1e871cd
0 Followers
0 Following
27 Posts
Joined October 31, 2022
LinkedIn:
What do you mean “based on consumption of whole ingredients”?! Who *ate* the green tea bags, tea box?! Who hurt you like this?!
Replying to
@danbjoseph@floss.social My shortlist is something like:
- Here's why they're not what they seem/ so compelling
- Here's how usage of these models will hollow out your organisational knowledge base (and what will happen to the org if that is allowed to happen)
- Here's how the economics of the thing preclude the long term strategy being "Oh we just become wholly reliant on it"
- Also, here's why (if you're a non-profit interested or positioned around a common good) this is fundamentally at odds with the moral standing of the organisation and the costs that would be involved from a reputational viewpoint.
I actually did a keynote along the lines of 1, 3, and 4 last year. Lots of fun.
Replying to
@blursedtexts@jorts.horse was interested enough that I found more context: https://www.livescience.com/nuclear-bunker-cannibal-ants.html
Tl;dr: they built an escape method for the ants and they all went back home ❤️
I have an insight for yall:
The #EU is the #Kubernetes of states.
I will not be elaborating further. Have a good weekend.
Replying to
@cR0w@infosec.exchange So excited to see how Iran will beat this offer. Maybe 25% but discounted to 15% if you subscribe to their patreon? I guess we’ll have to see.
RE: @ComicContext@mstdn.social
Any time I’m trying to debug networking problems remotely
Replying to
@peter@thepit.social love the “Ed’s argument is wrong, allow me to make Ed’s argument in contrast” vibes in this article.
Replying to
@uberduck @lawyersgunsnmoney @hacks4pancakes
Looks like this came second hand from ex-CIA analyst Larry Johnson. Not seeing any supporting evidence though, so maybe maybe not.
https://news.meaww.com/fact-check-did-trump-try-to-use-nuclear-codes-on-iran-but-was-stopped-by-dan-caine
Replying to
@wezm@mastodon.decentralised.social HeY gUyS, hAvE yOu HeArD aBoUt OuR nEw Ai NaNo BaNaNa?
Replying to
@glyph@mastodon.social I don’t like it, but unless things go *terribly wrong* for LLM companies (and that’s a real possibility not gonna lie), I can’t help but think that market forces are going to make most of corporate softeng a small team of humans shepherding swarms of agents.
Not cause it’s good, but because it will let them have less staff whilst passing off a chunk of the cost of development as opex.
Oh what, #Dragos is moving in across from us? I mean it’s a little small….
Tell you what if I see anyone here I’ll say hi, would be nice to have anyone that gets OT network monitoring to chat to over lunch.
Replying to
@ludicity@mastodon.sprawl.club bonus points: it’s the Apache server from Ubuntu 18.04.
Wow sophisticated technology of tomorrow running on the server of last decade, incredible.
Lol. So ima just say: that’s not a solution.
Today in commits:
[main xxxxxxx] Postgres 18 hates freedom, have rebased on communism
That is all.
So just read this about how #genai can't just up and recursively improve itself with it's own outputs:
https://arxiv.org/pdf/2601.05280v2
Neat, neat, still doesn't mean you can't have some autonomous training system with external validation of some kind... which is how I suspect they've been getting these things to agent better (with code agents anyway).
However, what if you built the #Matrix not to act as a power supply, but as a source of human input for training data? Something to think about 😅
Replying to
@lawyersgunsnmoney @uberduck @hacks4pancakes I pretty much posted this and then read you figuring it out 😂 sorry about that.
Replying to
@_elena@mastodon.social I’ve never felt comfortable being active in other social spaces, the interactions tend to be really highly strung, you’re a hero and a villain at the same time due to the algorithm feed. This place is so much more chill, and the conversations are way more thoughtful and respectful.
Replying to
Wow
https://www.ietf.org/archive/id/draft-thain-ipv8-00.html#name-ipv4-representation-in-ipv8
This is possibly the worst internet idea I've read in a while.
Has the author considered, possibly... just using IPv6?
If you’re like me and trying to get #xscreensaver working on #macos27, this patch may help:
https://cdn.redezria.net/xscreensaver-maze-macos27.patch.gz
Basically MacOS 27 does some absolutely ass things to the window initialisation/deinitialisation that you have to account for. Thankfully because it’s just the window itself this *should* work on any of the screensavers with a similar edit… I’ve not done that because I only use Maze on my laptop as nothing has ever caught my attention as much before or since 😂 And this was a fun side quest.
I can’t help you if you use one of those desktop-image-based screensavers, I would bet there’s some new cursed nightmares down there for you to solve. Good luck I guess!
Replying to
@glyph@mastodon.social Yeah I agree with this. My suspicion is that there's a lot of structural code that doesn't vary much between programs and exists mostly to let whatever cool thing you're working on do whatever it is it's doing. I think that abstracts too, there's a lot of really boilerplate applications as part of whatever stack that people are using that really only exist to help them do whatever cool thing they want to do.
You *don't* want LLM stuff to write that "cool" part though.
I think it'll take a loooong time for the effects to rock up however, as let's be fair a lot of the apps out there are shovelware as far as the code is concerned. Out here in cybersecurity my favourite game to play at trade shows is "Actual Product or Excel With Extra Steps?". Nearly everything is Excel with Extra Steps, which I suspect you can LLM out without any real issues. It's not like they're thinking hard about product anyway.
Doing something *new* and *interesting* though...