176 Followers
400 Following
3.7K Posts
Some #openbsd, some neuroscience, product security @ #Qualcomm, he/him, opinions are mine.
I don't think Anthropic really meant it this way but what an amazing demonstration of why generative AI's energy consumption is only getting worse, and dirtier. "Improvements" only come from nesting chatbots within chatbots, rather than any fundamental smarter design. It's wasteful at every single level. It's Bitcoin x 1000
Indeed, here is a good explanation on just HOW fucked up it was: bsky.app/profile/ryan... There is no way "An AI did it!" and it is journalistic malpractice to accept that story. "Tank Day", the numerology of the tumbler's volume, the slogan, etc... This was "hurr-durr we are fascists".

RE: https://bsky.app/profile/did:plc:ckrrdhtzt2dh3bfpaazbrgkw/post/3mmi5yrbqps25

First off, there’s the fact that you’re using the damn thing in the first place. That’s really not the flex you think it is.

Second, it clearly communicates to me that you don’t respect me. I put my own human time into looking at the code and giving feedback. You need to match energy. If you just give me a pile of LLM spew, that means you didn’t value my time and energy enough to respond in kind. This is basic human respect.

Third, it’s not Artificial Intelligence, it’s Automated Mansplaining. I’m an expert in my field. Mansplaining at me about my own field AND being obviously wrong about it is REALLY not the look you’re going for. (Like, the LLM “explanation” I got thrown at me this morning wasn’t even plausible. It contradicted itself in the first two sentences. It was 8 paragraphs.)

RE: https://hachyderm.io/@inthehands/116695414416647626

Oh! This is such a good framing for a misunderstanding about the role of code I don't think I've seen before.

Truth: Code is an encoding of intent. As with any kind of instruction, brevity is power. This, we know.

But, the fallacy: Code is a sort of fuel, made of symbols, that you pour into a computer to make it do stuff. Common sense tells us: more is better!

I can totally see how you’d start at one and arrive at the other. And it happens! So often! All the time! For our whole lives! Agh!

"The values described in Claude’s constitution sound very nice, but that hardly matters; it’s dishonest to suggest that Claude is capable of moral reasoning, because it’s not."

-Ted Chiang, No, Artificial Intelligence Is Not Conscious
https://www.theatlantic.com/philosophy/2026/06/no-artificial-intelligence-is-not-conscious/687378/

No, Artificial Intelligence Is Not Conscious

Taken to its logical conclusion, this line of thinking is absurd—and damning.

The Atlantic

For anyone in/around #PortlandMaine, I’m hosting a #Pride karaoke brunch in my enchanted garden next Saturday 6/13 from 11-1pm!

All are welcome, karaoke optional, feast is potluck.

Come to Peaks Island & celebrate with us!

(Take the 10:15 ferry, there’ll be signs with my address on them, or ask anyone)

Trump regime takes aim at crucial ocean monitoring network. Ships will be dispatched this month to remove the more than 900 deep-sea instruments that comprise the network, one of the world’s most trusted sources of climate and oceanic data. https://www.scientificamerican.com/article/trump-administration-takes-aim-at-crucial-ocean-monitoring-network/
Trump administration takes aim at crucial ocean monitoring network

The Ocean Observatories Initiative has been collecting data on physical, chemical, geological and biological conditions in the Atlantic and Pacific Oceans for the past decade

Scientific American

I found that crafted #MeshCore node names could compromise #HomeAssistant instances running meshcore-card, with an XSS leading to remote root access on the HA host. An attacker could then access anything controlled or visible through Home Assistant. The attacker doesn't need to be near the target, as MeshCore advertisements are repeated over the mesh, which is dense in NL.

This also affects around 20 public MeshCore analyzer websites. Some of those run CoreScope, where it looks like a vibecoding bot broke the XSS filter while hallucinating a bugfix. The analyzers are mostly public data though. In addition, the less popular MeshCore-Home-Assistant-Panel-v2 is likely also affected, but I was unable to make contact with the maintainer.

MeshCore node names are only 32 bytes, and each rendered in a different place in the page, so I had to be creative to run a more substantial payload. I found a way with three node names using an iframe feature I never heard of before.

https://mxsasha.eu/posts/meshcore-xss-home-assistant/

Rooting Home Assistant through MeshCore: XSS attacks with a LoRa node name

A crafted MeshCore node name could compromise any Home Assistant instance running meshcore-card as soon as someone viewed a dashboard with that card. MeshCore …

PBS did a documentary showing infrared/thermal drone footage of the massive gas turbine plants being built next to AI datacenters.

As I’ve pointed out in the past, despite greenwashing these massive datacenter campuses being built by so-called “hyperscalers” are being primarily fueled by on-site gas turbines, which in addition to climate crisis-causing CO2 also emit high amounts of NOx and particulate matter.

https://youtu.be/5p426fSlYH4

#climatecrisis

We Saw What AI Data Centers Don't Want You to See

YouTube

An internal Microsoft strategy document says that the plan for its just-announced “Scout” personal assistant AI is to “make people addicted” to the tool before rolling out additional functionality.

The document notes that “security and compliance” are important things to figure out moving forward.

https://www.404media.co/microsoft-wants-to-make-people-addicted-to-scout-its-new-ai-assistant-internal-documents-reveal/

Microsoft Wants to 'Make People Addicted' to its New AI Assistant, Internal Documents Reveal

Planning documents for "Scout" say the plan is to "make people addicted" to the tool before adding new features.

404 Media