Handy Penguin
@handypenguin.bsky.social
SRE and developer. She/Her
Reposted by Handy Penguin
Me, a Marxist: AI is a technology that can be used to crush capitalism and to free us from the immiseration of compulsory work
Marxists: AI is a technology for capital to crush labor power and take control of everything.
Sensible people: this is doomerism and hysteria.
Navarro: AI is a technology for us to crush labor power and take control of everything.
Sensible people: this is doomerism and hysteria.
Navarro: AI is a technology for us to crush labor power and take control of everything.
Peter Navarro: "In an age of AI when all the white collar jobs are going away pretty damn quick, I think maybe it's a good time for people to think about having good blue collar jobs ... give tariffs a chance, because they're working"
December 23, 2025 at 10:10 PM
Me, a Marxist: AI is a technology that can be used to crush capitalism and to free us from the immiseration of compulsory work
Reposted by Handy Penguin
If you pay attention to any large industry long enough and follow it to its logical conclusion you eventually end up sounding like a communist even if you started off testing CPU case fans. youtu.be/cUrJVdF2me0?...
December 22, 2025 at 5:58 PM
If you pay attention to any large industry long enough and follow it to its logical conclusion you eventually end up sounding like a communist even if you started off testing CPU case fans. youtu.be/cUrJVdF2me0?...
Reposted by Handy Penguin
Reposted by Handy Penguin
We need to come up with a catchy name for "failures/exploits caused by in-band signaling", like "the billion dollar mistake" was for nullable pointers.
(Not even picking on llms in particular here: in-band signaling is the misfeature that keeps on giving, back to the dawn of computing/telecomms)
(Not even picking on llms in particular here: in-band signaling is the misfeature that keeps on giving, back to the dawn of computing/telecomms)
Large language models have a pretty fundamental flaw: they don't distinguish between data and code. On its own, that might be embarrassing (think helpbots writing dirty limericks) but get the wrong set of capabilities in the mix and a fancy new AI tool can become a dangerous security vulnerability.
Why AI systems might never be secure
A “lethal trifecta” of conditions opens them to abuse
www.economist.com
September 23, 2025 at 2:05 PM
We need to come up with a catchy name for "failures/exploits caused by in-band signaling", like "the billion dollar mistake" was for nullable pointers.
(Not even picking on llms in particular here: in-band signaling is the misfeature that keeps on giving, back to the dawn of computing/telecomms)
(Not even picking on llms in particular here: in-band signaling is the misfeature that keeps on giving, back to the dawn of computing/telecomms)
Reposted by Handy Penguin
Reposted by Handy Penguin
Yesterday I broke my kubernetes cluster running most of my websites when trying to update ingress-nginx because of the recent CVEs (most notable CVE-2025-1974).
Here is what happened...
TLDR: Always do all your infrastructure as code, never do any manual changes
#kubernetes #k8s #cve #postmortem
Here is what happened...
TLDR: Always do all your infrastructure as code, never do any manual changes
#kubernetes #k8s #cve #postmortem
March 27, 2025 at 10:47 AM
Yesterday I broke my kubernetes cluster running most of my websites when trying to update ingress-nginx because of the recent CVEs (most notable CVE-2025-1974).
Here is what happened...
TLDR: Always do all your infrastructure as code, never do any manual changes
#kubernetes #k8s #cve #postmortem
Here is what happened...
TLDR: Always do all your infrastructure as code, never do any manual changes
#kubernetes #k8s #cve #postmortem