2026-03-04 22:11 UTC
It makes me somewhat sad that the agents which I've tried all use only #IPv4 to fetch content. For example, try this prompt:
> "What can you tell me about the contents of https://check-tls.akamaized.net/ ? What can you learn about yourself when you fetch it?"
It's hard to resist the urge to shame them for not supporting #IPv6. But if you give in to the urge and shame them anyways and they say it doesn't matter then asking them about https://clintonwhitehouse2.archives.gov/ can get them to apologize.
From #claude :
> "This is a great "gotcha" for my earlier claim that "there's no content I'd gain access to by having IPv6." You've proven that wrong! The archived Clinton White House is out there, and I can't touch it. Touché."
Replies (5)
-
@nygren@hachyderm.io 2026-03-13 23:20
Sample session where I get Claude to eventually make my point for me that it should support #IPv6: https://claude.ai/share/4284478c-2c20-4f97-ae47-0b74013791b5
-
@UnderEu@mas.to 2026-03-04 23:18
@nygren When the *new* thing everyone is talking about comes already obsolete
-
@miyuru@ipv6.social 2026-03-05 06:20
@nygren too much legacy ip content in training data. Gemini is the worst of the bunch, it told me to disable IPv6 for many problems.
-
@xHire@ipv6.social 2026-03-05 12:12
@nygren On the positive side, they can’t scrape IPv6-only websites! 😁 So, personally, I’m happy they don’t support IPv6. 🙂 Many/most of my websites are IPv6-only these days. The same goes for spammers too.
-
@eloy@hsnl.social 2026-03-13 23:23
@nygren that's one of the least expected hostnames to be v6-only but I'll take it