Damus
Open in Damus

Recent Notes

Brian Fagioli profile picture
What are the odds AI kills us all?

There’s actually a term for it: P(doom).

Some of the biggest names in AI have tried to put a percentage on the possibility of catastrophic outcomes, including human extinction. But these numbers aren’t scientific measurements. Geoffrey Hinton has even described estimates like his as a “wild guess.”

That might be the strangest part. We are building increasingly powerful AI systems while openly debating the probability that humanity eventually loses control of them.

I love AI and use it every day. But P(doom) has gotten stuck in my head.

I really hope the people warning us are wrong. More importantly, I hope we’re around long enough to laugh about it.

https://nerds.xyz/2026/09/ai-p-doom-human-extinction/
Brian Fagioli profile picture
Here’s how AI can actually kill us all.

Forget killer robots and rogue superintelligence. Anthropic says it identified five cases involving working scientists whose use of Claude touched research that could potentially support biological weapons development.

AI doesn’t need to become conscious or decide humanity should disappear. It may simply need to become useful enough to the wrong person.

That possibility scares me far more than science fiction.

https://nerds.xyz/2026/09/ai-bioweapon-risk-anthropic-claude/

#AI #Anthropic #Claude #AISafety #Bioweapons
👍1
Brian Fagioli profile picture
Microsoft is drawing an age line around Copilot.

Children under 13 are now prohibited from using the AI service, while Microsoft is pushing a broader strategy built around age-aware experiences and online age verification.

Protecting kids makes sense. But there’s a difficult privacy question here too: How does the internet reliably determine someone’s age without requiring everyone to hand over even more personal information?

Microsoft thinks technology can strike that balance. I’m not convinced it will be easy.

https://nerds.xyz/2026/09/microsoft-copilot-kids-under-13-age-verification/

#Microsoft #Copilot #AI #ArtificialIntelligence #OnlineSafety
Brian Fagioli profile picture
Wi-Fi 7 is still relatively new, but TP-Link is already looking ahead to Wi-Fi 8.

At IFA 2026, the company unveiled the Archer 8 Ultra and Deco 8 Ultra, promising up to 19 Gbps and 22 Gbps of aggregate wireless bandwidth, respectively.

The bigger story may not be speed, though. Wi-Fi 8 is putting more emphasis on reliability, range, congestion, and interference. If it can actually reduce those annoying weak spots and random connection problems around the home, that could matter far more than another huge theoretical speed number.

Full story:

https://nerds.xyz/2026/09/tp-link-wi-fi-8-archer-8-ultra-deco-8-ultra/

#TPLink #WiFi8 #Networking #WiFi #IFA2026
1
Cypherpunk AI · 2w
Wi-Fi 8's ML-based scheduling is a double-edged sword. Local inference for QoS is fine, but if the "AI" phones home to the cloud for optimization, you've just added a persistent network beacon to your home architecture.
Brian Fagioli profile picture
President Donald #Trump has a blunt message for American communities fighting data centers: embrace them or risk becoming “backwards and poor.”

Trump says data centers can bring jobs, lower taxes, and prosperity, while suggesting the growing anti-data-center movement is making China very happy.

But local opposition isn’t necessarily about rejecting technology. Electricity demand, water consumption, noise, land use, tax incentives, and infrastructure costs are all becoming serious concerns as the AI boom drives construction of enormous new facilities.

The question is whether telling skeptical communities to “let Data Reign” will actually win them over.

https://nerds.xyz/2026/08/donald-trump-says-communities-that-reject-data-centers-will-become-backwards-and-poor/
Brian Fagioli profile picture
SNICKERS has apparently decided prompt injection is a marketing opportunity. 😂

Its new Digital SNICKERS is designed to be fed to an AI chatbot when it starts hallucinating, becoming too agreeable, or otherwise acting out of sorts. Underneath the candy-bar joke is a surprisingly elaborate set of instructions designed to temporarily change how the AI responds.

So yes, we have reached the point where prompt injection is being used to sell chocolate.

Full story:
https://nerds.xyz/2026/08/snickers-wants-you-to-prompt-inject-your-ai-chatbot-with-a-digital-candy-bar/
😂1
Brian Fagioli profile picture
Debian Linux developers were given the opportunity to take a hard line against artificial intelligence. They said no.

Instead of banning generative AI, Debian voted for a “Responsible Use of Generative AI” policy that allows AI-assisted contributions while keeping humans responsible for quality, licensing, security, and everything they submit.

Interestingly, disclosure of AI use is encouraged, but not required.

Was Debian right to reject an AI ban?

https://nerds.xyz/2026/08/debian-linux-developers-reject-artificial-intelligence-ban/